In regulated electric utilities, there are three distinct numbers that all claim to measure the value of the same physical asset — the wires, poles, substations, transformers, generation plants, and hydroelectric dams that make up an electric utility system. These three numbers almost never agree, and understanding why they disagree is the entire intellectual basis of this paper.

The first number is book value: what the utility spent building or acquiring the asset, minus everything that has been depreciated away over the intervening years. A transmission line built in 1970 for $50 million, depreciated over 40 years, might have a net book value of $5 million in 2010 and zero in 2020. The accounting has systematically consumed the asset on paper while the wire continues carrying electricity through the same right-of-way.

The second number is the rate base: the regulatory asset value on which the utility earns its allowed return. This is similar to book value but not identical — regulators allow certain deferred costs, construction work in progress, and regulatory assets to be included, and certain items to be excluded. The rate base is the denominator on which the state public utility commission authorises the utility to earn, typically, 9% to 11% return on equity. In most cases, the rate base is somewhat higher than net book value but still reflects the historic cost paradigm.

The third number is replacement cost: what it would cost to build the identical asset from scratch today, using current labour costs, current steel and copper prices, current permitting and environmental compliance requirements, and current right-of-way acquisition costs. For infrastructure built in the 1950s through 1980s — during an era of cheap steel, cheap labour, and minimal environmental permitting — the replacement cost in 2026 can be 5 to 20 times the current net book value.

The owner’s analysis asks one question: which of these three numbers is relevant for deciding what the asset is worth to someone who wants to buy it today? The answer is obvious once stated: replacement cost. Because if a competitor wanted to provide the same service by building new infrastructure, they would have to spend the replacement cost. The existing asset’s competitive value — its economic moat — is defined by how much cheaper it is than rebuilding from scratch. No amount of accounting depreciation changes the cost of the steel and labour required to replicate the wire in the ground.

This paper is about the gap between what the accounting records, what the financial model produces, and what the replacement cost reveals.

Subscribe now

PART I: THE REGULATED UTILITY BUSINESS MODEL, FROM FIRST PRINCIPLES

A Monopoly with a Social Contract

An electric utility is not a normal business. It is a regulated monopoly operating under a social contract that is centuries old and deeply embedded in American law, politics, and culture. The contract has two sides. The utility agrees to serve every customer in its territory who asks for service, at rates set by a regulator, without the ability to raise prices unilaterally or refuse service to unprofitable customers. In exchange, the regulator agrees to set rates high enough to allow the utility to earn a reasonable return on its invested capital — the rate base — and to recover all prudently incurred operating costs. The regulator also guarantees the utility’s service territory against competition: no other company can string wires and serve the same customers.

This arrangement sounds restrictive, but it creates an asset class with characteristics that are genuinely unusual in the investment universe. The revenue is not cyclical — people use electricity regardless of the economic cycle, and the regulator must allow the utility to recover its costs. The customer base is captive — there is literally nowhere else for a residential customer in Columbus, Ohio or rural Louisiana to buy electricity. The rate of return is guaranteed by law — a regulator who routinely allows utilities to earn below their cost of capital will eventually face service deterioration, utility bankruptcy, and the political consequences of failing infrastructure. And the physical assets — the grid infrastructure — are monopolistic in the most profound sense: they cannot be replicated by a competitor even if that competitor had unlimited capital, because the right-of-way for a second set of transmission lines simply does not exist in most markets.

The income-generating asset of this business is the rate base — the sum of invested infrastructure on which the regulator allows the utility to earn its authorised return on equity. Every dollar the utility spends building or upgrading infrastructure, if approved by the regulator as prudently incurred, becomes part of the rate base. The rate base earns the allowed return, which flows through to the utility’s shareholders as earnings. The business model is simple: invest capital in approved projects, add it to the rate base, earn the authorised return, repeat. The complexity lies not in the business model but in the gap between what the rate base records and what the physical infrastructure is actually worth.

Subscribe now

Rate Base Is Not Replacement Cost — And That Gap Is Growing

The rate base formula, as understood in US regulatory accounting, is approximately: gross plant in service minus accumulated depreciation, plus construction work in progress, plus regulatory assets, less regulatory liabilities, plus working capital. This is a compilation of historical investments, adjusted downward for depreciation charged over time and upward for regulatory timing differences.

What rate base is decidedly not is replacement cost. The electric distribution infrastructure of a major US city — the underground cables, the pad-mounted transformers, the feeder substations, the service connections to individual buildings — was built over a period of decades, often primarily in the 1960s through 1990s. The construction costs at that time reflected the steel prices, labour rates, permitting requirements, and equipment costs of those decades. A distribution transformer that cost $3,000 to install in 1975 might cost $15,000 to replace today. A mile of underground distribution cable that cost $200,000 per mile in 1985 might cost $1.2 million per mile today in a dense urban environment, given labour rates, street opening permits, traffic control requirements, and equipment costs. The rate base carries the 1975 transformer at its historical cost minus 45 years of depreciation — approximately zero. The replacement cost of that transformer is $15,000.

The Electric Energy Institute, in its analysis of US utility capital spending, noted that approximately 70% of North America’s electric infrastructure is more than 25 years old. In 2024, EEI member utilities invested a record $186.4 billion in capital expenditure — the twelfth consecutive year of record capital spending — and the 2025 figure is projected to be $214.7 billion, a 24% increase. This accelerating investment reflects two realities simultaneously: first, the aging infrastructure is being replaced at costs that are dramatically higher than the historical costs at which it was originally built; second, the rate base is growing rapidly because new investment at higher current costs is flowing in, while the old cheap infrastructure depreciated away long ago. The replacement of old, cheap, fully depreciated assets with new, expensive assets at current prices is systematically increasing the rate base — and the replacement cost of the combined system — faster than any financial model based on historical cost accounting can capture.

Subscribe now

PART II: THE ACCOUNTING TREATMENT AND THE GAP IT CREATES

What GAAP Records About a Transmission Line Built in 1975

Under US GAAP (ASC 980, which governs regulated operations) and IFRS (IAS 16 with regulatory deferral under IFRS 14), electric utilities record their plant and equipment at historical cost minus accumulated depreciation. The standard utility accounting treatment uses a composite depreciation method — each category of plant (transmission structures, conductors, transformers, substations) is depreciated over its estimated useful life, which regulators typically set at 30 to 50 years for most major infrastructure categories.

A transmission line built in 1975 at a construction cost of $50 million, depreciated over 40 years on a straight-line basis, would be fully depreciated by 2015. If still in service in 2026 — which many such lines are, because properly maintained steel towers and ACSR conductors can last 60 to 80 years or more — its net book value is zero or near-zero. The rate base may carry a small residual value if regulators have allowed deferred recovery of certain costs, but it is still a small fraction of the asset’s economic importance.

The replacement cost of that same transmission line in 2026 is a different number entirely. Right-of-way acquisition costs in 2026 are not comparable to 1975 costs, because environmental review requirements, eminent domain litigation costs, and landowner compensation expectations have all increased dramatically. Steel prices and fabrication costs for transmission structures are higher. Skilled labour for high-voltage transmission construction is scarce and expensive. Project management, engineering, and permitting for a major transmission line now typically takes 7 to 15 years from initial routing study to energisation — and that timeline has a cost that the 1975 comparison line did not incur in anything like the same magnitude.

The Federal Energy Regulatory Commission (FERC) has published data showing that the cost of new interstate transmission in the United States has increased from approximately $1 million per mile for a 345-kV line in the late 1990s to $3 million to $6 million per mile for comparable new construction in the 2020s, with certain urban and mountainous routes exceeding $15 million per mile. A utility whose transmission system was built at the 1990s cost base, currently carried at a small fraction of its 1990s cost due to accumulated depreciation, is sitting on infrastructure whose replacement cost is 3 to 6 times what the most recent construction would have cost — which is itself already 3 to 4 times the historical book cost.

This is the core of the accounting gap: a transmission line does not become less physically valuable because an accountant has depreciated its book value. The wires are in the ground. The towers are in the fields. The right-of-way easement — which is the most valuable and irreplaceable component of the entire transmission system — was secured decades ago and would be extraordinarily difficult or impossible to replicate today. The accountant records zero. The owner’s replacement cost is in the hundreds of millions or billions of dollars.

Subscribe now

PART III: THE OWNER’S VALUATION — REPLACEMENT COST AND THE PERPETUAL WATER RIGHT

What a Business Owner Would Pay for an Electric Utility Today

A sophisticated utility acquirer does not build a DCF model and apply an earnings multiple. They count the kilowatts of capacity, the miles of transmission and distribution lines, the number of substations, the rating of each major transformer bank, and the permitted capacity of each generating station. They then estimate the cost of replicating that system from scratch in today’s regulatory and construction environment. That replacement cost estimate is the economic floor on the asset’s value — the value of having already built what would be extraordinarily difficult and expensive for any competitor to replicate.

The owner then adjusts this replacement cost estimate downward for two factors: the remaining useful life of the existing assets (which reduces the time until replacement capital must be spent) and the regulatory environment (which determines how quickly new capital spending can be recovered through rates). Against this replacement cost floor, the owner compares the current EV/Rate Base multiple being demanded by the seller.

In US utility M&A, recent transactions have priced at approximately 1.3 to 1.7 times rate base. But rate base, as explained above, is substantially below replacement cost for most mature utility systems. If a utility’s transmission and distribution system has a rate base of $5 billion — reflecting 50 years of historical cost accumulation minus depreciation — but its replacement cost is $20 billion (because the assets were originally built at 1970s costs and would cost 4 times as much to rebuild today), then a purchase price of $7 billion (1.4x rate base) is actually only 0.35x replacement cost. The buyer is paying 35 cents for every dollar of infrastructure value that cannot be replicated by any competitor on any timeline.

This is the gap the owner sees that the earnings model does not. The earnings model takes the current rate base, applies the allowed return on equity, generates a current earnings figure, and applies a multiple. The replacement cost analysis asks a more fundamental question: what would it cost to build this system today, and what am I paying relative to that cost?

Subscribe now

The Hydroelectric Case: Where the Gap Becomes a Chasm

The most extreme version of the replacement cost gap exists in hydroelectric generation, and Brookfield Renewable Partners provides the most documented and publicly disclosed example of this dynamic in the global markets.

Brookfield Renewable operates what it describes on its own website as “perpetual assets” — hydroelectric generating stations built primarily in the 1920s through 1950s, on rivers in the northeastern United States, Canada, Brazil, Colombia, and Europe. These stations produce electricity by capturing the kinetic energy of falling water, using turbines and generators that, while requiring periodic maintenance and occasional major refurbishment, have no fuel cost. The water flows from the mountains to the sea whether or not you capture its energy. The hydroelectric station’s job is simply to be in the path of that flow and to convert it into electricity. The Brookfield 2023 Annual Report describes PP&E totalling $64.0 billion across its entire renewable energy platform — but the carrying values of the oldest hydro assets, built a century ago, have been depreciated to tiny fractions of their original construction costs.

What would it cost to build these assets today? The answer is approximately never, at any price. In the northeastern United States, permitting a new large conventional hydroelectric dam requires navigating the Federal Energy Regulatory Commission’s (FERC) hydroelectric licensing process, environmental review under the National Environmental Policy Act, state environmental permits, consultations with tribal nations, fish passage requirements, and litigation from environmental groups and downstream water users. The timeline from initial application to operating license for a new major hydro facility in the US is typically 10 to 20 years. The environmental mitigation costs — fish passage facilities, minimum flow requirements, sediment management — are enormous and were not incurred by the 1920s and 1930s builders who simply received a permit from a then-compliant federal government and built their dam. The replacement cost of a new large hydro facility is therefore not merely the civil construction cost. It is the civil construction cost plus two decades of permitting plus tens of millions of dollars in environmental mitigation plus the opportunity cost of the capital during the waiting period.

Brookfield explicitly acknowledges this. The company’s investor materials consistently describe the hydroelectric portfolio as worth substantially more than its depreciated book value, and note that the assets are “irreplaceable” in the context of their licensed water rights and physical locations. This is not marketing language — it is an accurate description of the regulatory and physical reality. A water right that has been in continuous productive use since 1927 cannot be replicated by a competitor regardless of capital budget. The right itself, codified in a FERC hydroelectric licence that is renewed every 30 to 50 years, is an intangible asset that appears nowhere on the balance sheet at fair value. The civil works — the dam, the powerhouse, the penstocks — are depreciated to near-zero. The licensed water right, which is the reason the civil works were built there in the first place, is carried at whatever nominal cost was incurred in obtaining the original permit a century ago.

The owner who looks at this situation sees three layers of value that the balance sheet does not capture. First, the physical civil works at replacement cost — dams and powerhouses that would cost billions to build new and cannot be permitted for new construction in most US locations. Second, the licensed water right — a perpetual entitlement to capture a specified flow of water for power generation, renewed by FERC subject to environmental conditions, with no expiration and no competition from any other developer who could obtain a new licence for the same site. Third, the operational learning — a century of operational knowledge about the specific hydrology of each site, the optimal turbine settings for different flow conditions, the maintenance intervals for each specific piece of equipment in each specific environment. None of these three layers of value appears on the balance sheet at anything close to its economic reality.

Subscribe now

PART IV: THE WALL STREET PROBLEM — WHY EARNINGS MULTIPLES FAIL UTILITIES

The Circularity of Utility Earnings Models

The standard Wall Street approach to valuing a regulated utility begins with the current rate base, applies the allowed return on equity, generates a forecast of utility earnings, and then applies a price-to-earnings or price-to-book multiple derived from comparable utility transactions or sector averages. This approach has a circularity problem that is rarely acknowledged.

The P/E multiple applied in the utility sector is itself derived from the historical relationship between utility share prices and utility earnings. But utility earnings are themselves a function of the rate base — the regulatory allowed return on the historical cost asset base. This means the P/E multiple is calibrated to value a stream of earnings that is anchored to historical cost accounting. It is perfectly circular: it tells you what earnings are worth in a market where everyone is using the same earnings — which are based on the same historical cost accounting — to set prices. It cannot tell you whether the underlying physical assets are cheap or expensive relative to their replacement cost, because replacement cost never enters the calculation at any point.

The practical consequence is that a utility whose assets were built at 1970s costs, currently earning its 10% allowed return on a rate base that is 20% of replacement cost, looks identical in the P/E framework to a utility whose assets were built in 2010 at near-current costs, earning the same 10% return on a rate base that is 85% of replacement cost. Both earn approximately the same return on their respective rate bases. Both will trade at approximately the same P/E multiple. But the first utility’s physical infrastructure is worth 5 times the rate base in replacement cost terms, while the second utility’s is worth only 1.2 times. The investor who bought the first utility at 1.5x rate base (which equals 0.3x replacement cost) and the investor who bought the second at 1.5x rate base (which equals 1.28x replacement cost) paid completely different prices for completely different assets — and the P/E model treats them identically.

The EV/EBITDA framework fails for a related reason: EBITDA for a regulated utility is determined by the rate base and the allowed return, not by the competitive position of the underlying assets. A utility with $5 billion of rate base earning 10% ROE generates approximately $500 million in operating income before depreciation — regardless of whether the $5 billion rate base represents $25 billion of infrastructure at replacement cost (in which case the assets are extremely cheap) or $6 billion of replacement cost infrastructure (in which case they are fairly priced). EBITDA flattens the distinction. Only the replacement cost analysis preserves it.

Subscribe now

PART V: STORIES FROM THE GRID — THREE CASE STUDIES IN UTILITY VALUE ARBITRAGE

The Aging American Grid: When Deferred Depreciation Becomes Someone Else’s Windfall

The story of American electric transmission infrastructure since the energy deregulation wave of the 1990s is, in one sense, a story of extraordinary value accumulation that was invisible to the investors who owned it. During the 1990s and early 2000s, the prevailing consensus in the US utility sector was that electricity deregulation would restructure the industry into competitive retail markets, competitive generation, and regulated wires. Merchant power companies would compete for generation business on price. Transmission and distribution would remain regulated. The “wires” business — transmission and distribution alone, without generation — was considered boring, slow-growing, and unsuitable for investors who wanted growth.

Large industrial companies and foreign utilities — National Grid (UK), E.ON (Germany), Scottish Power — accumulated US transmission and distribution assets at prices reflecting this consensus. They bought aging New England and mid-Atlantic utility systems that had been built primarily in the 1950s through 1970s, at book values reflecting the historical construction costs of those decades, at a time when the regulatory framework was uncertain and the assumed trajectory was toward more competition, not more regulation.

What they were actually acquiring — though few framed it this way at the time — was physical infrastructure with replacement costs that were 5 to 10 times the carrying values on the balance sheet, in service territories where the regulatory framework (despite the deregulation rhetoric) continued to guarantee recovery of all prudently incurred costs and an adequate return on invested capital. The right-of-way corridors, in particular, were irreplaceable: a transmission corridor through densely developed New England suburbs that had been secured in the 1950s under much simpler permitting requirements could not be replicated at any cost in the 2000s, because the communities along its route had long since developed around and above the existing corridor and would resist any attempt to establish a new one.

National Grid’s acquisition of New England Electric System in 2000 for approximately $3.2 billion is illustrative. The combined New England transmission and distribution infrastructure — serving Massachusetts, Rhode Island, and New Hampshire — had a net book value at the time that reflected 40 to 50 years of accumulated depreciation on assets originally built at 1950s and 1960s costs. An owner building a replacement cost model would have arrived at a very different number — the replacement cost of the same wires, substations, and rights-of-way using 2000-era construction costs and regulatory requirements. That gap, invisible in the price-to-book framework used in the utility sector at the time, represented extraordinary durable value for a buyer with a long investment horizon.

Subscribe now

Brookfield and the Hundred-Year Hydroelectric Dam: When Accounting Age Meets Perpetual Physical Value

In 1999, Brookfield Asset Management began systematically acquiring North American hydroelectric generating stations that had been divested by vertically integrated utilities during the deregulation wave. Many of these facilities had been built between 1900 and 1950 — some were approaching or exceeding 80 years of age. Their book values, after decades of straight-line depreciation, ranged from modest to zero. Their FERC licences, while requiring periodic renewal and environmental compliance, were perpetual in the sense that the water flow they captured was perpetual. The rivers would not stop flowing.

The Brookfield thesis was not expressed in earnings multiples or EBITDA valuations at the time. It was expressed in physics and in hydrology. A cubic meter per second of water falling 100 meters generates a calculable number of kilowatt-hours of electricity. The facility that captures that energy earns revenue at the prevailing electricity price for however long the water flows, which is essentially forever. The operating cost of a fully depreciated hydroelectric facility — primarily maintenance, staffing, and transmission access charges — is a small fraction of the revenue. The capital cost of the facility has been paid once, depreciated once, and does not need to be incurred again unless the civil works physically fail, which properly maintained hydroelectric dams almost never do within a human-relevant time horizon.

The owner’s calculation for a 1930s-vintage hydroelectric facility is straightforward. Take the generating capacity in megawatts. Multiply by the capacity factor (the fraction of time the facility actually generates) to get annual megawatt-hours. Apply the prevailing power purchase agreement price or merchant electricity price per MWh. Subtract operating costs. The result is the annual operating income from a facility that cost nothing to build (in the sense that the capital has been fully depreciated) and that no competitor can replicate. The replacement cost of an equivalent greenfield hydroelectric facility in the same location, today, is the relevant floor on value — and that replacement cost, as established earlier, is functionally incalculable for most US hydro sites given the permitting environment.

Brookfield’s 2023 Annual Report disclosed that PP&E totalled $64.0 billion, reflecting cumulative acquisition and development capital invested across the global portfolio. Crucially, Brookfield reports its hydroelectric assets under IFRS, which allows but does not require the revaluation of PP&E to fair value under the IAS 16 revaluation model. Brookfield elects the cost model, meaning its hydro assets are carried at historical acquisition cost minus depreciation — not at replacement cost or fair value. The company explicitly notes in its investor materials that the hydroelectric portfolio is the most valuable and irreplaceable component of its asset base, generating long-duration, essentially free-fuel power from licensed water rights that no new entrant can obtain.

When Brookfield sells portions of its hydro portfolio — as it did with a 78% interest in a 378 MW US hydroelectric portfolio in Q1 2023 — the transaction prices reveal the gap between IFRS carrying values and market clearing prices. These sales, consistently transacted at prices well above IFRS book values, provide the most direct evidence of the replacement cost premium embedded in assets that the accounting system has been systematically writing down toward zero for 70 years.

Subscribe now

The First Energy Transmission Sale: What $4.85 Billion for Old Wires Revealed

In February 2021, FirstEnergy Corp announced the sale of a minority interest in FirstEnergy Transmission LLC (FET) to Brookfield Super-Core Infrastructure Partners for approximately $2.4 billion, implying a total enterprise value for FET of approximately $23 billion for a set of transmission assets serving West Virginia, Pennsylvania, New Jersey, Ohio, and Maryland.

The FET transmission system, built primarily between the 1960s and 1990s, had a rate base at the time of the transaction of approximately $9 billion. Brookfield paid for a minority interest at a price implying approximately 2.55x rate base for the total transmission enterprise. A standard utility sector model — which applies 1.3 to 1.7x rate base as a normalised transaction range — would have said Brookfield overpaid. But Brookfield was not applying the standard earnings model to the standard rate base.

Brookfield was applying the replacement cost framework to a transmission system whose right-of-way corridors through the densely populated mid-Atlantic region were assembled over 40 to 60 years, under environmental and land acquisition conditions that no longer exist. The cost of replicating those corridors today — even if money were unlimited — would be measured not in billions but in decades of litigation, environmental review, and community opposition. The existing corridors are the irreplaceable asset. The wires and towers strung on those corridors are replaceable. The right to put wires on those corridors in perpetuity, through suburban Pennsylvania and rural West Virginia, is not.

At 2.55x rate base and a rate base that is perhaps 25% to 30% of actual replacement cost for a 1970s-era transmission system in the mid-Atlantic, Brookfield paid approximately 0.64x to 0.77x replacement cost for an irreplaceable infrastructure monopoly — below replacement cost, for infrastructure that no competitor can replicate. The earnings model said 2.55x rate base looks expensive relative to the 1.5x sector norm. The replacement cost model said 0.64x to 0.77x replacement cost looks reasonable for a monopoly asset in a high-regulatory-barrier jurisdiction.

The lesson here is precise: the same transaction looks expensive or cheap depending entirely on which measurement framework you apply to it. The earnings model — applying a multiple to a rate-base-derived earnings figure — sees 2.55x and says expensive. The replacement cost model — dividing the purchase price by the estimated cost of reproducing the same infrastructure from scratch today — sees 0.64x to 0.77x and says reasonable. Brookfield, which has built its entire franchise on understanding this distinction, was applying the second framework. The market commentary at the time was largely applying the first.

Subscribe now

PART VI: THE THREE-LAYER VALUE FRAMEWORK FOR REGULATED UTILITIES

How an Owner Actually Measures What They Are Buying

Acquiring a regulated electric utility is not primarily a financial exercise. It is primarily an engineering exercise followed by a regulatory exercise, with the financial analysis in third place. The owner’s framework operates in three distinct layers, each of which captures a different component of value that the earnings model ignores entirely.

Layer one: the civil works replacement cost. This is the cost of building the generation, transmission, and distribution infrastructure from scratch, using current labour, materials, and equipment costs, but excluding the permitting process. It is the “bricks and mortar” replacement cost — the steel in the towers, the copper in the cables, the concrete in the dam. This layer is calculable from engineering cost databases, recent comparable construction projects, and utility capital expenditure disclosures. It is consistently well above the GAAP book value for infrastructure built more than 15 to 20 years ago, because construction costs have risen dramatically and the accounting has been depleting the old cost basis.

Layer two: the right-of-way and permitting premium. This is the value of having already secured the permits, rights-of-way, and regulatory approvals that allow the infrastructure to exist in its current location. For a transmission corridor through a densely populated suburban area, this layer may be the largest of the three — the physical cost of building a new transmission line may be $3 million per mile, but the right-of-way acquisition cost, permitting timeline, litigation exposure, and environmental compliance for a hypothetical new parallel corridor may add another $5 to $15 million per mile in implicit cost. This layer is what makes the existing infrastructure truly irreplaceable: even unlimited capital cannot compress the 10 to 20 year permitting timeline for major new transmission infrastructure. This layer does not appear on any balance sheet. It is the accumulated benefit of decisions made by utility engineers and regulators decades ago that could not be replicated today at any price in any reasonable timeframe.

Layer three: the licensed water right for hydroelectric assets. This is the most extreme version of the unrecognised value, and it applies specifically to regulated hydroelectric generating stations. A hydroelectric licence is not a depreciating asset. The water flows whether or not it is captured. The licensed right to capture it — granted by FERC for 30 to 50 year terms and historically renewed subject to environmental conditions — is a perpetual entitlement to a fuel source that has zero ongoing cost. A 100 MW hydroelectric facility with a 40% capacity factor produces approximately 350,000 MWh per year, which at $60/MWh generates $21 million per year in revenue — perpetually, with no fuel cost. The FERC licence that enables this revenue stream is the asset. The civil works that harness it are the tool. Both are irreplaceable. Neither is fully reflected on any balance sheet.

The practical metric that synthesises these three layers is the ratio of Enterprise Value to Estimated Replacement Cost — what the buyer is paying relative to what it would cost a competitor to reproduce the same infrastructure system. For mature, fully depreciated utility systems, this ratio is frequently below 0.5x, meaning buyers are paying less than half the cost of reproduction. For more recently built systems, the ratio approaches or exceeds 1.0x.

An owner applying this framework to a portfolio of 1960s-era electric distribution assets in a dense urban market might calculate: civil works replacement cost at $X per mile of cable times total cable miles, plus right-of-way premium based on current land and permitting costs in the service territory, plus the value of existing customer relationships and utility franchise rights. Compare the total to the acquisition price. If the acquisition price is below 0.6x of this total, the owner is acquiring physical infrastructure monopoly at below replacement cost — which, in a market where no new competition is possible, means they are acquiring an economic moat at a discount to the cost of competing with it.

Subscribe now

PART VII: CONFIRMED CATALYSTS — WHAT IS ALREADY ACCELERATING

Catalyst 1: The AI and Data Centre Load — Already Contracted, Not Forecasted

The most significant structural change in the US electric utility industry in a generation is happening right now, driven by the construction of artificial intelligence data centres and the hyperscalers’ insatiable demand for reliable, large-scale electricity. Microsoft, Google, Amazon, and Meta are collectively committing to build data centre campuses requiring hundreds to thousands of megawatts of power at individual sites, signed under long-term power purchase agreements directly with utilities, typically for 10 to 20 year terms.

The Gabelli utility research report, published in October 2025, noted that AEP forecasts 2025 to 2027 retail sales growth of 5.7%, 8.4%, and 8.9% respectively, driven heavily by commercial and industrial data centre load. NiSource expects to add new dispatchable generation to serve data centre customers who have committed to the utility’s service territory through 2032. ERCOT projects Texas peak load to rise from 86 GW in 2024 to 130 to 148 GW by 2030 — a 50 to 72% increase in six years.

This load growth is not a forecast that might or might not materialise. It is committed by customers who have signed power purchase agreements and are building data centres in the utility’s service territory right now. The utilities are required to serve these customers and are building generation and transmission infrastructure to do so, all of which flows into the rate base — at current construction costs, which are dramatically higher than the costs at which the existing system was built. Each dollar of new capital investment adds to the rate base at current cost levels, systematically expanding the rate base and widening the gap between rate base and the historical cost book value of the older portion of the system.

For the owner’s replacement cost framework: data centre load growth means the existing transmission and distribution infrastructure — which was sized for the historical load pattern — is now being supplemented with new capacity that will be built at current costs. The old infrastructure serving the existing load becomes, in effect, even more valuable relative to its book value, because the incremental load it serves is growing and the replacement cost of the entire expanded system is growing faster than the rate base as old assets are replaced at current prices.

Subscribe now

Catalyst 2: Grid Modernisation and Infrastructure Age — The Replacement Wave Is Already Funded

As noted by the Department of Energy and confirmed by the Edison Electric Institute, approximately 70% of North America’s electric grid infrastructure is more than 25 years old. The US electricity system — which was built primarily from the 1950s through the 1990s — is entering its replacement cycle, and that replacement is happening at current construction cost levels that are 3 to 5 times the original cost levels of the assets being retired.

In 2024, EEI member utilities invested $186.4 billion in capital expenditure — the twelfth consecutive record year. S&P Global projects 2025 capital expenditure at $214.7 billion across tracked utilities, a 24% increase in a single year. These are not aspirational numbers. They are actual capital budgets approved by utility boards, filed with state regulators in rate cases, and in many states pre-approved for recovery through forward-looking rate mechanisms that eliminate regulatory lag.

Each year of record capital expenditure at current cost levels is simultaneously two things: it is visible in the capital expenditure line of the utility’s financial statements, and it is invisible in the impact it will have on future rate base relative to the old historical cost assets being retired. When a utility retires a transformer that was installed in 1974 at $50,000 and replaces it with a new transformer at $180,000, the rate base increases by $130,000 for that single asset, and the rate base effect of every such replacement across thousands of similar assets accumulates systematically over a decade of infrastructure replacement. The EV/Rate Base multiple paid in 2026 for a utility’s current rate base will look inexpensive relative to what that rate base will be in 2035 after a decade of replacement at current construction costs.

Subscribe now

Catalyst 3: Transmission Permitting Reform — Already Enacted Under FERC Order 1920

The Federal Energy Regulatory Commission’s Order 1920, issued in May 2024, represents the most significant reform to US transmission planning in decades. The order requires transmission planning on a longer geographic horizon (20 years forward rather than 10), mandates cost allocation for regional transmission projects that serve multiple beneficiaries, and creates a framework for scenario-based planning that explicitly accounts for the load growth from electrification and data centres.

The practical consequence is that FERC has now established a regulatory pathway for large regional transmission projects that will add substantially to the rate bases of transmission-owning utilities over the next decade. These projects — driven by the need to integrate renewable generation and serve new large loads — were previously delayed or blocked by the absence of a coherent cost allocation framework. With Order 1920, the pathway from project identification to regulatory approval to construction to rate base addition is clearer than it has been at any point in the past 30 years.

For the replacement cost analysis: Order 1920 creates mandated new transmission investment at current construction costs. Every mile of new transmission built under the Order 1920 framework adds to the rate base at 2026 to 2035 construction costs — costs that are 3 to 5 times the historical costs of the existing transmission infrastructure. The gap between rate base and replacement cost for the combined system will narrow as new investment at current costs replaces old investment at historical costs — but in the near term, the existing old infrastructure continues to deliver its economic function at a book value that is a fraction of what a new entrant would pay.

Subscribe now

Catalyst 4: The Hydroelectric Relicensing Wave and FERC’s Renewable Preference

Between 2025 and 2035, a significant number of existing FERC hydroelectric licences are due for renewal. The FERC relicensing process has historically been the moment of greatest risk for hydroelectric operators — environmental conditions attached to renewed licences can add significant cost through fish passage requirements, minimum flow guarantees, and water temperature management. But the energy policy context has shifted dramatically: hydroelectric generation is now the only large-scale renewable resource that provides firm, dispatchable power — power that can be delivered on demand regardless of weather conditions. Wind and solar cannot make this claim.

The growing recognition among regulators, grid operators, and policymakers that dispatchable renewable generation is worth more than intermittent renewable generation is creating a policy environment in which hydroelectric facilities are being relicensed with significantly more regulatory goodwill than in previous decades. FERC’s recent relicensing approvals have reflected an explicit acknowledgement that the power system needs firm, dispatchable, renewable generation to balance the variability of wind and solar. This policy shift — already visible in recent FERC relicensing orders — is a confirmed catalyst for the retention and enhancement of hydroelectric asset value that the accounting system carries at near-zero book value.

Subscribe now

CONCLUSION: THE INFRASTRUCTURE THAT AGES ON PAPER BUT NOT IN THE GROUND

There is something almost philosophically confusing about the accounting for regulated electric utilities — and that confusion is the source of the value arbitrage described in this paper.

When a mine produces iron ore, it is genuinely depleting an asset. The ore extracted today is ore that cannot be extracted tomorrow. The UOP depletion method, whatever its other limitations, is directionally correct: the asset is being consumed as production continues, and the declining book value reflects a genuine physical reality.

When an electric utility operates a transmission line, nothing is being consumed in the same sense. The right-of-way corridor that was secured in 1968 is still a right-of-way corridor in 2026. The land it traverses has not been depleted. The towers standing on it are not meaningfully worn out if they have been maintained. The accounting depreciates the asset to zero over 40 years because the accounting standard requires a consumption assumption, but the physical asset — and more importantly, the competitive advantage embodied by the right-of-way and the regulatory franchise — does not diminish in proportion to the depreciation charge.

This means that, uniquely among physical asset classes, a fully depreciated electric utility infrastructure system is frequently worth more in replacement cost terms than a recently built one — because the old system was built when costs were lower, secured rights-of-way under more permissive conditions, and operates under regulatory relationships that have been established and stress-tested over decades. The age of the asset, which accounting treats as a reason to reduce its value, is often a reason to increase its competitive value.

An owner who understands this buys regulated utility infrastructure with a long time horizon, applies the replacement cost framework rather than the earnings multiple framework, and earns the returns that accrue to anyone who acquires an irreplaceable monopoly infrastructure asset at below its cost of replication. The grid will be rebuilt. It will be rebuilt at current costs. The owner of the existing rights-of-way, the existing regulatory franchises, and the existing licensed water rights will be the direct beneficiary of every dollar that higher replacement costs add to the replacement cost premium above book value.

The wire in the ground does not know what year the accountant depreciated it. It carries current with the same indifference it has carried current since the day it was energised. The replacement cost is what it is. And the gap between that replacement cost and the number the accounting system produces is, in the regulated utility sector, often among the largest and most durable such gaps in the entire investment universe.

This white paper is for educational and informational purposes only. Data on utility capital expenditure is drawn from the Edison Electric Institute’s annual financial review and S&P Global Market Intelligence tracking of investor-owned utilities. FERC Order 1920 is publicly available from the Federal Energy Regulatory Commission. FirstEnergy Transmission transaction data is drawn from FirstEnergy Corp’s Form 8-K filed February 2021. Brookfield Renewable data is drawn from the company’s 2023 Annual Report and 20-F filed with the SEC. AES Indiana rate case information is from the Indiana Utility Regulatory Commission’s publicly available case files for Cause No. 45911. Nothing herein constitutes investment advice.

If you like this read, check out my other writings below 👇

Before We Begin: A Story About the Same Asset Being Valued at Two Different Numbers on the Same Day

On April 14, 2020, Delta Air Lines’ stock closed at $22.46 per share. With approximately 645 million shares outstanding, the market capitalisation was roughly $14.5 billion. The market had decided, in the crucible of a global pandemic that had grounded 95% of commercial aviation, that the entire Delta Air Lines enterprise — aircraft, gates, routes, brand, customer relationships, loyalty programme, and everything else — was worth $14.5 billion.

On that same day, the slot portfolio that Delta held at New York’s LaGuardia Airport, Ronald Reagan Washington National Airport, and John F. Kennedy International Airport — the legal rights to take off and land at specific times at three of the busiest and most constrained airports in the United States — was sitting on Delta’s balance sheet. The disclosed carrying value for Delta’s domestic slots, as reported in its December 2019 10-Q impairment test, was $622 million. Delta’s own fair value assessment at that same December 2019 test showed that the fair value of those domestic slots exceeded the carrying value by 61% to 181%.

Apply even the low end of Delta’s own fair value disclosure — 61% above the $622 million carrying value — and the domestic slots alone were worth approximately $1 billion in Delta’s own reckoning at pre-pandemic testing. Add the international routes and slots, which Delta disclosed at a carrying value of $2,583 million with a fair value 15% to 29% above that, and the total slot and route intangible portfolio had a Delta-disclosed fair value of approximately $3.5 to $4.2 billion — based on the company’s own impairment testing numbers, filed with the SEC.

In April 2020, the market capitalization of the entire airline was $14.5 billion. The slot portfolio alone, at values Delta’s own accountants were comfortable defending to auditors just four months earlier, represented roughly 24% to 29% of the entire market cap of the company. The question the stock market was not asking — because the stock market was busy pricing in bankruptcy, zero revenue, and existential fear — is what is the slot portfolio worth on a standalone basis, and how much of the company’s total value does that represent at the current market price?

The answer, if you had asked it, was: more than the market thought. Much more. This paper is about why.

Subscribe now

PART I: WHAT AN AIRPORT SLOT ACTUALLY IS — BEFORE ANY FINANCE

The Physics of Congested Airspace

An airport runway can only handle a certain number of aircraft movements per hour. This is not a policy choice. It is physics, aerodynamics, and safety regulation combined. A runway requires a certain minimum spacing between successive aircraft to allow the preceding aircraft’s wake turbulence to dissipate. For a large aircraft like a Boeing 777 taking off ahead of a smaller regional jet, that spacing might be four to five minutes. For two large jets in sequence, perhaps two to three minutes. Factor in the approach sequencing, the taxi time, the runway occupancy time, and the departure roll, and you arrive at a maximum theoretical capacity for a single runway of roughly 30 to 45 aircraft movements per hour under ideal conditions.

At London Heathrow — two runways, four terminals, handling roughly 480,000 aircraft movements per year — this constraint is not theoretical. It is the fundamental reality around which the entire airport operates. Every take-off and landing at Heathrow has been planned, allocated, coordinated, and slotted into a precise window. If you arrive at a Heathrow runway without a slot, you do not take off. There is no queue you can join by simply showing up. The slot — the government-allocated authorisation to use the runway at a specified time — is the legal right to occupy a piece of the airport’s most finite and most irreplaceable resource: the runway itself.

At the six truly slot-controlled airports in the world — Heathrow, LaGuardia, Reagan National, Tokyo’s Haneda, Sydney Kingsford Smith, and a handful of others — the government has made a formal, permanent, legally binding determination that the runway capacity is fully allocated. There is no more. No new entrant can simply pay to use these runways during peak hours because there are no slots available for sale by the airport. The only way to obtain a slot at a slot-controlled airport is to buy one from a carrier that already holds one, negotiate an exchange, or wait for the essentially nonexistent event of an existing carrier surrendering a slot voluntarily.

This is the asset. Not the airline. Not the aircraft. Not the brand. The slot — the piece of government authorisation to use a runway at a specific time — is the primary income-generating asset at these airports, and its scarcity is absolute.

Subscribe now

Why Some Airports Are More Equal Than Others

Not every airport is slot-controlled, and not every slot-controlled airport is equally valuable. The economics of a slot are determined by two factors: the pricing power of the routes that can only be served from that airport, and the regulatory permanence of the constraint that makes the slot scarce.

Reagan National Airport in Washington, DC operates under both slot controls and a perimeter rule — airlines can only fly non-stop to destinations within 1,250 miles of the airport, with a small number of exceptions for limited beyond-perimeter destinations. The perimeter rule concentrates demand at DCA from exactly the markets where frequency and schedule convenience matter most: the business travellers on the Washington-New York, Washington-Boston, Washington-Chicago, and Washington-Atlanta corridors who will pay premium fares for the ability to fly into an airport 15-minutes from Capitol Hill rather than making the 45-minute journey from Dulles. A slot pair at Reagan National is not merely the right to take off and land. It is the right to offer the most convenient service to some of the highest-yield business travellers in the United States.

LaGuardia, similarly, serves New York’s business community with proximity that JFK and Newark cannot match for travellers heading to Midtown Manhattan. A shuttle from LaGuardia to Boston takes 45 minutes in the air and drops passengers within a reasonable taxi or subway ride of the financial district. The same journey from JFK adds 30 to 40 minutes of ground transportation. For the financial executive who makes this trip multiple times per week, the LaGuardia slot is worth paying a premium for — and that premium feeds directly into the economic value of the slot.

Heathrow’s value proposition is different but equally structural. It is the primary connecting hub for transatlantic traffic between Europe and North America. A carrier without Heathrow access cannot offer the connectivity — the ability to connect from dozens of European cities to dozens of North American cities through a single hub — that Heathrow provides. This is not merely a matter of passenger preference. It is a matter of commercial impossibility: the connecting traffic that makes a transatlantic network profitable requires the hub infrastructure that only Heathrow provides at the scale and geography that airlines need. A carrier without Heathrow slots cannot compete meaningfully for premium transatlantic connecting traffic. A carrier with Heathrow slots has access to a market that is genuinely closed to carriers without them.

Subscribe now

PART II: THE ACCOUNTING TREATMENT — HOW ZERO BECAME THE MOST IMPORTANT NUMBER IN AIRLINE FINANCE

The Origin Story of the Grandfather Slot

The slot system at US airports was established under the FAA’s High Density Rule, initially implemented in 1969 at four airports: JFK, LaGuardia, Reagan National, and O’Hare. The rule divided the available hourly capacity of each airport into a fixed number of slots and allocated those slots to the carriers then operating at each airport — at no charge. Airlines that were flying from LaGuardia in 1969 received their slots for free, as a regulatory gift from the federal government in the form of authorised operating rights.

These original slot allocations — and subsequent allocations made as part of the 1985 and 1993 slot control reforms — are carried on airline balance sheets at exactly what the airlines paid for them: zero. Nothing was purchased. No consideration changed hands. The FAA granted the rights administratively, and the accounting standards in place at the time (and since) require assets to be recognised at cost. The cost of a regulatory gift is zero. Therefore, the balance sheet value of these grandfathered slots is zero.

This creates the foundational absurdity at the heart of this paper: the asset that is most definitively not replicable by any competitor — an authorisation to use a runway at a permanently constrained airport, allocated by a government agency that has not created a new runway-capacity slot in decades — appears on the balance sheet of its owner at no value whatsoever.

Slots purchased in the secondary market since the post-1993 liberalisation of slot trading are treated differently. Under US GAAP, these slots are recognised as indefinite-lived intangible assets at their purchase cost, not amortised (because their useful life is indefinite), and tested annually for impairment — therefore resulting in a higher absolute value in the future, due to inflation. Delta’s domestic slot portfolio — which includes a mix of originally grandfathered slots and subsequently purchased slots — carried a net value of $622 million on its balance sheet. American Airlines’ intangible asset footnote in its 2023 10-K groups its airport slots within a broader intangibles disclosure totalling approximately $2 billion in net intangibles, which includes both slots and routes.

The critical point is that neither the carrying value of $622 million for Delta’s domestic slots nor the approximately $2 billion in total net intangibles for American’s portfolio bears any relationship to the transaction prices that these assets command in the secondary market. The carrying values reflect a combination of historical zeros (for grandfathered slots) and historical purchase prices from transactions that may be 20 or 30 years old. The market value of the assets as of any given day reflects current demand from airlines willing to pay today’s prices for access to today’s most congested airports.

Subscribe now

PART III: THE OWNER’S VALUATION — WHAT THE SECONDARY MARKET ACTUALLY PAYS

The Transaction Record That Wall Street Ignores

An owner valuing an airline’s slot portfolio does not need a financial model. They need a transaction database and the ability to multiply. The secondary market for airport slots has produced documented, arm’s-length transactions at prices that are publicly available from SEC filings, regulatory disclosures, and industry reporting. These prices are as objective as the current copper spot price or the LIFO reserve disclosed in a steel distributor’s footnote. They represent what a willing buyer and a willing seller agreed on in a competitive process, with full information about the asset.

At London Heathrow, the most liquid and most expensive slot market in the world, the transaction record shows consistent escalation. In 2007, a pair of Heathrow slots sold for approximately $30 million. By 2014, comparable slot pairs were transacting at $60 to $75 million. In 2016, American Airlines sold a pair of Heathrow slots to American Express Global Business Travel for approximately $60 million — a disclosed arm’s-length transaction. Virgin Atlantic, in December 2015, used its Heathrow slot portfolio as collateral for a £225 million private placement bond — effectively obtaining a market-based valuation of the slots from professional debt investors who needed to be confident in the collateral value before lending against it.

At Reagan National, the transaction record shows slot pairs at $20 to $40 million in the years preceding COVID. At LaGuardia, slot pairs have transacted at $10 to $20 million. These are not estimates or opinion. They are the prices at which airlines have actually transferred slots to each other, often disclosed in SEC filings because the transactions were material enough to require regulatory reporting.

Subscribe now

Applying the Transaction Database to a Specific Airline

Consider American Airlines’ disclosure in its June 2020 8-K filing with the SEC. Facing the existential financial pressure of COVID, American’s management produced a detailed asset-by-asset breakdown of the company’s unencumbered assets for the purpose of securing emergency financing. The filing disclosed the following third-party appraised values for its Slots, Gates, and Routes portfolio:

• Caribbean routes and associated slots: $2,520 million

• Asia and Australia routes and associated slots: $310 million

• Canada and Non-EU routes: $380 million

• Central America and Mexico routes and slots: $4,240 million

• Total Slots, Gates, and Routes (disclosed): $7,450 million

These values were not produced by American Airlines’ management for marketing purposes. They were produced by independent third-party appraisers — the same appraisers that lenders and investors use when deciding whether to lend money against an asset. They represent the professional opinion of people whose livelihoods depend on producing accurate valuations that can withstand scrutiny, because the alternative — overvaluing collateral — is what gets appraisers sued and lenders bankrupted.

This is not a theoretical argument about what the assets might be worth under optimistic assumptions. It is the comparison between independently appraised values that American itself was presenting to professional lenders — people who were deciding whether to extend $1.2 billion in emergency financing secured by the LGA/DCA slots — and the market price of the equity at the same moment.

Subscribe now

PART IV: STORIES FROM THE TARMAC — THREE CASE STUDIES IN SLOT VALUE ARBITRAGE

American Airlines’ COVID Financing: When the Company Itself Proved the Asset’s Worth

The most authoritative evidence that airline slots have substantial standalone value independent of the airline’s operating condition came not from an investor thesis or an analyst report. It came from American Airlines’ own crisis financing in 2020.

In July 2020, with global air travel at approximately 5% of normal levels, no meaningful revenue, and a deeply uncertain future, American Airlines needed to raise emergency capital to survive. The company had two choices: sell equity at deeply diluted prices, reflecting the crisis-level market capitalisation, or find assets to pledge as collateral to secured lenders. Management chose the latter.

On July 23, 2020, American filed an 8-K disclosing that Goldman Sachs had committed to purchase $1.0 billion in IP Notes and $200 million in LGA/DCA Notes — the latter secured by a first-lien interest in American’s slots at LaGuardia and Reagan National. The LGA/DCA slots, which on the balance sheet were carried at historical cost values reflecting a mix of zero (for grandfathered slots) and old acquisition prices, were being used as collateral for $200 million in emergency financing at a moment when the entire airline was under existential financial stress.

The lenders extending $200 million against these slots were professional credit investors who were not motivated by sentiment or narrative. They were motivated by the recovery value of the collateral if American Airlines defaulted. Their willingness to lend $200 million against the LGA/DCA slots — at the darkest moment of the airline industry’s history in generations — confirmed that independent, financially-sophisticated market participants believed the slots had a standalone value substantially in excess of $200 million. Lenders do not extend credit against collateral at 100% of their assessed value; they extend it at a loan-to-value ratio that protects them from downside. A $200 million loan against collateral appraised conservatively might require $300 to $400 million in slot value to support the lending decision. The slots delivered that, and they delivered it in April 2020, when zero aircraft were flying from those airports.

JetBlue, simultaneously, raised $750 million using airport slots at JFK and LaGuardia — along with other collateral — as security for a secured term loan. The lenders to JetBlue and American were making the same bet in parallel: that the regulatory right to use these specific runways at these specific times was worth real money regardless of what was currently happening to airline revenues, because that right would return to full productive use when the pandemic ended, and at that point, the slot’s value would reflect the same secondary market prices that had prevailed in 2019.

They were correct. Both loans were repaid. The slots retained their value. The airlines that survived — which most major US carriers did, albeit with significant government assistance — emerged from COVID with their slot portfolios intact.

Subscribe now

The Heathrow Premium: When Slots Are Worth More Than the Planes

In 2012, Japan Airlines (JAL), emerging from a 2010 bankruptcy and restructuring, made a decision that attracted relatively little attention in the Western financial press but that illustrated the economics of slot ownership with extraordinary clarity. JAL, as part of its operational restructuring, chose to exit certain of its London Heathrow slots rather than maintain the routes those slots enabled. The slots were sold in the secondary market. The proceeds were, by JAL’s own accounting, material to its post-bankruptcy balance sheet.

The price paid for those JAL Heathrow slots — disclosed in filings with Japanese regulators — was equivalent to the purchase price of multiple new narrow-body aircraft. The slot rights, which took up no hangar space and required no maintenance, no fuel, no crew, and no insurance, were worth as much as several tangible, physical aircraft. A buyer trying to replicate the route connectivity that those Heathrow slots enabled could not have done so by buying the equivalent aircraft at any price, because without the Heathrow slots, the aircraft had nowhere to land during the peak hours required to make the transatlantic route commercially viable.

This transaction illustrates what the earnings model perpetually obscures: the slot is not merely a tool for the airline to serve a route. It is the precondition for serving the route at all. An aircraft without a Heathrow slot is a plane that cannot land at Heathrow. A carrier without Heathrow slots is a carrier that cannot compete for premium transatlantic traffic through London. The slots come first. The routes, the revenue, and the earnings follow. But an earnings model (e.g. DCF, Relative PE.etc) counts the revenue and prices it with a multiple, completely bypassing the question of whether the precondition for generating that revenue — the slot — is cheap or expensive relative to its replacement cost, which is functionally infinite.

Subscribe now

Delta’s COVID Trough and What a Patient Investor Saw

Return to April 14, 2020. Delta Air Lines trades at $22.46 per share, $14.5 billion total market cap. The company’s December 2019 10-Q impairment test has disclosed:

Domestic slots: $622 million carrying value, fair value 61% to 181% above that. International routes and slots: $2,583 million carrying value, fair value 15% to 29% above that.

The picture that emerges for an investor who is reading the asset footnotes rather than the earnings model is strikingly different from the narrative the market was pricing. The market was pricing: airline industry in crisis, revenues near zero, bankruptcy risk elevated, sell. The asset analysis was pricing: a slot and route portfolio worth $4 to $5 billion by the company’s own fair value testing worth many multiples of the current equity price, in a company that the US government was explicitly supporting through the CARES Act specifically to prevent from failing. The CARES Act Payroll Support Programme provided Delta with $5.4 billion in grants and loans. The government had effectively already decided that Delta would not be permitted to fail.

An investor who had bought Delta at $22 per share in April 2020 and held through the recovery watched the stock reach $52 per share by November 2021 — a 136% return in 19 months. The slot portfolio had not changed. The route network had not changed. What changed was the market’s willingness to price temporary disruption as permanent impairment, and then to reverse that mistake as the disruption proved temporary and the assets proved durable.

Subscribe now

PART V: THE ACCOUNTING GAP — WHY CARRYING VALUE IS ALMOST MEANINGLESS FOR SLOTS

Why Zero Plus Old Acquisition Cost Is Not a Useful Number

The balance sheet representation of an airline’s slot portfolio is, by any honest assessment, nearly useless for determining economic value. It is the sum of two fundamentally different numbers with no common basis of comparison:

The first component is the grandfathered slots — allocated by the FAA at zero cost in 1969, 1985, or 1993, carried at zero on the balance sheet, never revalued regardless of what secondary market prices have done in the intervening decades. A Reagan National slot pair that the FAA allocated for free in 1985 is carried at zero in 2026, despite the fact that comparable slot pairs have transacted at $20 to $40 million in recent secondary market transactions.

The second component is purchased slots — acquired in the secondary market at historical transaction prices, carried at those historical prices as indefinite-lived intangibles and not amortised. A LaGuardia slot pair purchased for $8 million in 2001 is carried at $8 million in 2026, despite the fact that comparable slot pairs have transacted at $15 to $20 million recently. The acquisition cost is fixed in time; the market value has continued to move.

The result is a carrying value that is neither the current market value nor a consistent methodology. It is a historical accident — a combination of regulatory history (which slots were granted for free) and acquisition history (when purchased slots were bought and at what prices) that has no predictive relationship to the portfolio’s current economic value.

Under GAAP (ASC 350 for intangibles), indefinite-lived intangible assets are not amortised but must be tested annually for impairment. The impairment test asks: is the carrying value above the asset’s fair value? If not, no adjustment is made. This is a one-directional test. The carrying value can be written down to fair value if fair value falls below cost, but it cannot be written up if fair value rises above cost. This means that a grandfathered slot carried at zero — whose fair value has risen to $20 million per pair — remains at zero on the balance sheet permanently. The appreciation is invisible, disclosed only obliquely in the impairment test footnote when the company notes that fair value exceeds carrying value by a specified percentage.

Subscribe now

PART VI: CONFIRMED CATALYSTS — WHAT MAKES THE VALUATION GAP IRREVERSIBLE

Catalyst 1: Heathrow Expansion Is Off the Table — Confirmed by Multiple Governments, Multiple Times

The Heathrow expansion debate in the United Kingdom has persisted for decades. A third runway at Heathrow — the single addition that would meaningfully increase slot capacity and potentially reduce slot scarcity — has been proposed, studied, approved, rejected, appealed, and reapproved at various points across forty years of aviation policy. As of 2025, the third runway remains unbuilt and its timeline, if any, extends well beyond a decade.

In the meantime, every year that passes without a new Heathrow runway is a year in which existing Heathrow slots appreciate. The demand for slots — from airlines wanting to serve the world’s most-visited city and Europe’s most important hub — grows with global aviation traffic. The supply of slots is fixed. The price follows the relationship between fixed supply and growing demand.

This is confirmed as a catalyst in the most concrete sense: the physical and regulatory constraints on Heathrow expansion are not speculative future risks. They are documented, litigated, and politically settled realities. There will not be a third Heathrow runway operational in any timeframe that is commercially relevant for current slot holders. The scarcity is permanent within any reasonable investment horizon.

Subscribe now

Catalyst 2: US Airport Slot Controls Are Entrenched by a History of Opposition to Expansion

At Reagan National, the perimeter rule and slot controls have been in place for decades, survived multiple legislative challenges, and are actively defended by the political representatives of the District of Columbia and Northern Virginia — jurisdictions that have no interest in increasing noise, traffic, or disruption around their airport and every interest in maintaining the premium business-travel environment that slot control creates. Every time Congress has considered relaxing slot controls at Reagan National, the opposition from local representatives and the airline incumbents who benefit from existing slots has been sufficient to prevent any substantive change.

The DOT’s process for slot allocation and trading is well-established and creates a liquid secondary market without creating new slot supply. New slots at Reagan National are simply not available. The FAA does not have a mechanism for creating new runway capacity without building new runways, and new runways at Reagan National would require expanding into the Potomac River — which is not on any currently contemplated infrastructure agenda.

It is however, interesting to note, that in the future, new engine technology (which reduces the sound of flights dramatically) may result in an increase in slot supply

Catalyst 3: The Post-COVID Consolidation Has Reduced the Number of Slot Holders

The COVID pandemic, despite its devastation to airline revenues, produced one structural consequence that benefits remaining slot holders: it accelerated the exit of weaker carriers from slot-controlled airports. Airlines that could not survive the revenue shock surrendered slots or sold them to larger carriers at distressed prices. The concentration of slots at the largest US and European carriers increased. Each surviving carrier emerged with a larger proportional share of the total slot pool than it had before — without paying market prices for the incremental slots in every case.

Delta’s slot position at LaGuardia and Reagan National, and American Airlines’ position at the same airports plus Heathrow, is larger and more defensible post-COVID than it was pre-COVID. The secondary market for slot trading has resumed, and prices have recovered to or above pre-COVID levels for prime peak-hour slots at Heathrow and the key US slot-controlled airports.

Subscribe now

Catalyst 4: Regulatory Action Against Slot Hoarding Creates Forced Transfers at Market Prices

Both the FAA and the European Commission enforce “use it or lose it” rules for airport slots: carriers must use their slots a specified minimum percentage of the time (typically 80% under the IATA Worldwide Slot Guidelines) or risk forfeiture. These rules periodically create forced transfers of slots when carriers cannot maintain minimum utilisation — as happened during COVID when regulators temporarily suspended use-it-or-lose-it requirements to prevent mass forfeiture. The resumption of normal use-it-or-lose-it enforcement creates a predictable channel for slot transactions to occur at market prices, documenting and updating the transaction database on which the owner’s valuation is based.

When a slot pair transacts — whether voluntarily between two carriers or as a result of regulatory enforcement — the transaction price enters the public record and updates the market’s understanding of current slot values. Each documented transaction confirms or updates the secondary market price and reduces the uncertainty about what an entire slot portfolio would be worth in a monetisation event.

Subscribe now

Catalyst 5: The Airline Industry’s Continuing Need to Monetise Non-Operational Assets

Major US carriers have repeatedly demonstrated willingness to sell or securitise slot portfolios when capital needs arise. American’s 2020 use of LGA/DCA slots as collateral, JetBlue’s $750 million slot-backed financing, and historical outright slot sales at Heathrow all confirm that the secondary market for slot transactions is liquid, functional, and capable of processing material volumes in short timeframes. The infrastructure for slot monetisation exists. The appetite from buyers — both operating airlines and financial investors — is documented.

As airlines manage their balance sheets in the post-COVID era, with high debt levels from pandemic financing and capital requirements from fleet renewal, the economics of selling non-core slot positions or using peak slot pairs as collateral for lower-cost financing are attractive and proven. Each such transaction at current market prices documents the gap between the carrying value and the economic value — making the arbitrage increasingly visible with each disclosed transaction.

Subscribe now

CONCLUSION: THE ASSET THAT GETS MORE VALUABLE AS THE BALANCE SHEET IGNORES IT

The airport slot story is, in one sense, the simplest value arbitrage in this series of papers. You do not need to understand geological engineering or electrolytic chemistry or regulatory rate-of-return frameworks. You need to understand one physical reality — runways have finite capacity and some airports will never add capacity — and one accounting consequence — the slots allocated free by the government in 1969 will forever be carried at zero, regardless of what they trade for in the secondary market.

Between those two realities lives the arbitrage. The slots appreciate in value as global aviation traffic grows and slot supply remains fixed. The balance sheet shows the same zero it has always shown. The earnings model applies a multiple to current airline earnings, which are cyclically volatile and temporarily terrible during crises, and produces a price target that has no systematic relationship to the standalone value of the most non-replicable asset in the portfolio. And periodically — every seven to twelve years in the history of commercial aviation — a crisis arrives that makes airline earnings terrible, drives the stock price toward the floor, and creates a window in which an investor can buy a slot portfolio worth $3 to $5 billion at the fair values the company’s own accountants were comfortable defending just months earlier, embedded in a company priced at $14 to $15 billion in total equity value.

That window closed in 2021 as COVID receded and airline stocks recovered. It will open again — because it always does, in every sufficiently severe aviation cycle. When it does, the question that will separate investors who earn extraordinary returns from investors who earn ordinary ones is the same question that separated them in April 2020: are you looking at the earnings model, or are you looking at the asset?

The slot is the asset. The earnings are its temporary, cyclically-impaired output. The gap between what the slot is worth in the secondary market and what the balance sheet records is not a mistake or an oversight. It is a mechanical consequence of accounting standards designed for a world in which assets have purchase prices. When the government gives you a slot for free, the accounting records a zero, and the zero persists forever, while the world changes around it.

Delta’s domestic slots: $622 million on the balance sheet, $1.0 to $1.7 billion in Delta’s own fair value testing. American’s LGA/DCA slots: pledged as collateral for $200 million in emergency financing when no one was flying. Virgin’s Heathrow slots: £225 million of debt raised against them in a private bond market. The market kept being told the same thing by the same transactions. It kept looking at the earnings model instead.

This white paper is for educational and informational purposes only. Delta Air Lines impairment test data is drawn from Delta’s Form 10-Q filed with the SEC for the period ended September 30, 2020, Note 4 (Goodwill, Identifiable Intangibles, and Other Assets). American Airlines appraised asset values are drawn from American Airlines Group Inc. Form 8-Ks filed June 12, 2020 and July 23, 2020 with the SEC. Virgin Atlantic’s slot-backed financing is documented in contemporaneous press reports of the December 2015 transaction. JetBlue’s slot-backed financing is documented in ACC Aviation’s published analysis of COVID-era airline capital markets activity. Heathrow slot transaction prices are drawn from publicly disclosed SEC filings and mba Aviation’s published industry analysis. Nothing herein constitutes investment advice.

If you like this read, check out my other writings 👇

This paper applies the same discipline used throughout this (sector value arbitrage) series: identify the primary income-generating asset of the business, value it the way a business owner would — at what it produces today without a financial model — and then compare that to what the accountant records on the balance sheet. The gap between those two numbers is the arbitrage.

In banking, this discipline requires us to ask a question that almost nobody who covers bank stocks asks: what is the liability side of the balance sheet worth as an asset?

That sounds paradoxical. Deposits are liabilities. A bank owes its depositors money. How can a liability be an asset?

The answer is the core of this paper, and once you see it clearly, you cannot look at a community bank balance sheet the same way again. A deposit is a liability in accounting terms. In economic terms, it is a funding source — the raw material the bank uses to make loans. And like every raw material, it has a cost. The difference between what a bank pays for its deposits — sometimes zero, for non-interest-bearing checking accounts — and what it would cost to fund itself from the open market is real money, earned every single day, from an asset that shows up on the balance sheet at exactly zero.

That zero is where the arbitrage lives.

Subscribe now

PART I: THE BUSINESS OF BANKING, STRIPPED TO ITS ESSENTIALS

Where a Bank’s Income Actually Comes From

To understand the deposit as an asset, you first need to understand what banking actually is as a business. Strip away the branches, the ATMs, the digital apps, and the marketing campaigns, and a bank is doing one thing: borrowing money at one rate and lending it at a higher rate. The difference — the spread — is the net interest margin, and it is the source of almost all the income a traditional commercial bank generates.

A community bank in rural Ohio gathers deposits from its customers. Mrs Johnson deposits her savings at 4% per year. Mr Chen deposits his business operating account — a checking account — earning nothing at all, because business checking accounts rarely pay interest. The bank takes those funds and makes loans: mortgages at 7%, business loans at 8%, agricultural loans at 7.5%. The spread between what the bank pays for deposits and what it earns on loans flows through the income statement as net interest income. This is the bank’s operating revenue. It is as simple and as mechanical as a water mill: money flows in one side at one price, flows out the other side at a higher price, and the difference is what the bank earns.

Now here is the question a business owner asks before any other: of all the different types of deposits a bank holds, which one costs the least? And the answer is so obvious that it barely registers as a strategic insight, but its implications are enormous. The cheapest deposit is the one that pays nothing. The non-interest-bearing checking account — the business operating account, the personal current account, the payroll processing account — where the customer deposits money and earns 0% interest in return.

If a bank holds $1 billion in non-interest-bearing checking accounts, it is funding $1 billion of its loan book at a cost of exactly zero. If the 5-year US Treasury note yields 4.5%, then the bank’s $1 billion in zero-cost deposits is generating a funding advantage of $45 million per year over what it would cost to fund the same $1 billion by issuing medium-term notes in the wholesale market. Forty-five million dollars per year in funding advantage. From deposits that cost nothing. From customers who are not going anywhere because they have had their business accounts at this bank for twenty years and switching is a hassle and the banker knows their name.

This is the core deposit intangible. It is not a theoretical concept. It is forty-five million dollars per year falling from the sky, from a relationship asset that the balance sheet records at zero.

Subscribe now

PART II: THE ASSET THE ACCOUNTANT REFUSES TO RECOGNISE

Why GAAP Records the Value of This Franchise at Exactly Nothing

Under US GAAP — specifically ASC 805 (Business Combinations) and the broader framework governing intangible asset recognition — a company can only recognise an intangible asset on its balance sheet when it has purchased that asset from a third party in an arm’s-length transaction. This is called the “purchased intangible” rule, and it has a sensible rationale: intangibles created internally are difficult to measure reliably, so conservative accounting standards require a market transaction to establish a verifiable cost.

For banks, this rule creates an accounting absurdity that compounds year by year. A community bank that has been gathering non-interest-bearing checking accounts from local businesses for sixty years has built something extraordinarily valuable — a stable, sticky, low-cost funding base that no competitor can walk in and replicate tomorrow. It took sixty years of relationships, sixty years of service, sixty years of the local banker attending the Chamber of Commerce breakfast and helping the hardware store owner set up his payroll processing. That franchise is real. It generates measurable economic value every day. It is the primary reason the bank earns a superior net interest margin relative to competitors that fund themselves from brokered deposits or wholesale markets.

But because the bank built this franchise organically — it did not purchase it from anyone — GAAP requires it to record the value of the franchise at zero. The balance sheet says nothing about it. The income statement reflects it implicitly through the net interest margin, but never isolates it or names it. An investor looking at the balance sheet sees: loans, securities, cash, and a small amount of physical premises at depreciated cost. The deposit franchise built internally — the thing that makes the bank worth buying — is invisible.

The moment this same bank is acquired by a competitor, the accounting changes completely. Under ASC 805, the acquirer must perform purchase price allocation and assign fair values to all assets and liabilities acquired, including identifiable intangible assets. The deposit franchise — the very same asset that was invisible the day before the acquisition — is now recognised as a core deposit intangible (CDI) on the acquirer’s consolidated balance sheet. It is valued using a present value methodology that estimates the future cost savings from the acquired deposit base relative to alternative funding sources, discounted to the acquisition date. It is then amortised over its estimated useful life — typically seven to ten years.

One day before the merger: $0 in intangibles for the deposit franchise. One day after the merger: $40 million, $60 million, $100 million in recognised CDI (Core Deposit Intangible) — depending on the size and quality of the deposit base.

Nothing changed about the deposits. Nothing changed about the customers or their relationships with the bank. Nothing changed about the funding cost advantage those deposits represent. The only thing that changed was who owns the bank and what the accounting standards require the new owner to recognise.

This is the gap. It is structural, it is systematic, and it is disclosed — in detail — in every bank merger proxy statement and every acquirer’s first financial statement following a bank acquisition. The analysis that reveals the deposit franchise’s value is not secret or proprietary. It is performed by investment bankers, documented in regulatory filings, and available to any investor who reads beyond the headline financials. But for a standalone bank — one that has not yet been acquired — that analysis is never performed publicly, and its result never appears in the financial statements. The asset is invisible until the moment it is sold.

Subscribe now

PART III: THE OWNER’S VALUATION — WHAT A BANK ACQUIRER ACTUALLY PAYS FOR

The Arithmetic of the Deposit Franchise

A business owner considering the acquisition of a community bank does not start with the P/E ratio or the price-to-book multiple. They start by answering a question that has nothing to do with earnings: what does this bank’s funding cost me, and what would it cost me if I had to replicate this funding base from scratch?

The answer to the first part — current funding cost — is straightforward. The bank discloses its average cost of funds: total interest expense divided by total average interest-bearing liabilities. For a well-run community bank with a strong deposit franchise, this number might be 1.5% to 2.0% on interest-bearing deposits, and 0% on non-interest-bearing deposits. The blended cost of the total deposit base might be 1.0% to 1.5%.

The answer to the second part — what replacement funding would cost — is the current market rate for an equivalent funding source: a short-term Treasury bill, a Federal Home Loan Bank advance, or a brokered CD. If the risk-free rate is 5% and the bank’s blended deposit cost is 1.5%, the funding advantage is 3.5% on every dollar of deposits. On a $1 billion deposit base, this is $35 million per year in recurring, structural, cost-of-funds advantage — before any credit for the bank’s loan underwriting, its fee income, or its operational efficiencies.

This is the CDI in its simplest form: the difference between what the bank pays for its deposits and what it would pay for equivalent funding in the open market, applied to the total deposit base. It is real money. It is ongoing. And for a bank with a genuinely sticky deposit franchise — one built on multi-decade business relationships, payroll processing, and the kind of embedded operational dependency that makes switching both painful and risky — it is as durable as any physical asset in any industry.

Mercer Capital, the pre-eminent independent advisory firm covering bank M&A transactions, has tracked CDI values in bank acquisitions over multiple decades. Their 2025 CDI Update, based on S&P Capital IQ Pro data through mid-September 2025, shows CDI values averaging approximately 2.47% to 2.70% of core deposits in recent whole-bank acquisitions. Their 2024 update showed an average of 2.73%. Their 2023 update showed 2.58%. The historical long-run average from the early 2000s — before the distortions of the post-crisis near-zero interest rate environment — was 2.5% to 3.0%.

What this means in practice: every $1 billion of core deposits a bank holds organically has a transaction value — the value that an acquirer assigns and pays for — of approximately $25 to $30 million in CDI alone. Add to this the deposit premium: the excess of total purchase price over tangible book value, expressed as a percentage of the core deposit base. From 2015 to 2023, deposit premiums in whole-bank acquisitions ranged from 6% to 10% of the core deposit base. On $1 billion of deposits, that is $60 to $100 million in total deposit franchise value above tangible book. Together — CDI plus deposit premium — the total value of the deposit franchise paid by acquirers has historically been 9% to 13% of the total deposit base.

A community bank with $2 billion in total deposits — let us say $600 million of which are non-interest-bearing checking accounts — has an embedded deposit franchise value, as measured by arm’s-length M&A transactions, of approximately $180 to $260 million above its tangible book value. If that bank’s tangible book value is $250 million and it trades at 1.2x tangible book ($300 million market cap), the market is assigning $50 million in value above book. The deposit franchise, by M&A transaction comparables, is worth $180 to $260 million above book. The market is capturing roughly 19 to 28 cents of every dollar of deposit franchise value that an acquirer would actually pay.

That is the arbitrage.

Subscribe now

PART IV: WHY WALL STREET’S TOOLS PRODUCE THE WRONG ANSWER

The Problem with P/E and P/TBV for Banks

When a bank analyst covers a community bank, the standard toolkit involves two metrics: price-to-earnings (P/E) and price-to-tangible-book-value (P/TBV). Both have serious limitations when applied to banks with strong deposit franchises, and both systematically undervalue the banks where the deposit franchise is most valuable.

The P/E ratio has an obvious problem: it uses current earnings as the measure of value. But current earnings are heavily influenced by the current interest rate environment, the current credit cycle, and the current mix of assets and liabilities. A community bank that funded itself in 2021 at near-zero deposit costs while holding mortgages yielding 3% to 4% looks very different in its earnings profile than the same bank in 2024, holding the same mortgages but now paying 2% to 3% on deposits as rates normalised. The P/E ratio in 2024 looks worse than 2021 — but the deposit franchise is unchanged, the customer relationships are unchanged, and the long-term funding advantage of the deposit base, relative to alternative funding sources, may be even more valuable when rates are high. The earnings metric is measuring the outcome of one point in the rate cycle, not the value of the franchise asset. These are not the same thing.

The P/TBV ratio has a more fundamental problem, which is the one this paper is principally about: tangible book value does not include the core deposit intangible for a standalone bank. The “tangible” in tangible book value specifically excludes intangible assets — and since the CDI of a standalone bank is zero (because it was never purchased), the tangible book value does not capture the deposit franchise value at all. A bank trading at 1.0x tangible book value could be extraordinarily cheap if its deposit franchise is worth 50% or 100% of tangible book. A bank trading at 2.0x tangible book could be expensive if it has no deposit franchise to speak of and its funding comes from brokered deposits and FHLB advances.

The analysis that correctly values the deposit franchise is not a financial model. It requires no forecast. It requires three pieces of information: the total deposit base, the composition of the deposit base (what fraction is non-interest-bearing or low-cost core deposits), and the long term average market deposit premium and long term average CDI percentages. Apply those percentages to the deposit base, add to tangible book, and compare to the market price. The resulting ratio — market price to CDI-adjusted tangible book — is the measure that tells you whether you are paying a fair price for the deposit franchise.

This is the owner’s calculation. It is entirely backward-looking and present-tense. No forecast required. No terminal value. No discount rate debate. Just: what are the deposits worth, and what am I paying for them in the public equity market?

Subscribe now

PART V: STORIES FROM THE VAULT — THREE CASE STUDIES IN DEPOSIT FRANCHISE VALUE ARBITRAGE

The Savings and Loan Crisis and the Birth of the Premium Market

The modern understanding of deposit franchise value as an explicitly tradeable, explicitly priced asset was forged in the wreckage of the US Savings and Loan crisis of the 1980s and early 1990s. Between 1986 and 1995, approximately 1,043 savings and loan institutions failed, at a total estimated cost to the federal government of approximately $160 billion. The Resolution Trust Corporation — the federal agency created to manage the disposition of failed thrift assets — was tasked with selling thousands of branch networks, deposit bases, and loan portfolios to healthy acquirers.

In the process of these sales, something important was documented for the first time at scale: buyers were consistently willing to pay meaningful premiums above the book value of the deposit base specifically for the right to assume those deposits. A branch with $100 million in stable savings accounts and checking accounts was worth more than the physical building and the furniture inside it. The difference — the premium for the deposit relationships — was the market price of the core deposit intangible.

The FDIC and RTC transactions of this era produced the first systematic dataset on deposit premiums, and they revealed that buyers considered stable, low-cost deposits valuable enough to pay for even in distressed circumstances. Healthy banks acquiring failed thrift branches would typically pay 3% to 8% of the deposit base as a premium above the deposit liability face value, simply for the right to assume those customer relationships and their associated funding advantage.

This market-clearing evidence established that a bank’s deposit base is not merely a source of funding — it is a franchise with a knowable market price. The crisis created the intellectual infrastructure for valuing something that had previously been invisible, and the valuation methodology that emerged has been refined and updated through every subsequent credit cycle.

The lesson from the S&L era was simple: the market for bank assets has always recognised the deposit franchise as something worth paying for. The public equity market has never consistently done so. The gap between those two realities — the M&A market’s deposit premium and the public equity market’s P/TBV multiple — is where the money has been made by investors who understood what they were actually buying.

Subscribe now

The Washington Mutual Collapse: What Happens When the Deposit Franchise Is Hollow

To understand why a genuine, sticky deposit franchise is valuable, it helps enormously to study its opposite — a bank whose deposit base turned out to be fundamentally fragile, despite appearing healthy on the balance sheet.

Washington Mutual (WaMu) was the largest US savings institution by assets when it failed in September 2008 — the largest bank failure in American history at the time. On its balance sheet, it held approximately $188 billion in assets and had a retail deposit base of approximately $143 billion from roughly 2,200 branches across the United States. By conventional metrics, WaMu looked like a bank with a substantial deposit franchise: large deposit base, broad geographic coverage, tens of millions of retail customers.

But WaMu’s deposit base was not what it appeared. The bank had aggressively expanded its branch network through the 1990s and 2000s into markets where it had no historical customer relationships, competing for deposits through rate promotions and aggressive marketing rather than through the embedded relationships that characterise genuine core deposit franchises. A significant portion of its retail deposits were “hot money” — accounts opened by rate-sensitive customers who would move their funds to any competitor offering 25 basis points more. WaMu’s checking account retention rates, its percentage of non-interest-bearing accounts, and its customer longevity metrics were all weaker than a bank that had spent generations building relationships in a single community.

When the bank’s mortgage losses became public knowledge in mid-2008, depositors began withdrawing funds — not through the slow, orderly run that the franchise-based deposit model is designed to resist, but through a rapid electronic exodus that drained approximately $16.7 billion in deposits over ten days in September 2008. The deposit franchise — which had looked impressive in terms of total size — proved to be brittle because it had been built on rate competitiveness rather than on relationship stickiness.

JPMorgan Chase acquired WaMu’s banking operations from the FDIC on September 25, 2008, for $1.9 billion — effectively paying for the physical branch network and the healthy performing portion of the loan portfolio, but implicitly assigning zero premium to a deposit base that had just demonstrated it was not a genuine franchise. The CDI recorded by JPMorgan in the WaMu acquisition was a fraction of what comparable deposit volumes would have fetched for a bank with genuine community banking relationships.

The lesson from WaMu is the crucial qualifier on everything else in this paper: not all deposits are created equal, and not all deposit bases carry the same franchise value. The CDI premium — 2.5% to 3.0% of core deposits in a typical community bank acquisition — is warranted only for deposits that are genuinely sticky. The diagnostic for stickiness is not deposit volume. It is the proportion of non-interest-bearing checking accounts, the average customer tenure, the rate sensitivity (measured by how much the deposit mix shifted during the 2022 to 2023 rate increases), and the operational embeddedness of the customer relationships (payroll processing, business cash management, trust services that are practically impossible to move without significant disruption).

A community bank in a small Midwestern town whose business customers have been depositing their operating accounts there for thirty years, whose mortgage customers also have checking accounts there, and whose deposits barely moved during the 2022 rate cycle — that bank has a genuine franchise. A bank that has grown its deposits 40% in three years through rate promotions and digital account opening — that bank may have a large deposit base but a small or zero franchise premium.

Subscribe now

First Republic’s Collapse and JPMorgan’s Bargain: The Deposit Franchise Valued in Real Time

On May 1, 2023, JPMorgan Chase acquired substantially all of the assets and assumed substantially all of the deposits and certain liabilities of First Republic Bank from the FDIC, following the California Department of Financial Protection and Innovation’s closure of First Republic on that date. The transaction produced one of the most explicit, documented, and publicly disclosed valuations of a deposit franchise in modern banking history — because JPMorgan’s 8-K filings disclosed the transaction’s mechanics in detail.

First Republic had been a genuinely differentiated bank. Its business model was built on serving high-net-worth individuals and their associated businesses in major US metropolitan markets — San Francisco, Los Angeles, New York, Boston, and Palm Beach. Its customers were wealthy professionals: doctors, lawyers, technology executives, and venture capitalists. Its relationship banking model created genuine stickiness: First Republic’s bankers provided concierge-level service, knew their clients personally, and structured banking relationships that spanned mortgages, business accounts, investment services, and trust accounts. Its non-interest-bearing deposits were a substantial fraction of its total deposit base, funded at zero cost, from customers who valued the relationship more than the interest rate.

This was, in isolation, a valuable franchise. But First Republic had a structural problem that the banking world only fully understood during the 2022 to 2023 rate cycle: it had funded extraordinarily cheap fixed-rate jumbo mortgages using short-duration deposits, creating a duration mismatch that became catastrophically visible when the Federal Reserve raised rates from near zero to 5% in fourteen months. As its fixed-rate mortgage assets remained anchored at low yields, its deposit costs rose and its available-for-sale securities portfolio was marked to market at devastating losses. When Silicon Valley Bank’s failure in March 2023 triggered a flight to safety among uninsured depositors broadly, First Republic — which had a high proportion of uninsured deposits given the wealth of its customers — lost approximately $100 billion in deposits in days. The franchise was still real; the balance sheet was insolvent.

JPMorgan’s 8-K filed in conjunction with the May 1, 2023 acquisition disclosed that the transaction included approximately $92 billion in deposits and approximately $173 billion in loans and $30 billion in securities. JPMorgan paid $10.6 billion to the FDIC and received $50 billion in FDIC-provided term financing plus loss-share agreements on the loan portfolio. The post-closing balance sheet disclosed by JPMorgan included $1.1 billion in intangibles — the CDI recognised on acquisition — for approximately $92 billion of acquired deposits. That CDI represents approximately 1.2% of the acquired deposit base, notably below the typical 2.5% to 3.0% for a healthy community bank, reflecting the reality that some of those deposits were the most rate-sensitive in First Republic’s portfolio and the customer relationships, while genuine, had been stress-tested severely.

The transaction was immediately accretive. JPMorgan disclosed that the acquisition was expected to generate more than $500 million of incremental net income per year — entirely from the interest rate spread on the acquired loan portfolio and the funding advantage of assuming $92 billion in deposits at their contractual cost rather than at wholesale market rates. JPMorgan also recognised a bargain purchase gain of $2.7 billion — the excess of the fair value of assets acquired over the consideration paid — reflecting the FDIC’s urgency in resolving the situation and JPMorgan’s willingness and ability to absorb the complexity.

The bargain purchase gain tells you something important. JPMorgan — the most financially sophisticated bank acquirer in the world, with the largest balance sheet and the most analytical resources — looked at First Republic’s deposits, its loan portfolio, and its customer relationships, and concluded that the whole package was worth materially more than what the FDIC was charging for it. Even in a distressed, rushed, weekend-auction context, the deposit franchise had enough residual value that the acquirer received a gain simply by assuming it. The deposit base was not worthless just because the institution had failed. It had failed because of an asset-liability mismatch problem, not because the customers were gone or the relationships were severed.

The public equity investor who held First Republic shares through 2022 — watching the stock fall from its mid-2021 peak of approximately $220 per share to $20 per share by early 2023 — was watching the market price in terminal failure risk. The market was correct that the institution was insolvent on a mark-to-market basis. But the market was not pricing the deposit franchise separately from the institution. An owner who could have carved out only the deposit relationships and customer franchise — removing the problem loan book and the duration mismatch — would have found something worth billions. The equity market, unable to separate the franchise from the flawed balance sheet structure, simply sold both together at zero.

Subscribe now

The Community Bank in the Ozarks: Sixty Years of Nothing on the Balance Sheet

The most instructive case study for the pure deposit franchise arbitrage is not a large bank or a failed institution. It is the community bank in a secondary market that has never made headlines, never had an investment banking relationship, and never been the subject of an analyst report. There are thousands of them across America — in rural Oklahoma, in small-town Texas, in Appalachian Kentucky — and they represent the clearest version of the asset-valuation gap described in this paper.

Consider a $1.5 billion community bank in a mid-sized regional market, founded in 1962. Over sixty years, it has gathered a deposit base that is heavily weighted toward non-interest-bearing checking accounts: local businesses, agricultural operations, local governments, and long-tenure retail customers who have maintained their accounts for thirty or forty years and will not change banks for 25 basis points. Non-interest-bearing demand deposits constitute approximately 35% of its $1.3 billion total deposit base — roughly $455 million in zero-cost funding. The remaining deposits carry an average cost of perhaps 1.8% in a normalised rate environment.

The bank’s blended deposit cost is approximately 1.2%. The one-year Treasury bill in a normalised environment yields approximately 4.5%. The funding advantage — the annual economic value of the deposit franchise — is approximately 3.3% on $1.3 billion, or approximately $43 million per year.

The bank’s tangible book value is $185 million. It trades in thin, infrequent market transactions at approximately 1.1x tangible book — about $200 million in market capitalisation. At $200 million, you are paying roughly $154 per $1,000 in deposits for the whole bank.

Now apply the M&A market’s pricing. CDI at the historical average of 2.7%: $1.3 billion × 2.7% = $35.1 million. Deposit premium at the historical average of 7%: $1.3 billion × 7% = $91 million. Total deposit franchise value above tangible book: $126 million. Against tangible book of $185 million, the CDI-adjusted tangible book value of this bank is approximately $311 million. It is trading at $200 million. You are buying a verified, documented, M&A-comparable deposit franchise at a 35% discount to what a strategic acquirer would pay for it in cash.

Importantly, this is not a forecast of what the bank might earn in the future. It is not a model of what the deposit franchise might be worth under various rate scenarios. It is what another business owner would pay for now. The analysis is entirely backward-looking and present-tense. No forecast required.

The bank’s management that discloses the deposit composition in detail. The total deposits, the non-interest-bearing portion, the time deposit portion, and the interest cost of each category are all public information, reported with 30 to 45 days of lag. The CDI and deposit premium percentages come from Mercer Capital’s annual CDI update, which is published and freely available. Any investor with internet access and a calculator can perform this analysis in approximately twenty minutes. But almost no one does, because the attention is on the earnings per share, the price-to-book ratio, and whether the bank is going to grow its loan portfolio. The franchise is invisible because the accountant says it is zero, and the analyst’s model starts with the accountant’s number.

Subscribe now

PART VI: THE RATE CYCLE AND THE CDI — HOW INTEREST RATES REVEAL THE FRANCHISE

2022 to 2023: The Great Stress Test of Every Deposit Franchise in America

The Federal Reserve’s rate-hiking cycle from March 2022 to July 2023 — which took the Federal Funds Rate from near zero to 5.25 to 5.50% in sixteen months — was the most aggressive monetary tightening in forty years. For bank investors, it was also the largest real-world stress test of deposit franchise quality ever administered in a single concentrated period. No theoretical model, no earnings projection, and no historical scenario analysis produced a more accurate measure of which deposit franchises were real and which were not than simply watching what happened to each bank’s deposit base between Q1 2022 and Q4 2023.

Banks with genuine core deposit franchises — strong non-interest-bearing proportions, deeply embedded business relationships, customers whose deposits were operationally necessary rather than rate-driven — experienced moderate deposit repricing. Their customers, who had been depositing money at 0% or near-0% for years, upgraded some portion of their savings to money market accounts, but retained their operating accounts and maintained their primary banking relationships. These banks saw their non-interest-bearing deposit percentage decline from, say, 35% to 25% of total deposits — a meaningful shift, but not a collapse. Their cost of deposits rose, but their franchise proved durable.

Banks without genuine franchises — those whose deposit growth had been driven by rate promotions, digital acquisition channels, or wholesale brokered deposits — experienced severe outflows. When the rate environment changed and competitors offered better yields, rate-sensitive customers left. These banks were forced to replace outflowing deposits with FHLB advances, brokered CDs, and other wholesale funding at market rates — eliminating the funding cost advantage that had previously made their net interest margins appear attractive.

The rate cycle did not create this distinction. It revealed it. The franchise quality difference between a genuine community bank with sixty years of embedded relationships and a newer, rate-driven deposit gatherer existed before March 2022. It just was not visible until the stress test exposed it.

For the investor using the CDI-adjusted tangible book methodology, the rate cycle provides the most important diagnostic tool available: what actually happened to deposit volumes and non-interest-bearing ratios during the 2022 to 2023 period? A bank whose non-interest-bearing deposits declined by less than 20% from their peak during the cycle, and whose total deposit volumes remained stable, has a demonstrated, stress-tested franchise. A bank whose non-interest-bearing deposits halved and whose total deposits declined materially has told you directly that its franchise is weaker than the balance sheet’s geographic breadth might suggest.

The 2022 to 2023 cycle also interacted with CDI values in a way that the Mercer Capital data documents clearly. CDI values — which are driven by the spread between a bank’s deposit cost and the prevailing market interest rate — were artificially depressed during the 2021 near-zero rate environment (averaging only 0.63% for all of 2021, per Mercer Capital’s data) because the spread between zero-cost deposits and near-zero market rates was minimal. As rates rose and the spread between core deposit cost and market rates widened dramatically, CDI values surged: averaging 1.61% for all of 2022, 2.58% for 2023, and 2.73% to 2.74% for 2024. The deposit franchise became more valuable as interest rates rose, because the economic advantage of holding low-cost deposits relative to the market rate became larger.

This creates a specific investment dynamic that the P/E ratio cannot capture. Community bank stocks often decline in rate-rising environments because earnings are initially compressed — asset yields reprice slower than liability costs for many banks, temporarily squeezing net interest margins. The P/E ratio rises as earnings fall. Investors see an “expensive” stock. But the underlying deposit franchise has become more valuable in present-value terms, because the funding cost advantage has widened. The P/E ratio is pointing in the wrong direction at exactly the moment the franchise value is increasing most rapidly.

Subscribe now

PART VII: HOW TO VALUE THE DEPOSIT FRANCHISE — THE OWNER’S FRAMEWORK

Four Steps, No Model Required

The business owner’s analysis of a community bank’s deposit franchise does not require a financial model, a DCF, or any forecast of future interest rates. It requires four pieces of information, all publicly available:

Step one: total deposit base and its composition. Deposits disclosed by US banks are mainly separated into 2 categories: non-interest-bearing demand deposits, interest-bearing demand deposits, savings accounts, time deposits under $250,000, and time deposits over $250,000. This breakdown tells you the composition of the funding base. The non-interest-bearing demand deposits are the highest-quality funding — the franchise’s crown jewel. The time deposits over $250,000 are the lowest quality — the most rate-sensitive, least sticky, most likely to reprice or leave.

Step two: the bank’s actual deposit cost versus the risk-free rate. The fillings of the company should disclose the total interest expense on deposits. Divide this by average deposits and you have the bank’s actual blended deposit cost. Compare this to the current federal funds rate or the one-year Treasury yield. The spread between these two numbers is the current annual funding advantage per dollar of deposits. This is the CDI’s cash flow, calculated at current rates, with no forecast.

Step three: apply the M&A-documented CDI and deposit premium percentages to the core deposit base (note that you can even skip this step to be safe). Using Mercer Capital’s annual CDI Update, take the current CDI average (approximately 2.5% to 2.7% as of 2024 to 2025) and the average deposit premium range (6% to 10%) and apply them to the bank’s core deposit base. This produces the deposit franchise’s M&A-comparable value. Add this to tangible book value to produce the CDI-adjusted tangible book value.

Step four: compare the market price to the CDI-adjusted tangible book. If the market is pricing the bank at 1.1x or 1.2x unadjusted tangible book, but at 0.7x to 0.9x CDI-adjusted tangible book, the market is not correctly pricing the franchise. The discount to CDI-adjusted tangible book represents the portion of the franchise value that the market is giving you for free.

Subscribe now

PART VIII: CONFIRMED CATALYSTS — WHY THE FRANCHISE VALUE GETS RECOGNISED

Catalyst 1: The Bank M&A Wave Is Already in Motion and Structurally Accelerating

Community bank consolidation in the United States has been accelerating for decades and shows no sign of reversing. There were approximately 14,500 FDIC-insured institutions in 1990. By 2024, that number had fallen below 4,600 — a 68% reduction over 34 years, driven almost entirely by voluntary mergers rather than failures. The pace of consolidation shows no sign of slowing.

When a community bank is acquired — voluntarily, not in distress — the deposit franchise is valued explicitly and paid for in cash. The CDI is recognised on the acquirer’s books. The deposit premium is paid to the target’s shareholders. Every merger proxy statement walks through exactly this analysis, disclosed in the fairness opinion section where the investment banker explains what the deposit franchise is worth and why the offer price is fair.

For shareholders of community banks in consolidating markets — which is effectively every market in the United States — the M&A catalyst is not speculative. It is a statistical certainty over a five to ten year holding period for most community bank investments. The question is not whether the deposit franchise will eventually be valued at M&A comparable prices. It is when.

This creates a specific investment framework: identify community banks trading at significant discounts to CDI-adjusted tangible book, with demonstrated franchise quality (stable non-interest-bearing proportions through the 2022 to 2023 rate cycle), located in markets with active M&A consolidation, and hold them. The catalyst — a merger announcement at a price that reflects the deposit franchise’s full economical value — is a confirmed, documented, recurring event in the US banking industry at a rate of approximately 200 to 300 transactions per year.

Catalyst 2: The Silicon Valley Bank Aftermath Raised the Price of Safe Deposits — The Quality Of Your Deposit

The failures of Silicon Valley Bank, Signature Bank, and First Republic in March and May 2023 changed something fundamental about how bank regulators and large depositors think about banking relationships. The speed with which uninsured deposits fled these institutions — enabled by digital banking and social media — demonstrated that uninsured deposits at certain institution types were not as stable as previously assumed. The regulatory response included elevated scrutiny of uninsured deposit concentrations, new reporting requirements, and ongoing discussions about extending deposit insurance to business operating accounts.

The institutional memory of 2023 has created incremental demand for insured, relationship-based deposit relationships — the kind that community banks with genuine franchises provide — from business customers who previously would have moved freely between institutions for marginal rate differences. A business owner who watched colleagues scramble to move payroll deposits out of a failing bank in March 2023 is now more likely to value the stability and insurance coverage of a relationship-based community bank, even at a modest cost in yield. This structural shift — away from rate-driven deposit gathering toward franchise-relationship banking — advantages exactly the institutions whose deposit franchises are most undervalued by conventional financial metrics.

Subscribe now

Catalyst 3: Rising Rates Have Widened CDI Spreads, Making Franchises More Valuable — Before the Market Has Caught Up

As documented above, CDI values rise when the spread between core deposit costs and market interest rates widens. The post-2022 rate environment has been the most favourable for CDI valuations in two decades. Banks with strong non-interest-bearing deposit franchises — which continued to cost zero to 0.5% even as the federal funds rate reached 5.25% — were funding their loan books at costs dramatically below market rates, generating extraordinary net interest margins from a funding advantage that flows entirely from the franchise.

The public equity market, focused on near-term earnings compression from the initial rate repricing, did not fully reward these banks in 2022 and 2023. Investors saw deposit betas (how much a bank changes the interest rates it pays on deposits in response to changes in central bank policy rates) rising, net interest margins compressing, and unrealised losses on bond portfolios. They sold. But for the banks with genuine franchises — those whose deposit betas were low because their customers were not moving — the compression was temporary and the underlying franchise value was growing. The M&A market, which prices deposits on CDI and premium percentages, has recognised this by paying CDI values in 2024 of 2.73% — well above the 0.63% of 2021. The public equity market has been slower to reprice.

Subscribe now

CONCLUSION: THE FRANCHISE THAT LIVES IN THE FOOTNOTE

The core deposit intangible is, in one sense, the simplest asset in this series of papers. You do not need geological engineering to measure it. You do not need thermodynamic calculations. You do not need to understand the Hall-Héroult process or the unit of production depletion method.

You need to know one thing: what a business owner pays for deposits, versus what the stock market is charging for those same deposits in the public equity market. The gap between those two numbers — is the arbitrage. It has existed consistently, in every interest rate environment, across every credit cycle. It exists because GAAP requires a transaction to occur before an intangible can be recognised, and because the public equity market takes the GAAP balance sheet as its starting point rather than performing the owner’s analysis from scratch.

The woman who built a career in community banking, the bank board member who has watched three generations of a family move their accounts to the same institution, and the regional bank M&A advisor who prices deposit franchises for a living — they all know what the deposit base is worth. They have known it for decades. The number is not hidden. It is in every bank merger proxy statement. It is disclosed in the footnotes of fillings.

The public equity market simply does not look for it. It looks for P/E, P/TBV, and earnings per share. And in looking for those things, it consistently assigns zero value to the most durable, most irreplaceable, most structurally advantaged asset a community bank possesses: the relationship it has built, over decades, with the business owners, farmers, and families who have trusted it with their money and their operating accounts, and who will be back next year, and the year after that, and the year after that.

That relationship costs nothing to maintain. It generates tens of millions of dollars per year in funding advantage. It is worth 6% to 10% of deposits in cash in the M&A market. And on the balance sheet, it is worth exactly zero.

That zero is where the arbitrage is.

This white paper is for educational and informational purposes only. CDI and deposit premium data cited is drawn from Mercer Capital’s publicly available Core Deposit Intangibles Update series (2017 through 2025), based on S&P Capital IQ Pro data. JPMorgan Chase acquisition data is drawn from JPMorgan’s Form 8-K filings with the SEC dated May 1, 2023 and the company’s Q2 2023 earnings presentation filed with the SEC. FDIC deposit composition data is publicly available through the FDIC’s Statistics on Depository Institutions database. Nothing herein constitutes investment advice.

If you like this read, check out my other writings 👇

In the 1998 Berkshire Hathaway Annual Report — published the same year Berkshire acquired General Re for $22 billion and suddenly became one of the largest insurance enterprises on earth — Buffett wrote this in his letter to shareholders:

“Impressive as the growth in our float has been — 25.4% compounded annually — what really counts is the cost of this item. If that becomes too high, growth in float becomes a curse rather than a blessing. At Berkshire, the news is all good: Our average cost over the 32 years has been well under zero.”

Read that last sentence again. Thirty-two years of average cost of float that was not merely low, not merely zero, but below zero. Buffett was not describing an asset on Berkshire’s balance sheet. He was describing something that does not appear as an asset anywhere in the financial statements. He was describing a pool of capital — belonging legally to policyholders, recorded as a liability — that had effectively paid Berkshire to hold it, for three decades, at growing scale.

By the end of 2024, that float had grown to $171 billion. Berkshire had operated at an underwriting profit for 14 consecutive years, accumulating $28 billion in pre-tax underwriting gains over that period alone. In the same two decades, float had grown from $46 billion to $171 billion. The $171 billion sits on Berkshire’s balance sheet as a liability — as “unpaid losses and loss adjustment expenses” and “unearned premiums” — because that is where accounting standards require it to sit. The $171 billion generates investment income every year from being deployed into stocks, bonds, and whole businesses. That investment income flows through the income statement as earnings. But the capital pool itself — the $171 billion that Berkshire does not own — appears as something owed to others rather than something of value to the owner.

This is the central intellectual puzzle of insurance finance, and it is the source of the value arbitrage this paper describes.

Subscribe now

PART I: THE BUSINESS MODEL FROM THE GROUND UP

Why Insurance Has a Different Clock Than Every Other Business

Most businesses operate on a simple financial rhythm. You spend money to make something. You sell it. The revenue arrives. You count the profit. The sequence is: cost first, revenue close behind. A factory that makes cars knows exactly what the car cost to build by the time the customer drives it off the lot.

Insurance runs the sequence entirely backwards, and this reversal creates everything important about the economics. You collect the premium first — typically the full annual premium upfront on day one of the policy. Then, if a claim occurs, you pay it later. For a homeowner’s insurance policy on a house that does not burn down, you collect the premium and pay nothing. For a workers’ compensation policy on an employee who develops occupational disease thirty years into retirement, you collect the premium in 1995 and are still paying claims in 2025. The gap between collection and payment can be a few weeks for a car fender-bender or several decades for asbestos liability. The 2024 Berkshire annual report noted: “We are still making substantial payments on asbestos exposures that occurred 50 or more years ago.”

During that entire gap — whether it is three months or fifty years — the insurance company holds and invests the premium. The float is the aggregate of all those gaps across all active policies. It is not a fund set aside for a specific purpose. It is the natural consequence of the timing mismatch between premium receipt and claim payment. Every insurance company has float. The question is how much it costs them to maintain it.

Subscribe now

The Cost of Float: The Only Number That Distinguishes a Good Insurance Business from a Bad One

The cost of float is the most important metric in insurance analysis, and it is the metric that conventional financial models handle most poorly. It is calculated as follows: take the insurer’s underwriting result for the year — the combined ratio minus 100%, expressed as a profit or loss as a percentage of earned premiums — and divide by the average float. If the insurer ran at a combined ratio of 97% (meaning it paid out 97 cents in claims and expenses for every dollar of premium collected), the underwriting profit was 3% of premiums. If that 3% profit represented $900 million against an average float of $60 billion, the cost of float was negative 1.5%. The float not only cost nothing — the policyholders effectively paid the insurer $900 million to hold their money.

Compare this to the cost of any other form of capital. A company issuing ten-year corporate bonds in 2024 might pay 5% to 5.5% per year in interest. A company drawing on a revolving credit facility pays the bank’s spread above SOFR. An equity issuance dilutes existing shareholders at the current cost of equity, which might be 8% to 10% in a normalised rate environment. Float from a well-underwritten insurance operation costs, at its best, less than zero. It costs, at the industry average, approximately 3% to 5% — which is roughly equivalent to investment-grade corporate borrowing. And it comes with a distinctive characteristic that no other form of capital provides: it does not need to be repaid on any fixed schedule. Claims are paid when they occur. The float simply revolves continuously, with new premiums replacing paid claims in an endless cycle.

Subscribe now

PART II: WHY THE ACCOUNTING SYSTEM GETS THIS EXACTLY WRONG

A Liability That Behaves Like an Asset

Under US GAAP (ASC 944 for insurance entities) and IFRS (IFRS 17, which took effect in 2023), insurance liabilities are measured at their projected future settlement values. An insurer that has collected $100 million in premiums on a portfolio of long-tail liability policies will carry approximately $100 million (plus or minus actuarial adjustments) in reserves on the balance sheet. The standard requires the liability to reflect the expected undiscounted value of future payments.

Here is the accounting absurdity: a $100 million loss reserve that will not be paid for ten years has a present value of approximately $67 million at a 4% discount rate. The economic truth is that the insurer holds $100 million today but only needs $67 million to cover the obligation — the remaining $33 million is the investment income that will be earned by investing the $100 million over ten years. The balance sheet shows the liability at $100 million and makes no provision for the fact that the $100 million of float, invested for ten years at 4%, will grow to approximately $148 million — covering the $100 million obligation with $48 million left over.

To restate the arithmetic with precision: if you take $67 million today and compound it at 4% per annum for ten years, you arrive at exactly $100 million. The balance sheet liability is $100 million. The present-value reality of what you need today to cover that liability is $67 million. The difference — $33 million — is the off-balance-sheet investment income that the float will generate between now and the payment date. It belongs economically to the insurer. It does not appear as an asset anywhere in the financial statements. It is invisible.

This invisibility compounds across an insurer’s entire reserve portfolio. For a large property-casualty insurer with $30 billion in loss reserves and an average payment duration of five to eight years, the present-value discount embedded in the undiscounted reserves might represent $3 to $6 billion in off-balance-sheet economic value — investment income the company will earn on money that the balance sheet pretends it has already spent.

Subscribe now

PART III: THE WALL STREET PROBLEM — WHY CONVENTIONAL METRICS MISS THE FLOAT

What the Analyst’s Model Looks Like

When an equity analyst covers an insurance company, the standard model takes the form of a combined ratio forecast, a premium growth assumption, and an investment yield assumption. The analyst projects: earned premiums of $X, a combined ratio of Y%, generating underwriting profit of $Z, plus investment income on the portfolio of $W. Total pre-tax earnings are $Z + $W, taxed at the effective rate, producing net income and then an earnings per share figure. Apply a sector P/E multiple — typically 10 to 14 times for property-casualty insurers — and you arrive at a price target.

This model treats the float entirely as the source of investment income rather than as the asset itself. The analyst counts the income the float generates but never asks: what is the float worth as a capital source, independent of this year’s investment yield? The float generates $W in investment income this year at whatever the current investment yield happens to be. But the float is also $10 billion or $30 billion or $171 billion in permanent, revolving capital that does not appear as equity and does not cost what equity costs. The P/E model applies a multiple to current earnings, which depend on current interest rates and current underwriting conditions, and produces a price. It does not value the float as a structural capital advantage.

The Price-to-Book ratio, similarly, fails because book value for an insurer equals the investment portfolio plus other assets minus all liabilities, including the full undiscounted loss reserves. An insurer with $30 billion in undiscounted reserves that are economically worth only $23 billion at present value has $7 billion in off-balance-sheet economic assets that reduce the effective cost of its capital structure — and none of that appears in the P/B calculation. The P/B ratio is telling you the price relative to shareholders’ equity after all the accounting deductions. It is not telling you the price relative to the economic capital structure, which includes the float valued at its cost advantage.

The single metric that most clearly reveals whether an insurer’s float represents value or a burden is one that almost no sell-side analyst includes in a standard model: the ratio of cost of float to the current risk-free rate. If an insurer’s cost of float is 2% and the ten-year Treasury yields 4.5%, the float is effectively leverage at less than half the cost of the risk-free rate. If the cost of float is 5% and the ten-year yields 4.5%, the float is costing slightly more than the risk-free rate — it is still useful, but the advantage is thin. If the cost of float is persistently 8% to 10%, the float is destroying value.

An owner buying an insurance company does not primarily ask: what will the earnings per share be next year? They ask: what is the cost of float, how large is the float, and how durably low is that cost? The answers to those three questions determine the economic value of the insurance franchise, independent of any forecast.

Subscribe now

PART IV: STORIES FROM THE UNDERWRITING ROOM — FOUR CASE STUDIES IN FLOAT VALUE ARBITRAGE

The $8.6 Million Purchase That Changed Everything: National Indemnity, 1967

In March 1967, Berkshire Hathaway paid $8.6 million to acquire National Indemnity Company, a specialty commercial auto and general liability insurer based in Omaha, Nebraska. The seller was Jack Ringwalt, who had founded the company in 1940 and had built it into a respectable regional insurer over 27 years. The purchase price was modest. The acquired float was small — a few tens of millions of dollars. The transaction attracted no significant attention from the financial press.

Fifty-seven years later, National Indemnity is the legal entity through which Berkshire Hathaway holds most of its investment portfolio. It is the structural heart of an operation that holds $171 billion in float, has produced $28 billion in pre-tax underwriting profits over the past 14 years alone, and generates investment income on a pool of capital that includes Apple shares, short-term Treasury bills, and dozens of wholly owned businesses. The $8.6 million purchase price has compounded into a contribution to Berkshire’s enterprise that is measured in the hundreds of billions.

The insight that made this possible was not about National Indemnity’s earnings in 1967. It was about the structure of the insurance business model itself. Float, properly managed, is a permanent, revolving source of investment capital that the owner does not have to contribute from their own pocket. Every dollar of float is a dollar that can be invested. If the float grows, the investable pool grows. If the underwriting discipline holds, the cost of float stays low or negative. The earnings model in 1967 would have valued National Indemnity based on its underwriting results and investment income — a relatively ordinary small insurance company by those measures. The owner’s analysis, applied by Buffett, valued it as a platform for accumulating permanently revolving low-cost investment capital. That is a fundamentally different asset.

The 2024 Berkshire annual report noted that over the past two decades, Berkshire’s insurance operations generated $32 billion in after-tax profits from underwriting — roughly 3.3 cents per dollar of premiums written — while float grew from $46 billion to $171 billion. The investment income on $171 billion at even 4% is $6.84 billion per year, on capital that Berkshire does not own. If you add the underwriting profit — which is currently substantial — to the investment income on float, you are looking at a combined economic contribution from the insurance operation that reflects an asset no GAAP balance sheet shows in its true form: a $171 billion permanent, revolving, free or negative-cost pool of investment capital.

Subscribe now

The General Re Disaster and the Lesson About Float Quality

Not all float is created equal. The General Re acquisition in 1998, for $22 billion in Berkshire stock, taught a lesson about float quality that is worth examining in detail — because it demonstrates that float from careless underwriting is not merely unprofitable but actively destructive, and that the earnings model is particularly poorly suited to identifying this distinction.

Berkshire acquired General Re — then one of the world’s largest and most prestigious reinsurance companies — in late 1998. The strategic rationale was clear and correct: General Re held approximately $15 billion in float, providing Berkshire with an enormous incremental investment pool. The business was respectable, well-regarded, and apparently conservatively run.

Within months of the acquisition, the problems emerged. General Re had been running a combined ratio that looked acceptable in the years leading up to the acquisition but concealed deteriorating reserve adequacy in its long-tail liability lines.

Long-tail liabilities in insurance are claims that take a significant amount of time—often years or decades—to manifest, report, and settle after the initial insured event occurs.

In 1999, underwriting losses ballooned to $1.2 billion on $6.9 billion in premiums. Losses continued in 2000. Then September 11, 2001 added $1.9 billion in losses directly related to the terrorist attack. Berkshire’s 2000 annual report acknowledged a float cost of 6% for the year — meaning the combined underwriting losses had turned the float from a free capital source into capital that cost 6% annually, more than risk-free bond rates at the time.

What the earnings model had shown before the acquisition was: a well-regarded reinsurer with a long operating history, reasonable combined ratios, and a large and growing float. What the owner’s analysis should have probed was: how were those combined ratios constructed? Were the loss reserves adequately funded for the tail risk in the portfolio? Was the pricing discipline genuine or was it underwriting results being smoothed by releasing prior-year reserves?

Buffett’s 2000 letter was frank about what happened: General Re had been underpricing risk before the acquisition, and the shortfall was only becoming apparent years later as long-tail claims developed. The float was large but its cost was rising as the inadequate underwriting was revealed. The 2001 letter acknowledged that corrective action — repricing policies, non-renewing unprofitable business, strengthening reserves — had begun, but that the repair would take years.

By 2016, the repair was complete. General Re’s float stood at $17.7 billion and produced an underwriting profit of $190 million. The same float that had cost Berkshire 6% in 2000 was generating a positive return by 2016. The underlying economics of the float — as free, revolving investment capital — were always present. The problem had been the underwriting discipline that determines the cost of that float. Once the underwriting was corrected, the float’s structural advantage reasserted itself.

The lesson for investors evaluating an insurance company’s float is precise: the size of the float matters, but the quality of the underwriting that produces it matters more. An insurer running a combined ratio of 104% on $50 billion of float is not generating free leverage — it is paying 4% above zero for the right to invest other people’s money. An insurer running at 97% on $10 billion of float is generating negative-cost leverage at 3% profit on the float. The second is the better business by a wide margin, even though it has five times less float. The earnings model, comparing the two companies by P/E or EV/EBITDA, might prefer the first because the absolute investment income from a larger float produces higher reported earnings. The owner’s analysis of the float quality prefers the second unambiguously.

Subscribe now

GEICO: The Float Machine That Was Always There, Waiting to Be Capitalised

GEICO’s story is the best available illustration of how a business can have a structurally extraordinary float advantage that the market completely fails to price because the earnings model anchors on current underwriting results rather than on the long-term economics of the float franchise.

By the early 1970s, GEICO — Government Employees Insurance Company — had built a large and profitable auto insurance operation selling directly to government employees and military personnel, primarily by mail. Its direct-to-consumer model eliminated the broker commissions that added 15% to 20% to the cost of policies sold through agents, giving GEICO a structural cost advantage that no conventional agency-distribution competitor could match. In 1972, it was highly profitable and growing rapidly.

Then, in the mid-1970s, GEICO nearly destroyed itself through a combination of rapid expansion into unfamiliar markets, inadequate pricing discipline, and catastrophically inadequate loss reserves. The company had been booking inadequate reserves for its liability policies — presenting better underwriting results than the underlying business actually supported. By 1976, the true extent of the underreserving was becoming clear. GEICO was technically insolvent, its share price had collapsed from $61 to $2, and most of the financial community had concluded it was finished.

Warren Buffett had started buying GEICO at around $2 per share in 1976. By 1980, Berkshire’s stake had been built at an average cost that would prove to be a fraction of GEICO’s eventual value. In 1996, Berkshire paid approximately $2.3 billion for the remaining 49% of GEICO it did not yet own — valuing the whole company at roughly $4.7 billion.

At the point of acquisition, the earnings model for GEICO would have shown: an auto insurer with growing premiums, improving combined ratios, and competitive investment yields on a growing reserve portfolio. A reasonable P/E or Price-to-Book multiple applied to those numbers would have produced a valuation in the several-billion-dollar range — which was broadly what Berkshire paid.

But what was the float worth as a structural asset? In 1995, the year before full acquisition, GEICO’s float was approximately $2.9 billion. The combined ratio in recent years had averaged around 96% to 98%, meaning GEICO was generating a modest underwriting profit and its float cost was slightly negative to modestly positive. More importantly, GEICO’s direct-distribution model — no agents, no commissions, lower cost per policy — gave it a structural advantage that would allow it to grow float profitably over time, because it could price more competitively than agency-channel competitors while maintaining similar or better underwriting margins.

Since 1996, GEICO has grown from roughly 3 million policies to over 18 million, from $2.9 billion in float to a float of approximately $26 billion within Berkshire’s consolidated insurance operation. GEICO’s underwriting profit in 2024 was approximately $7.8 billion — a single year’s underwriting profit that exceeds the total acquisition price Berkshire paid for the entire company in 1996. The structural float advantage — the direct distribution model producing lower costs that enable both competitive pricing and underwriting profitability — was visible in the economics in 1996, even if the earnings model priced only the current results. The owner who priced the float franchise rather than the current earnings bought an asset worth many times the purchase price.

Subscribe now

The Long-Tail Problem and How Patience Becomes Return

There is a category of insurance float that is particularly misunderstood by conventional financial analysis because the income it produces does not appear in the short-term results that earnings models emphasise. Long-tail liability insurance — workers’ compensation, medical malpractice, environmental liability, asbestos claims — has payment periods that routinely extend 20 to 30 years from the date the original policy was written. The float from these policies sits on the balance sheet for decades, generating investment income year after year, while the corresponding liability reserve gradually declines as claims are paid.

An investor looking at a specialty insurer with a large book of workers’ compensation insurance would see, in the earnings model: premium income from current-year policies, investment income from the portfolio, and claims payments on both current and prior-year policies. What the model does not highlight is the extraordinary present-value advantage embedded in the long-tail reserves. A workers’ compensation reserve of $1 billion, representing claims that will be paid out over 25 years at an average duration of 12 years, has a present value of approximately $620 million at 4% discount. The insurer holds $1 billion in investments against a $1 billion liability but only needs $620 million to fund the obligation in present-value terms. The $380 million difference represents investment income the insurer will earn over the next 12 years — investment income on float that belongs on the liability side of the balance sheet but generates all its value on the asset side.

The owners who understand this have structured their insurance businesses to maximise long-tail float — not because long-tail claims are easier to manage (they are not), but because the duration of the float is itself a source of value. Buffett noted in his 2016 annual letter: “One reason we were attracted to the P/C business was its financial characteristics: (certain) P/C insurers receive premiums upfront and pay claims later, often much later.” The “often much later” is doing a lot of work in that sentence. “Much later” means more years of investment income on the float. It means more compounding. It means more present-value discount between the liability’s face value and its economic cost.

A majority of Horace Mann's insurance are short-tail insurance — which means that they only have about 12 months of float and claims are made really quickly. On the other hand, the P/C insurance that Buffet bought in his previous investments was long-tail. In complex commercial P/C lines (like asbestos, environmental hazards, or multi-party liability), decades can pass before courts or adjusters finalize a claim. This long time gap provides Berkshire with a permanent pool of compounding investment capital. The low float duration of Horace Mann's float does not.

Subscribe now

PART V: THE OWNER’S FRAMEWORK — HOW TO VALUE A FLOAT-GENERATING INSURANCE BUSINESS

What a Business Owner Actually Calculates When Buying an Insurer

A business owner acquiring an insurance company does not apply a P/E multiple to current earnings. They perform a float analysis that has five steps, none of which requires a forecast of future insurance prices or investment yields.

The first step is to measure the float: total loss reserves plus unearned premiums (prepaid insurance premium that the insurer has not yet "earned" because the policy term hasn't fully elapsed) minus agents’ balances, prepaid acquisition costs, and similar deductions. This is the investable pool — the capital the business holds but does not own, which is available for investment. Berkshire discloses this figure explicitly in its annual letter. Most insurers do not provide the float figure directly, but it can be calculated from the balance sheet components.

The second step is to calculate the historical cost of float over the past five to ten years: average annual underwriting profit or loss (via combined ratio), divided by average float. This produces a percentage — the average cost per dollar of float. Compare this to the average 10-year Treasury yield over the same period. If the cost of float has consistently been below the 10-year yield, the float is providing cheaper capital than the risk-free rate. If it has been persistently above, the float is expensive leverage. Buffett’s 1998 letter disclosed that Berkshire’s average float cost over 32 years had been well under zero — an unambiguous signal of the quality of the underlying underwriting franchise.

The third step is to value the float explicitly as a capital pool. At a zero cost of float and a 4.5% investment yield on a $30 billion float, the insurer generates $1.35 billion per year in investment income on capital it does not own. Capitalised at, say, 12 times (reflecting the perpetual nature of the float and the zero cost of capital), this component of value alone is worth $16.2 billion — from a $30 billion capital pool that shows up as a liability on the balance sheet.

The fourth step is to assess the durability of the float. Is this a business with genuine underwriting discipline — disciplined pricing, conservative reserving, no pressure to grow premiums for volume’s sake — or is the float a product of aggressive growth that will reverse in the next soft market? The General Re case illustrates the danger of float that appears large but whose cost is concealed by inadequate reserving.

The indicator for reserve adequacy is the consistency of prior-year reserve development: an insurer that consistently reports favourable prior-year development — meaning claims on old years came in below the original reserve estimate — is demonstrating conservative reserving. An insurer that consistently reports adverse development is signalling the opposite.

Subscribe now

PART VI: CONFIRMED CATALYSTS — WHAT MAKES THE FLOAT MORE VALUABLE NOW THAN BEFORE

Catalyst 1: The Rate Environment Has Permanently Repriced the Investment Returns on Float

From 2009 to 2021, the Federal Reserve maintained interest rates near zero. During this period, the investment income on insurance float was severely compressed. An insurer with $10 billion in float earning 1% on its bond portfolio generated $100 million in annual investment income. The same $10 billion at 4.5% generates $450 million. The difference — $350 million per year — is the direct consequence of the interest rate normalisation that began in 2022.

The rate cycle has confirmed that the investment income component of float value is highly sensitive to the interest rate environment — and that the decade of near-zero rates dramatically suppressed the economic value of float in a way that earnings models and P/E ratios captured partially but float analysis captures precisely. Insurers that maintained their float through the near-zero rate period — paying claims, keeping policyholders, writing new business at or near combined ratio breakeven — were positioned to capture the full benefit of rising rates in 2022 to 2024. The companies that maintained underwriting discipline during the lean years and avoided the temptation to chase underwriting profit by cutting prices and extending terms owned the same float at the same near-zero balance sheet cost — but now earn 4% to 5% rather than 1% on it. The float’s value, expressed as investment income per year, has quadrupled without any change in the float’s size.

This is a confirmed, already-realised catalyst. Berkshire’s insurance investment income line — the direct output of deploying float into investable assets — grew dramatically from 2022 to 2024. The market began to price this in as rising rates flowed through to insurer earnings, but the structural message is more durable: interest rates are unlikely to return to near-zero levels in any near-term scenario, meaning the elevated investment income on float is structural rather than cyclical.

Subscribe now

Catalyst 2: The Hard Insurance Market Has Improved Float Quality Across the Industry

Following several years of significant catastrophe losses — California wildfires, Atlantic hurricanes, mid-continent convective storms, and the COVID-related liability uncertainty — the global property-casualty insurance market entered a sustained hardening cycle from approximately 2020 to 2024. Insurance prices rose substantially across most lines, with commercial property rate increases exceeding 20% to 30% in some categories and personal lines auto insurance raising rates 30% to 40% to catch up with inflation in repair costs.

A hard insurance market does two things simultaneously that expand float value. It increases premium volume, which directly increases float as more premium is collected and held. And it improves underwriting profitability, which reduces the cost of float — if the combined ratio falls from 100% to 94%, the cost of float drops from zero to negative 6%, meaning policyholders are paying the insurer 6% to hold their money. Berkshire’s 2024 annual report disclosed 14 consecutive years of underwriting profit, with $28 billion in cumulative pre-tax underwriting gains — the direct consequence of underwriting discipline through a period of market hardening. The Berkshire 2024 operating earnings were $47.4 billion, with insurance underwriting contributing approximately $9 billion and insurance investment income contributing further.

Subscribe now

Catalyst 3: Catastrophe Mispricing Creates Episodic Windows to Acquire Float Cheaply

When a major catastrophe — a hurricane, an earthquake, a pandemic — strikes, insurance stocks typically decline immediately as the market prices in expected losses. The response is usually correct in the short term: claims will be paid, earnings will be affected. But the market frequently overcorrects, selling insurance stocks as if the catastrophe impairs not only current earnings but the long-term value of the float franchise.

The float does not disappear after a catastrophe. The loss reserves increase — reflecting the claims to be paid — but the investable float remains invested and earning returns. The underwriting franchise that generated the float continues operating, writing new business, collecting new premiums. In many cases, a major catastrophe actually strengthens the long-term value of insurance franchises by prompting widespread premium increases — hardening the market precisely because capacity has been reduced and prices must rise to attract new capital back into the industry. The insurer that survived the catastrophe, maintained its balance sheet, and remains a going concern benefits from higher prices on all subsequent business — directly improving the future cost of float.

This mechanism has repeated itself after every major catastrophe cycle in modern insurance history. The investors who bought quality insurance franchises in the aftermath of Hurricane Andrew in 1992, the 9/11 attacks in 2001, Hurricanes Katrina and Rita in 2005, and COVID in 2020 — at prices reflecting panic-driven selling of float-rich franchises — consistently earned above-market returns as the hard market cycle that followed each event improved underwriting profitability and directly reduced the cost of the surviving insurers’ float. The confirmed pattern is: catastrophe creates a temporary discount to float franchise value; the hard market that follows confirms and enhances the franchise; patient capital earns the spread.

Subscribe now

CONCLUSION: THE BALANCE SHEET LIABILITY THAT ACTS LIKE AN ASSET

Berkshire Hathaway’s 2024 annual report contains a sentence that is, in retrospect, the most important sentence in the entire document: “Our float has grown from $46 billion to $171 billion” over the past two decades. The sentence is presented as a data point. It is actually a summary of one of the greatest feats of capital accumulation in the history of finance.

In those two decades, Berkshire did not issue $125 billion in debt to grow its investment pool. It did not dilute shareholders by $125 billion in equity issuance. It took in insurance premiums, paid claims, and held the difference — the float — for investment. The $125 billion in incremental float was provided by policyholders who needed insurance, under contracts that allowed Berkshire to invest the premium float during the period between collection and claim payment. And because the underwriting was profitable — combined ratios below 100% for most of those twenty years — the policyholders paid Berkshire for the privilege of providing this capital.

The balance sheet shows $171 billion in insurance liabilities. The economic reality is $171 billion in permanent, revolving investment capital at zero or negative cost. The gap between those two descriptions is not a philosophical distinction. It is the source of Berkshire Hathaway’s status as one of the largest and most consistently compounding businesses in the history of capital markets.

The P/E ratio tells you nothing about this. The Price-to-Book ratio tells you very little. EV/EBITDA, applied to the insurance segment alone, would produce a number that reflects current underwriting results and current investment yields — useful information but not the relevant question. The relevant question is: what does the float cost, how large is it, how durable is that cost, and what does the market price imply about the value of that capital pool? When the market’s implied price for the float — backed into from the total enterprise value and the non-insurance asset value — is below the present value of the investment income the float will generate over its remaining life at current yields, the insurer’s equity is cheap by the only measure that matters for this particular type of business.

Insurance is the only business in the world where, if you are genuinely good at what you do, your customers pay you to borrow their money. The accounting calls it a liability. The economics call it the most valuable thing on the balance sheet.

Buffett figured this out in 1967, for $8.6 million. The market is still catching up.

This white paper is for educational and informational purposes only. Berkshire Hathaway financial data is drawn from publicly available Annual Reports filed on berkshirehathaway.com, including the 1998, 2000, 2001, 2016, and 2024 Annual Reports. General Re float and underwriting data is drawn from Berkshire Hathaway’s shareholder letters for the relevant years. GEICO premium and float data is drawn from Berkshire’s consolidated insurance disclosures. The float calculation methodology described is consistent with the methodology disclosed by Berkshire Hathaway in its annual reports. Nothing herein constitutes investment advice.

If you like what you read, check out my other writings 👇

PART I: WHAT A STEEL DISTRIBUTOR ACTUALLY DOES, AND WHY THE INVENTORY IS EVERYTHING

The Middleman with the Warehouse

To understand why steel distributors’ inventory is so consequential, you need to understand what these companies actually do — because it is not glamorous, and the business model is not complicated, which is precisely why the accounting distortion within it is so easy to miss.

A steel service centre — companies like Ryerson Holding Corporation, Olympic Steel, Metals USA, or the smaller pure-plays like Worthington Steel and Friedman Industries — sits between the steel mills and the industrial end-users. The mills produce steel in enormous quantities: hot-rolled coil, cold-rolled sheet, galvanised steel, structural beams, plate, and tube. They sell in large minimum order quantities, often measured in railcar or truckload lots. A manufacturer that needs 5,000 pounds of stainless steel sheet in a specific gauge and finish does not buy directly from the mill. The mill’s minimum order might be 50,000 pounds, its lead time might be eight to twelve weeks, and its products come in standard sizes that may not match the manufacturer’s specifications.

The service centre solves this problem. It buys large quantities from the mill at prevailing market prices, warehouses the material, and resells it in smaller quantities, to tighter specifications, with shorter lead times, to thousands of customers across automotive, construction, manufacturing, and energy industries. It earns a spread — a margin between what it paid the mill and what it charges the customer. This spread is not fixed. It fluctuates with steel prices, with supply-demand dynamics at the mill level, and with competitive pressure from other distributors and from direct mill sales.

The primary asset of this business is not the warehouse. It is not the cutting equipment or the delivery fleet. It is the inventory — the steel sitting on the racks and coil cradles in the warehouse, purchased at some point in the past and held until a customer needs it. The inventory is what allows the distributor to promise next-day delivery on a 500-pound order. It is what justifies the service centre’s existence as a business. Remove the inventory and the business evaporates. Every dollar of revenue the service centre will ever generate passes through the inventory first.

This makes the valuation of inventory the single most important question in assessing what a steel distribution business is worth. And this is precisely where the accounting system introduces a gap between what the balance sheet says and what the business is actually worth.

Subscribe now

PART II: LIFO — THE ACCOUNTING ELECTION THAT CHANGED EVERYTHING

How a Tax Decision Became an Information Crisis

The Last-In, First-Out (LIFO) inventory accounting method was not designed to obscure asset values. It was designed to be economically sensible from a tax perspective in inflationary environments. The logic is straightforward: if prices are rising, the most recently purchased inventory is the most expensive. Under LIFO, when you sell a unit of inventory, you match the sale revenue against the cost of the most recently purchased inventory — the highest cost. This reduces reported gross profit in an inflationary period, which reduces taxable income, which reduces taxes owed. For a steel distributor operating during the extended inflationary period of the 1960s and 1970s, this was an extremely attractive feature. Hundreds of US companies adopted LIFO and have never switched back, because switching would create an enormous tax liability from the recognition of the accumulated LIFO reserve.

The accounting consequence of LIFO, however, is one that most financial analysts do not fully internalise. In an inflationary environment, LIFO means that the inventory sitting on the balance sheet is valued at the oldest, cheapest cost layers. Recent purchases — at higher current prices — flow through to Cost of Goods Sold immediately. Old purchases — at lower historical prices — accumulate in the balance sheet as the residual “unsold” inventory. This residual is called the LIFO base stock. Over decades of inflation, the LIFO base stock can be valued at prices from the 1970s, 1980s, or 1990s, depending on when the company established its inventory position.

The difference between what the inventory is worth today at current replacement cost and what LIFO shows on the balance sheet is the LIFO Reserve — a number that every US LIFO company is required to disclose in the notes to its financial statements. It is not hidden. It is disclosed. It is just in a footnote, expressed in a way that most investors ignore or misunderstand.

Subscribe now

The Mechanics of the Gap: An Actual Story

Imagine you run a steel service centre and you bought 1,000 tonnes of hot-rolled coil steel in 2005 for $400 per tonne. You still have that steel on your books. Under LIFO, that 1,000 tonnes appears on your balance sheet at $400/tonne — $400,000 in inventory value. But it is now 2022, and hot-rolled coil steel is trading at $1,800 per tonne (which it actually did in mid-2021, briefly reaching that level). The replacement cost of your 1,000 tonnes is $1.8 million. The balance sheet says $400,000. The LIFO reserve for that tonnage is $1.4 million — the gap between what the inventory is worth and what the balance sheet says it is worth.

A business owner buying this distribution company asks: what is the inventory worth? The answer is $1.8 million. The company can sell those 1,000 tonnes tomorrow at $1,800/tonne. That is what they are worth. The GAAP balance sheet says $400,000. The footnote — buried in Note 4 or Note 6 of the financial statements — discloses the LIFO reserve and allows you to reconstruct the true inventory value. But you have to look for it.

This is the gap. It is systematic. It is structural. It is not an error or an anomaly. It is a predictable consequence of an accounting election that hundreds of US companies made decades ago and have maintained ever since, because the alternative — switching to FIFO and recognising the accumulated LIFO reserve as income — would trigger a tax bill that could exceed the company’s annual earnings for several years running.

Subscribe now

PART III: THE NUMBERS, FROM ACTUAL FILINGS

Ryerson: The $245 Million Footnote in 2022

Ryerson Holding Corporation (NYSE: RYI) is one of the largest steel and metals service centres in North America, with approximately $5.1 billion in revenue in 2023. In its December 31, 2022 10-K filing, Ryerson disclosed the following in Note 4 (Inventories): “If current cost had been used to value inventories, such inventories would have been $245 million higher and $303 million higher than reported at December 31, 2022 and 2021, respectively.”

Read that again carefully. The inventory sitting in Ryerson’s warehouses on December 31, 2022 — steel that would be sold to customers in the following weeks and months — was worth $245 million more at current replacement cost than what appeared on the balance sheet. Not in 20 years. Not after a series of favourable business developments. Today, at prices already established in the market. The inventory had already been purchased. It was sitting in the warehouse. A buyer could walk in and offer to buy it at current market prices and receive $245 million more than the balance sheet suggested it was worth.

In 2021, the gap was even larger at $303 million. Ryerson’s total inventory on the balance sheet at year-end 2021 was $832.1 million. Adding back $303 million means the inventory was actually worth approximately $1.135 billion — 36% more than the balance sheet showed.

Ryerson’s market capitalisation during much of 2022 ranged from approximately $600 million to $1.1 billion. At the lower end of that range, an investor buying all of Ryerson’s equity was paying $600 million for a business whose inventory alone — the primary working asset — was worth over $1 billion at replacement cost. The equity below the inventory. This is not a forecast. This is what the 10-K disclosed.

During the same period, Ryerson’s reported gross margin was being crushed by LIFO expense. In Q4 2021, Ryerson reported LIFO expense of $76 million — meaning $76 million of cost that in economic reality did not exist (no cash left the company, no inventory was lost) flowed through the income statement and reduced reported earnings. Analysts looking at the earnings model saw compressed margins and reduced their price targets. The inventory footnote told a completely different story: the inventory was appreciating, not depreciating. The earnings model was penalising the company for owning more valuable inventory than the LIFO method acknowledged.

Subscribe now

The Wild Swing Back: Q4 2023

The opposite phenomenon appeared in Q4 2023. Ryerson disclosed LIFO income of $59.3 million in that quarter. Steel prices had declined from their 2022 peaks, meaning the cost of recently purchased inventory had fallen below the cost of older inventory layers on the books. Under LIFO mechanics, when prices fall, the recent cheap purchases flow through to Cost of Goods Sold, while the older expensive layers remain on the balance sheet. This generates LIFO income — a non-cash accounting gain that inflates reported earnings without any economic benefit.

Analysts who anchored to the earnings number saw improving results. The earnings model showed strong gross margins in Q4 2023. The inventory footnote told the counterpoint: the LIFO reserve had shrunk because inventory replacement cost had fallen, meaning the inventory was worth less in absolute terms than it had been in 2022. The earnings model was flattering the company at exactly the moment the underlying asset was weakening.

This is the pattern that repeats across every steel price cycle and every steel distributor on the LIFO system: earnings are systematically overstated when prices are falling (LIFO income artificially inflates margins) and systematically understated when prices are rising (LIFO expense artificially compresses margins). The earnings model is exactly backwards as a predictor of inventory value — it tells you the inventory is most valuable when it is falling in price, and least valuable when it is rising.

Subscribe now

Nucor: The $1.1 Billion You Cannot See on the Balance Sheet

Nucor Corporation (NYSE: NUE) is the largest steel manufacturer in the United States and also operates steel distribution and fabrication businesses through its various subsidiaries. In its 2022 annual report, Nucor disclosed a LIFO reserve of approximately $1.1 billion. This single footnote item means that Nucor’s inventory — as disclosed on the face of the balance sheet — was understated by $1.1 billion relative to current replacement cost.

Tax-effect this at the current US corporate rate of 21%: the after-tax impact on net asset value is approximately $869 million. This is not an estimate. It is the disclosed difference between LIFO-stated inventory and FIFO replacement cost, adjusted for the tax liability that would be triggered if the LIFO reserve were recognised as income. An analyst computing Nucor’s Price-to-Book ratio using reported book value is comparing market price to a book value that is approximately $869 million lower than the economically correct figure.

The perversity of this becomes clear when you run through the cycle. In 2020, when steel prices were depressed by COVID-related demand destruction, Nucor’s LIFO reserve was much smaller — perhaps a few hundred million dollars. Reported book value looked relatively high relative to economic book value. The P/B ratio looked lower, suggesting the stock was relatively cheap by that metric. But the inventory was less valuable in economic terms. In 2021 and 2022, when steel prices surged to multi-year highs, the LIFO reserve expanded dramatically — meaning the inventory became dramatically more valuable. But reported book value grew slowly because the LIFO reserve was not recognised. The P/B ratio looked higher, suggesting the stock was more expensive. But the inventory was more valuable in economic terms. The metric pointed precisely the wrong direction, at precisely the wrong time, every time.

Subscribe now

PART IV: THE MARKET’S TOOLS AND WHY THEY FAIL HERE

What Happens When the Analyst Runs the Earnings Model

The standard sell-side approach to a steel distributor is built around revenue per tonne, gross margin per tonne, operating expense per tonne, and an earnings multiple. The analyst models the steel price, estimates volumes, calculates spread, subtracts operating costs, arrives at EBITDA, applies a multiple, and produces a price target. This model has three structural problems when applied to a LIFO company in a volatile steel price environment.

The first problem is that LIFO expense and income make reported gross margin a particularly unreliable indicator of underlying business performance. When Ryerson reported $76 million of LIFO expense in a single quarter, its gross margin looked terrible by historical standards. But the gross margin, excluding LIFO, was actually quite healthy — Ryerson disclosed its EBITDA excluding LIFO explicitly because management understood that the LIFO-adjusted earnings figure was the economically meaningful one. Analysts who used reported gross margin were measuring an accounting artifact, not a business outcome. The inventory was getting more valuable. The spread between purchase price and sale price — the actual economics of the business — was widening. The earnings model said the opposite.

The second problem is that the P/B ratio, which investors use as a valuation shortcut for asset-heavy businesses, is calculated on a LIFO-stated book value that systematically understates the value of the primary asset. When steel prices are high and inventory is most valuable, the LIFO reserve is largest, book value is most understated, and P/B looks most expensive. When steel prices are low and inventory is least valuable, the LIFO reserve is smallest, book value is least understated, and P/B looks cheapest. The ratio signals the market to buy when the inventory is cheapest and sell when the inventory is most valuable — exactly the reverse of what an asset-value-oriented investor should do.

The third problem is that none of the standard metrics — P/E, EV/EBITDA, P/B — answer the owner’s question. The owner’s question is not “what is this company earning this quarter?” It is: “what is the inventory worth today at replacement cost, and what am I paying for it?” That question has a concrete, calculable, non-forecasted answer. It requires reading one footnote and doing one subtraction.

Subscribe now

PART V: STORIES FROM THE WAREHOUSE — CASE STUDIES IN STEEL DISTRIBUTION VALUE ARBITRAGE

The Reckoning of 2021: When Steel Went Vertical and the Earnings Model Got Lost

The story of the US steel market in 2021 is one of the most dramatic commodity price episodes in recent American industrial history, and it produced the clearest possible illustration of why LIFO accounting hides value precisely when the value is greatest.

In January 2021, US hot-rolled coil steel was priced at approximately $700 to $800 per short ton. By August 2021, the price had reached approximately $1,900 per short ton — a roughly 150% increase in seven months. The causes were multiple and simultaneous: COVID-era fiscal stimulus created massive demand for manufactured goods that consume steel; automotive production disruptions from semiconductor shortages created unusual patterns of mill output; and the large integrated steel mills, having taken production offline in the COVID lockdowns of 2020, were slow to restart capacity. Supply was tight, demand surged, and prices moved in a way the market had not seen in modern memory.

For steel service centres, this environment was economically extraordinary. A company that had purchased hot-rolled coil at $700/tonne in Q4 2020 and was selling it at $1,800/tonne in Q3 2021 was earning margins that would be almost unrecognisable in a normal market. The spread between purchase price and sale price — the distributor’s gross profit — had expanded enormously.

But the earnings model, applied naively, told investors an alarming story. Ryerson’s full-year 2021 LIFO expense was enormous — the company was buying new inventory at $1,800/tonne and flowing that high cost through to Cost of Goods Sold immediately under LIFO, while the older, cheaper inventory layers remained on the balance sheet. Reported gross margins looked compressed relative to what they would have been under FIFO accounting. Ryerson disclosed that on a FIFO basis, diluted earnings per share would have been dramatically higher than the LIFO-reported figure.

Ryerson also disclosed its LIFO reserve grew from $303 million at year-end 2021, compared to a much smaller figure in 2020 — meaning the inventory the company was holding had appreciated by hundreds of millions of dollars in a single year. That appreciation was entirely invisible to an investor looking at the income statement. The inventory line on the balance sheet grew in dollar terms because volumes stayed roughly constant while prices rose, but the LIFO reserve — buried in the footnote — captured the true magnitude of the inventory appreciation.

Investors who read the footnote, added back the LIFO reserve to book value, and compared the adjusted inventory value to Ryerson’s market capitalisation found a company whose primary working asset was worth substantially more than the market was pricing the whole enterprise. They bought. The company’s shares, which had started 2021 at approximately $14 per share, finished the year at approximately $22 per share — not quite capturing the full inventory appreciation, because LIFO accounting continued to suppress reported earnings — but meaningfully up nonetheless. The footnote readers were rewarded. The earnings-model readers were confused.

The lesson: in an inflationary steel price environment, LIFO accounting creates a systematic optical illusion. Reported earnings look worse than economic reality because LIFO expense flows through the income statement. Inventory on the balance sheet looks flat or only modestly higher because the appreciation is captured in the LIFO reserve, not in the stated inventory value. And the LIFO reserve itself grows — in the footnote — representing hundreds of millions of dollars in real, tangible, immediately realisable value that the standard metrics completely miss.

Subscribe now

Friedman Industries: The Tiny Company with the Transparent Footnote

Friedman Industries, Inc. (NYSE American: FRD) is a steel processor and service centre headquartered in Longview, Texas, with annual revenues of roughly $500 to $700 million — a fraction the size of Ryerson, and largely ignored by institutional investors and sell-side analysts. This makes it a particularly clean case study, because there is very little Wall Street noise to obscure the underlying inventory arithmetic.

Friedman operates hot-rolled coil processing facilities at several locations in the US South, purchasing coils directly from the mills, slitting and cutting them to customer specifications, and selling flat-rolled steel to manufacturers primarily in the pipe and tube, agricultural equipment, and industrial machinery sectors. The company has used LIFO inventory accounting consistently across its history.

In its fiscal year 2022 annual report (Friedman’s fiscal year ends March 31), the company disclosed steel inventories valued at approximately $78 million under LIFO. The LIFO reserve at fiscal year end 2022 was disclosed in the notes as approximately $28 to $30 million — meaning the inventory was worth approximately $106 to $108 million at current replacement cost, roughly 37% more than the balance sheet showed.

Friedman’s market capitalisation during fiscal 2022 ranged from approximately $100 to $160 million. At the lower end of that range, an investor was paying approximately $100 million for a business whose inventory alone was worth $106 to $108 million at replacement cost, plus processing equipment, accounts receivable, and the operational franchise of a 50-year-old customer relationships with regional manufacturers. The enterprise value implied by the market was, at times, essentially equivalent to the replacement cost of the inventory stack alone — the processing equipment, the real estate, the customer relationships, and the ongoing business operations were being offered for free.

The earnings model was not helpful here. Friedman’s reported earnings in fiscal 2022 were compressed by LIFO expense — the steel price run-up in 2021 had caused large LIFO charges that flowed through the income statement. P/E looked high. P/B looked slightly elevated because book value was understated by the LIFO reserve. EV/EBITDA looked middling. None of the standard metrics communicated what the footnote communicated: this is a business where the inventory alone is worth close to the entire market capitalisation.

What happened next validated the owner’s analysis. As steel prices peaked and inventory turns continued, Friedman generated substantial cash earnings from liquidating high-value inventory that had been purchased at lower prices. The company distributed the profits through a combination of dividends and share buybacks that, over a 24-month period, returned a significant fraction of its market capitalisation directly to shareholders. The shareholders who had read the footnote and understood that the inventory was worth more than the market price had effectively been paid to wait.

Subscribe now

Olympic Steel and the $20 Million Reserve That Told a Different Story

Olympic Steel, Inc. (NASDAQ: ZEUS) is a Cleveland-based flat-rolled steel distributor with approximately $2 billion in annual revenues, serving customers primarily in the manufacturing, automotive, and construction industries. It is mid-sized by industry standards, covered by a handful of sell-side analysts, and largely ignored by generalist investors.

In its December 31, 2022 10-K filing, Olympic disclosed a LIFO reserve of $20.3 million — meaning its LIFO-stated inventory of $416.9 million would have been $437.2 million under FIFO at current replacement cost. Compared to Ryerson or Nucor, the absolute dollar amount of the reserve is modest. But the proportion matters more than the absolute number.

What is more instructive about Olympic Steel’s filing is what happens when you compare the LIFO reserve across years. At December 31, 2020, Olympic had a LIFO debit of $2.1 million — meaning inventories under LIFO were actually slightly higher than FIFO cost, because prices had recently fallen and recent (lower-cost) purchases had been absorbed into the LIFO cost stack while older (higher-cost) layers lingered. By December 31, 2021, the reserve had swung to a positive $19.7 million. By December 31, 2022, it stood at $20.3 million.

This swing from a $2.1 million debit to a $20.3 million reserve — a $22.4 million change in the LIFO reserve — occurred in exactly two years, across a period when US hot-rolled coil prices went from approximately $500/tonne to over $1,800/tonne and back to approximately $800/tonne. The LIFO reserve was measuring, in real time, the gap between what the inventory cost when purchased and what it is worth today. When the reserve is growing, the inventory is accumulating value that the balance sheet does not show. When the reserve is shrinking, the inventory value is reverting toward cost.

An investor who read the LIFO reserve disclosure across two consecutive years — not even across a full cycle — could determine the direction and magnitude of inventory value movement without running a single earnings model. The footnote was doing the work that the income statement obscures.

Olympic Steel’s share price over this period moved from approximately $20 in early 2020 to a peak of approximately $55 in early 2022, before retreating toward $35 as steel prices normalised. The investors who tracked the LIFO reserve expansion rather than the reported earnings had a better framework for understanding when the company was accumulating hidden value and when that value was being realised.

Subscribe now

PART VI: WORTHINGTON STEEL — WHERE THE LIFO RESERVE REVEALS A VALUATION STORY

Worthington Steel, Inc. (NYSE: WS) was spun off from Worthington Enterprises in December 2023, separating the steel processing business from the diversified consumer products and building products businesses that had historically been combined under the Worthington Industries umbrella. The spin-off created a pure-play steel processor and service centre for the first time — a company focused entirely on value-added steel processing for automotive, construction, and industrial customers.

The creation of a standalone Worthington Steel was an event that itself illustrated the owner’s vs. market valuation gap in a direct way. The combined Worthington Industries entity had been valued by the market primarily on the earnings multiple of its diversified businesses, which included pressurised cylinders, cooling products, and building systems — businesses with different margin profiles and valuation frameworks than steel processing. The steel processing segment, with its LIFO-distorted inventory accounting and cyclical earnings, was being valued as an appendage of a diversified industrial company rather than as a standalone asset base.

The spin-off forced the market to value the steel assets separately. The initial market capitalisation of Worthington Steel post-spin was approximately $750 million to $900 million. Its disclosed LIFO reserve at the time of separation — inherited from the Worthington Industries steel processing segment — represented meaningful hidden inventory value that would now be visible to investors evaluating a pure-play steel service centre rather than a conglomerate.

As of today (2026), Worthington Steel is worth almost $2 billion in market cap.

Subscribe now

PART VII: IRON ORE — THE UPSTREAM PARTNER IN THE SAME STORY

Where the Steel Comes From

Iron ore is the raw material from which steel is made. Every tonne of flat-rolled coil that fills a steel service centre’s warehouse began as iron ore — typically magnetite or hematite rock mined in Australia, Brazil, West Africa, or the US, shipped to a steel mill, reduced to molten pig iron in a blast furnace, refined to steel in a basic oxygen furnace or electric arc furnace, rolled into sheet or plate, and then distributed through the service centre network.

The economics of iron ore mining have their own version of the same accounting gap we have seen in steel distribution — and the same disconnect between what the balance sheet records and what a business owner would pay for the asset today.

The Ore Body Valuation Gap

Iron ore miners carry their ore bodies on the balance sheet at historical development cost, depleted against production using the Units of Production method. This accounting treatment effectively treats a world-class geological discovery as a wasting expense rather than a massive, appreciating inventory. The ore body itself is never recognized at fair value; what appears on the balance sheet is merely the “sunk cost” of the steel and concrete used to access it.

Subscribe now

Rio Tinto: The Multi-Billion Tonne Disconnect

Rio Tinto’s Pilbara operations are not just a production machine pumping out 330 million tonnes a year; they are a massive geological warehouse. As of recent disclosures, Rio Tinto maintains Proved and Probable (P&P) Reserves in the Pilbara totaling approximately 1.5 to 1.7 billion tonnes, with an additional Resource base that extends significantly further.

The Smaller Producers Where the Gap Is More Acute

The gap between “book value” and “reserve value” becomes an abyss for smaller, high-grade producers. Champion Iron (ASX: CIA), operating the Bloom Lake mine in Quebec, offers a concentrated example of this mispricing.

Bloom Lake isn’t just a 15–20 million tonne-per-annum (Mtpa) producer; it is a Tier-1 asset with a reserve life that spans decades.

• The Reserve Base: Champion Iron reports Proved and Probable Reserves of approximately 700 million tonnes at Bloom Lake.

• The Grade Premium: Because it produces a 66.2% Fe concentrate, the “embedded margin” per tonne is higher than the Pilbara benchmark, estimated at ~$72 per tonne (at $120/t benchmark prices).

The “Shadow” Balance Sheet

If we multiply the 716 million tonnes of reserves by the $72/tonne margin, we find an embedded value of roughly $51 billion.

Yet, Champion Iron’s market capitalization typically sits between $2 billion and $3 billion. Even after accounting for future capital expenditures, taxes, and the time value of money (discounting), the market is essentially valuing the company at less than 10% of the net margin value of the ore it has already proved to exist.

The “Units of Production” depletion method on the balance sheet treats each tonne mined as a loss of value. In reality, the “proven” nature of these decades-long reserves provides a margin of safety and a duration of cash flow that traditional earnings-per-share (EPS) models, which often look only 12–24 months ahead, fail to capture.

Subscribe now

PART VIII: CONFIRMED CATALYSTS — WHAT IS ALREADY IN MOTION

Catalyst 1: US Section 232 Steel Tariffs — Confirmed and Structurally Active

The Trump administration’s Section 232 tariffs on steel imports, first imposed in March 2018, have been maintained, expanded, and reinforced through multiple subsequent administrations. The reimposition and strengthening of Section 232 tariffs in early 2025 — under the new Trump administration — added an additional layer of import protection on top of the existing baseline. As of 2025, most foreign steel entering the United States faces a 25% ad valorem tariff, with country-specific quota arrangements for a small number of allied nations.

The effect of sustained Section 232 tariffs is to create a domestic steel price floor. When global steel prices are below US prices plus the tariff, imports are non-competitive and domestic mills can maintain higher prices than the global market would otherwise support. For steel service centres holding domestic inventory — purchased at domestic mill prices that already reflect the tariff floor — this means their inventory value is supported by a government-maintained price floor that does not exist in most commodity markets.

This is not a cyclical factor. It is a policy factor that has now persisted for seven years across two administrations with opposing economic philosophies, suggesting bipartisan durability rooted in domestic political economy rather than abstract trade theory. Service centres whose LIFO reserves were built at pre-tariff price levels hold inventory whose replacement cost is permanently elevated relative to global commodity prices by the tariff differential. The LIFO reserve measures this gap precisely, and the gap is sustained by enacted, currently-operative law.

Subscribe now

Catalyst 2: Infrastructure Spending and the Steel Demand Baseline

The US Infrastructure Investment and Jobs Act (IIJA) of 2021 authorised approximately $1.2 trillion in federal spending over five years, including roughly $550 billion in new spending on transportation, broadband, water infrastructure, and the power grid. The specific steel content of this infrastructure investment is substantial: bridges, structural beams for elevated highways, rebar for concrete construction, pipe for water systems, and structural steel for transmission towers all require domestic steel supply under the Buy America provisions of the act.

Infrastructure spending works through steel distributors for a significant fraction of its requirements. Large construction contractors and fabricators purchase structural steel, plate, and other products from service centres for customisation and just-in-time delivery to construction sites. The IIJA’s multi-year spending profile creates a sustained demand baseline that reduces the volatility in steel service centre revenues relative to purely cyclical demand patterns. For companies like Olympic Steel, which focuses on flat-rolled products including those used in manufacturing and construction equipment, the infrastructure programme creates a multi-year demand tailwind that makes the inventory — and the LIFO reserve attached to it — more defensible as an asset.

Subscribe now

Catalyst 3: Automotive Electrification and the Hidden Steel Story

The transition from internal combustion engine vehicles to battery electric vehicles involves an often-overlooked fact about steel consumption: EVs require substantially more electrical steel — specifically non-grain-oriented (NGO) electrical steel — for their motors than ICE vehicles require for any equivalent application. The motor stators and rotors in an EV motor use laminated electrical steel to manage magnetic flux. A single EV traction motor might require 30 to 60 kilograms of electrical steel sheet. Large EV motors — in performance vehicles, SUVs, and commercial trucks — require more.

Steel service centres that have invested in processing capabilities for electrical steel — precision slitting to tight width tolerances, surface protection to prevent corrosion, flatness control to meet motor lamination requirements — are positioned to benefit from the EV transition in a way that the standard automotive steel distributor is not. Worthington Steel has explicitly identified electrical steel processing as a growth focus, having invested in capabilities to serve motor lamination customers. The installed processing capability — the slitters, tension levellers, and quality control systems capable of handling electrical steel — represents an asset that is valued on the balance sheet at historical capital expenditure cost, not at the forward value of the customer contracts and market position it enables.

Subscribe now

Catalyst 4: The LIFO Reserve as a Buyback Mechanism — The Self-Liquidating Value

There is a final catalyst for steel distribution value arbitrage that is less obvious than policy or demand, but more direct in its effect on shareholder returns: the LIFO reserve itself creates the conditions for highly accretive share buybacks.

Here is the mechanism. When steel prices are elevated and the LIFO reserve is large, service centres generate substantial cash earnings from their operations — the spread between purchase price and sale price is wide, inventory is turning at elevated prices, and cash is accumulating. Management can choose to return this cash to shareholders through buybacks. The stock price, artificially depressed by the earnings model’s failure to credit the LIFO reserve, means the buyback is purchasing shares at a discount to their economically adjusted book value. The company is essentially converting excess inventory value — recognised in the footnote but not in the market price — into permanent per-share value through repurchasing shares at below-economic-book-value prices.

Ryerson specifically highlighted this dynamic during its capital allocation discussions in 2021 and 2022, when the company was generating substantial free cash flow from its inflated-price inventory liquidation and using a portion of that cash to repurchase shares. The share count declined during a period when the share price did not fully reflect the inventory value. Each repurchased share at below-economic-book-value permanently increases the per-share claim of remaining shareholders on the inventory value — whether or not the market ever correctly prices it.

The catalyst here is confirmed in a simple way: these companies have historically been consistent repurchasers of their own shares during periods of elevated steel prices, and the buybacks have consistently been executed at prices that, on an LIFO-adjusted basis, were below economic book value. The pattern is documented in proxy statements, earnings releases, and annual reports. It does not require a forecast to expect to continue — it follows mechanically from the financial incentive structure.

Subscribe now

CONCLUSION: THE FOOTNOTE AS THE MOST HONEST NUMBER IN THE DOCUMENT

There is something almost absurd about the structure of the value arbitrage described in this paper. The information that creates the opportunity is not hidden. It is disclosed. It is required to be disclosed by US GAAP, in every annual report filed by every LIFO-method company, in the notes to the financial statements. The LIFO reserve tells you, in plain numbers, exactly how much the inventory is understated relative to current replacement cost. It is the most honest number in the document — it is the number that says, “here is how much more the inventory is actually worth than what the balance sheet shows.”

And yet the market persistently underweights it, because the earnings model does not use it, because the P/E ratio ignores it, because the EV/EBITDA calculation is built on reported earnings rather than LIFO-adjusted earnings, and because institutional investors are trained to look at the income statement first, the balance sheet second, and the notes to the financial statements almost never.

The business owner who is considering buying a steel service centre does not start with the income statement. They walk the warehouse, count the tonnes, check today’s steel price on the spot market, calculate the replacement cost of what is on the racks, and compare that to the acquisition price. The LIFO reserve footnote is that owner’s calculation, disclosed in regulatory-mandated format, sitting in plain view on EDGAR for any investor who cares to look.

The pattern is consistent across the companies in this sector. Ryerson’s inventory was worth $245 million more than the balance sheet showed at year-end 2022. Nucor’s was understated by $1.1 billion. Friedman’s by approximately 37% of its stated inventory value. Olympic Steel’s reserve grew from negative $2 million to positive $20 million in two years, tracking the steel price cycle in a way that the reported gross margin did not. Every one of these numbers was in a footnote. Every one of them told a more accurate story about asset value than the earnings model did.

The iron ore producers tell the same story from a different direction: ore bodies carried at depleted historical cost, generating annual operating income many multiples of their balance sheet value, priced by the market on earnings multiples that are most compressed precisely when the ore price is highest and the balance sheet value is furthest below replacement cost.

In both cases — the steel distributor with its LIFO reserve and the iron ore producer with its depleted mineral property — the accounting is doing what accounting was designed to do: recording historical cost conservatively. The market is doing what it usually does: anchoring on the most visible, most frequently reported metrics. The business owner is doing something simpler and more direct: asking what the asset is worth today, at today’s prices and today’s costs, and comparing that to what the market is charging for it.

The gap between those two numbers is where the returns have consistently been made.

This white paper is for educational and informational purposes only. All financial figures cited are drawn from publicly available company filings including 10-Ks, 20-Fs, 40-Fs, and earnings releases as filed with the SEC or equivalent regulatory bodies. Specific footnote disclosures cited include Note 4 (Inventories) of Ryerson Holding Corporation’s December 31, 2022 10-K and the corresponding disclosures of Olympic Steel and other companies named. Nothing herein constitutes investment advice.

If you like this read, check out my other writings 👇

Every other white paper in this series addresses a situation where the primary income-generating asset is invisible to accounting — the uranium reserve that never appears on a balance sheet, the core deposit intangible carried at zero, the landfill airspace that sits at depleted historical cost while tipping fees rise. In each of those cases, the investor’s challenge is to find a valuation methodology that the accountant cannot provide, because the accounting standard simply refuses to recognise the asset at its economic value.

Maritime shipping is categorically different, and the difference is what makes it simultaneously the most straightforward and the most instructive case study in the series. In shipping, the primary asset — the vessel — has a live, weekly, arm’s-length market price published every Monday by the Baltic Exchange. Third-party brokers including Clarkson Research Services, VesselsValue.com, and Xclusiv Shipbrokers maintain continuously updated valuations for every named vessel in the global fleet, based on documented comparable sales transactions. The asset does not need a financial model. It does not need a geological report or a regulatory filing. It has a current market price, quoted in dollars per vessel, that any investor can look up in approximately thirty seconds.

And yet the market consistently misprices shipping equities — sometimes dramatically, sometimes for years at a time — because Wall Street’s default tools for valuing businesses are earnings multiples and book values, and neither of those tools is calibrated to the asset that matters in this industry. The vessel has a spot price. The P/E ratio ignores it. The Price-to-Book ratio misrepresents it. And the gap between what the secondhand market says a fleet is worth and what the equity market implies about the same fleet is where the returns have been made, repeatedly, by investors who understood what they were actually buying.

Subscribe now

PART I: THE ASSET, PRECISELY DEFINED

Why the Ship Is Everything and Everything Else Is Nothing

A dry bulk shipping company — one that operates Capesize, Panamax, or Supramax vessels carrying iron ore, coal, grain, and fertiliser from producing nations to consuming ones — is the simplest conceivable business from a capital structure perspective. You own ships. You charter those ships to industrial customers who need to move large quantities of bulk commodities across oceans. The customer pays you a daily rate for the use of the vessel. At the end of the charter, the ship comes back to you and you charter it to someone else.

Every dollar of revenue this company generates flows directly from the physical vessel. There is no software. There is no brand. There is no customer relationship that survives the vessel being sold — if you sold the ship, the charterer would simply charter an equivalent vessel from another owner. The intellectual property is zero. The competitive moat, to the extent it exists, comes from the age, quality, and fuel efficiency of the fleet — all characteristics of the physical vessel — and the experience of the operating team in maintaining and chartering those vessels. But if you stripped out the vessels and left everything else, there would be nothing left that generates income.

This makes shipping unusually tractable for the owner’s valuation framework. There is no need to assess the value of a customer relationship, or a drug pipeline, or a regulatory franchise. You count the ships, look up their current market values in the secondhand broker database, subtract the debt, and you have what the business is worth to a liquidating buyer today.

The size of the gap between that number and the equity market price — and the reason that gap exists and periodically becomes enormous — is the subject of this paper.

Subscribe now

The Ship Market That Wall Street Ignores

The secondhand ship market is one of the world’s oldest and most liquid markets for major capital goods. Ships have been bought and sold between owners for as long as there have been ships. The modern institutional version of this market — with professional shipbrokers intermediating transactions, the Baltic Exchange providing weekly reference valuations, and broker databases tracking every sale by IMO vessel number — is a mature, liquid, and transparent market that processes hundreds of transactions per year.

The Baltic Exchange Sale and Purchase Assessments, published every Monday, provide the definitive reference price for each major vessel class. The Monday assessment for a five-year-old Capesize bulk carrier (approximately 172,000 DWT) reflects transactions that brokers have observed, participated in, and recorded in the previous week. It is adjusted for the specific vessel’s attributes — fuel efficiency, installed equipment, survey status, flag, and class society certification — to arrive at individual valuations for specific named vessels. VesselsValue.com, Clarkson Research Services, and Xclusiv Shipbrokers each maintain their own methodologies, but all are rooted in the same empirical foundation: what have comparable vessels actually sold for recently?

In the first nine months of 2024 alone, approximately 594 bulk carriers changed hands in the secondhand market — an average of 66 transactions per month, per Lloyd’s List data citing Xclusiv Shipbrokers. This is not a thin, illiquid market where one or two transactions set a reference price. It is a market processing more transactions in a single month than most commercial real estate markets process in a year. The price it produces is as real and as current as the copper spot price.

The specific transaction record confirms this. In late 2025, a 2019-built Newcastlemax bulk carrier — the Mineral Shougang International, 206,400 DWT, scrubber-fitted — was reported sold for $63.5 million, per Lloyd’s List citing shipbroker reports. In 2023, Clarkson Research assessed a five-year-old Capesize at approximately $54.5 million, close to 90% of the then-prevailing newbuild price of $65 million. A ten-year-old Capesize traded at approximately $19 million at the COVID trough of early 2020, then surged to $33 million by 2022 as the freight market tightened. The same physical vessel — the same iron, steel, engine, and propeller shaft — was worth 74% more in 2022 than in 2020, not because anything changed about the ship, but because the supply-demand dynamics of the freight market had shifted and the cash flows the vessel would generate during the remainder of its operating life were now substantially higher.

Subscribe now

PART II: THE ACCOUNTING TREATMENT AND THE PECULIAR WAY IT FAILS IN SHIPPING

Straight-Line Depreciation and a Market That Never Moves in a Straight Line

Under US GAAP (ASC 360) and IFRS (IAS 16 cost model, which virtually all listed shipping companies elect), vessels are capitalised at historical acquisition cost and depreciated straight-line over their estimated useful life — typically 20 to 25 years — to a residual scrap value. The mechanics are simple and consistent. A Capesize acquired in 2010 for $50 million, depreciated over 25 years to a $5 million scrap residual, produces an annual depreciation charge of $1.8 million. After 12 years of operation through 2022, the carrying value is $28.4 million. That number is completely unrelated to whether the secondhand market values a comparable 12-year-old Capesize at $8 million, as it did in 2016 during the dry bulk trough, or at $28 to $35 million, as it did in 2022 when the freight market recovered. The accounting balance sheet produces a straight diagonal line declining from $50 million to $5 million over 25 years. The secondhand market produces a jagged, volatile curve that bears no relationship to the accounting line’s slope, direction, or endpoint at any given moment.

This creates a specific and asymmetric problem that the GAAP framework handles in the worst possible way for investors trying to assess fleet value. The impairment rules under ASC 360 require companies to test vessels for impairment only when circumstances indicate that the carrying amount may not be recoverable — and the recovery test is done on an undiscounted cash flow basis, not on a market value basis. This means that during a severe market downturn, a vessel with a carrying value of $30 million but a secondhand market price of $12 million does not require an impairment if the undiscounted projected future cash flows from the vessel over its remaining life still technically exceed $30 million. The accounting does not have to acknowledge the market’s assessment. The balance sheet continues to show $30 million for something the market says is worth $12 million.

But there is no equivalent mechanism when the market recovers. If that same vessel’s secondhand price rises from $12 million back to $28 million as the freight cycle turns — as happened dramatically from 2020 to 2022 — there is no provision under GAAP or under the elected IFRS cost model for writing the vessel back up to reflect the recovery. The asset stays at its straight-line depreciated cost. The market says it is worth $28 million. The balance sheet says it is worth something between the original $30 million and the $5 million scrap residual, declining mechanically toward the latter. The recovery is invisible.

The consequence for investors is that the Price-to-Book ratio in shipping is not a stable measure of cheapness or expensiveness. It is, instead, a lagging indicator of where the secondhand market was relative to the historical cost of the fleet at the time it was acquired. When shipping stocks look expensive on a P/B basis, it often means vessels have been appreciating in the secondhand market while the GAAP book value has been declining through depreciation — the P/B multiple expands because the market price has risen while the book value has fallen. When shipping stocks look cheap on a P/B basis, it often means vessels were bought at peak prices and have since depreciated in the secondhand market faster than the accounting depreciation — the P/B looks low but it is a trap, because the book value overstates the asset’s actual market price.

This last point requires emphasis because it runs counter to every value-investing instinct applied to this sector by investors who do not understand the secondhand market. During the 2016 dry bulk trough, many shipping stocks were trading at or below 1.0x book value. Screens for “cheap” shipping equities produced long lists of companies with P/B ratios below 0.8x. This looked like value. In many cases, it was a trap. The book value was a phantom figure reflecting peak-cycle acquisition prices that had been insufficiently depreciated toward the trough-cycle market prices. The real “book value” — what the fleet was actually worth in the secondhand market — was dramatically below the GAAP figure, and the stock was not cheap; it was expensive relative to the true liquidation value of the asset.

Value Investor’s Edge, using VesselsValue.com data in Q3 2016, documented this with precision for Diana Shipping (NYSE: DSX). Diana’s 20-F stated a fleet book value of approximately $1.42 billion including newbuild advances. The independent broker valuation of the same fleet using actual comparable sales data from VesselsValue produced an estimate of approximately $675 million — a 54% overstatement of fleet value on the balance sheet relative to market clearing prices. Diana was trading at a discount to book. It was trading at a premium to the real fleet value. The investors who used P/B as their screening tool walked into the wrong side of the trade.

Subscribe now

PART III: THE OWNER’S VALUATION — A FRAMEWORK THAT ACTUALLY WORKS

Fleet NAV: The Five-Step Method That Requires No Model

The owner’s valuation of a shipping company is not merely simpler than conventional financial analysis — it is more accurate, because it starts from a directly observable market price rather than from an accounting artifact or an earnings forecast. The methodology has five steps, each of which requires only arithmetic applied to publicly available data.

The first step is to identify every vessel in the fleet by IMO number, vessel type, deadweight tonnage, year of build, and equipment specification — scrubbers, eco-engines, fuel efficiency rating. This information is disclosed in the company’s annual reports and filings, and is also available from publicly accessible databases including Clarksons’ World Fleet Register and IHS Markit’s Maritime intelligence. For a company like Diana Shipping or Safe Bulkers, the fleet list is disclosed explicitly in the 20-F or annual report, often vessel by vessel.

The second step is to look up the current secondhand market value for each vessel type from the Baltic Exchange BSPA weekly assessments or other valuation/ brokerage sites, adjusted for each vessel’s specific age using broker age-adjustment curves. The Baltic publishes benchmark values for five-year-old vessels in each major class. The age adjustment — which reflects the decline in value per year of age, calibrated to the current market relationship between new and old vessel prices — converts the benchmark price to an estimated value for each specific vessel in the fleet.

The third step is to sum the individual vessel valuations to produce the Fleet Market Value — the total value of the fleet if every vessel were sold in the current secondhand market at broker-assessed prices.

Note: it is important to take the lowest vessel valuation in a cycle as the valuation is dependant on market cycle as well. This helps us to identify the floor.

The fourth step is to subtract net debt — total debt minus cash and cash equivalents — to arrive at the implied equity value of the fleet on a liquidation basis.

The fifth step is to compare this Fleet NAV to the current market capitalisation of the equity. The ratio of market cap to Fleet NAV — the Price-to-Fleet-NAV multiple — is the single most informative metric available for evaluating shipping equity value. When this ratio falls below 0.7x, the market is pricing the equity at a discount to the immediate liquidation value of the fleet. This means: if you bought all the equity and all the debt at current prices, you could immediately sell every ship in the secondhand market, retire all the debt, and have money left over. The business itself — the management team, the chartering relationships, the operational history — has implied negative value in the market’s pricing. That condition has historically been temporary.

When the Price-to-Fleet-NAV ratio exceeds 1.3 to 1.5x, the opposite is true. The market is pricing the equity at a premium to the fleet’s immediate liquidation value, implying that the investor is paying for future charter rates not yet reflected in current vessel prices. This premium may be justified if freight markets are in an early upcycle and vessel prices are lagging. It is not justified if the freight market is near a peak and vessel prices already reflect cycle-high assumptions. The discipline is the same either way: look at the asset price, not the earnings multiple.

Subscribe now

PART IV: STORIES FROM THE WATERLINE — THREE CASE STUDIES IN SHIPPING EQUITY MISPRICING

The Dry Bulk Supercycle and Its Aftermath: How the Same Ships Were Worth $90 Million, Then $12 Million, Then $30 Million

The dry bulk shipping cycle of 2003 to 2016 is the most dramatic and best-documented case study in shipping equity valuation history, and it illustrates the Fleet NAV framework with a clarity that no financial model can replicate.

In the years from 2003 to 2008, China’s industrial expansion created an unprecedented demand shock for seaborne dry bulk commodities. Iron ore and coal imports into China grew at double-digit annual rates. Every Capesize vessel in the world was fully employed, every day, at rates that were extraordinary by historical standards. The Baltic Dry Index, a composite of bulk carrier time charter rates, rose from approximately 1,000 points in early 2002 to a peak of 11,793 points in May 2008 — an eleven-fold increase in daily charter rates over six years. Vessel owners who had bought ships in 2001 and 2002 at depressed prices were generating cash flows that made the purchase price look trivial within a couple of years.

The secondhand market for vessels reflected this. A five-year-old Capesize that had been worth approximately $20 to $25 million in 2001 was worth $90 to $95 million by 2007 to 2008. Shipping companies that had bought vessels cheaply in the early 2000s were sitting on fleets that had appreciated 4 to 5 times in the secondhand market — appreciation that appeared nowhere on the GAAP balance sheet because there is no mechanism for writing vessels up to market value under the cost model. The earnings were high, the dividends were large, and shipping stocks traded at significant premiums to their GAAP book values. But Fleet NAV — the actual liquidation value of the fleet — was several times book value for companies with old, cheap vessels purchased in the prior trough.

Then the global financial crisis hit in late 2008. Chinese industrial growth slowed abruptly. Simultaneously, the newbuilding orders placed during the 2003 to 2008 boom began delivering — shipping companies and speculators had ordered thousands of new vessels at the height of the cycle, and those vessels arrived in 2009, 2010, and 2011 into a market with dramatically lower demand. The Baltic Dry Index fell from its May 2008 peak of 11,793 to 663 by December 2008 — a 94% collapse in daily charter rates over seven months. Vessel prices followed.

By 2016, a five-year-old Capesize that had been worth $90 million in 2007 was worth approximately $20 to $25 million. A ten-year-old Capesize — the vessels that had been five years old in 2011, purchased at the elevated prices of the 2007 to 2011 period — was worth approximately $8 to $12 million in the secondhand market. The companies that had taken on debt to acquire these vessels at 2007 to 2008 prices, and had financed them with loans against their elevated valuations, were in various states of financial distress. Their GAAP balance sheets showed vessel carrying values that were declining slowly through straight-line depreciation from their original high acquisition costs — values that were still well above the secondhand market, because the accounting depreciation had not been steep enough to track the collapse in vessel prices.

In this environment, the trap for conventional value investors was precisely described above: P/B ratios below 1.0x that appeared cheap were actually expensive relative to the real Fleet NAV. Diana Shipping’s $1.42 billion fleet book value versus $675 million broker valuation — a 54% overstatement — meant that a stock trading at 0.8x GAAP book was actually trading at 1.7x the true fleet value. Cheap by one measure. Expensive by the correct measure.

The Fleet NAV framework identified the distinction correctly. Companies with fleets whose GAAP book value had already been written down close to secondhand market values — either through impairments taken during the downturn or through operating histories that had built up significant accumulated depreciation — were trading at Fleet NAV multiples below 1.0x, and represented genuine value. Companies with overvalued legacy fleets at inflated GAAP book values were not cheap regardless of where their P/B appeared to be.

The cycle turned. From 2016 to 2019, freight rates recovered modestly. Then, in 2020 to 2022, a combination of COVID disruptions — which initially crushed demand but then created massive supply chain bottlenecks and demand surges as economies reopened — produced an extraordinary freight cycle. The BDI, which had been languishing below 1,500 for much of 2019 and early 2020, reached 5,650 by October 2021 — the highest in a decade. The ten-year-old Capesize that had been worth $12 million at the COVID trough in early 2020 was worth $33 million by 2022. The same iron and steel, the same engine, the same propeller — worth 175% more in 24 months because the market price of the freight service it provided had tripled.

The companies that survived the 2016 trough with manageable debt levels, rebuilt their balance sheets through the modest recovery of 2017 to 2019, and held their fleets through the pandemic disruption emerged into the 2021 to 2022 period with fleets trading at large premiums to their GAAP book values — premiums that the GAAP model could not show because there is no write-up mechanism. Fleet NAV told the story at every point in the cycle. GAAP book value told the wrong story at every critical inflection.

Subscribe now

Scorpio Bulkers at $2.98: What the Broker’s Quote Sheet Said That the Earnings Model Could Not

In late 2016, J. Mintzmyer of Value Investor’s Edge published a fleet NAV analysis of Scorpio Bulkers (then NYSE: SALT, later renamed Scorpio Holdings) using VesselsValue.com broker data. The stock was trading at approximately $2.98 per share. The company had a fleet of Ultramax and Supramax bulk carriers — smaller than Capesize but more flexible, able to serve a wider range of ports and cargo types.

The analysis was not about earnings. Scorpio Bulkers was earning very little at 2016 freight rates — the Baltic Dry Index was near multi-decade lows, and a bulk carrier owner with fixed operating costs and minimal charter revenue was barely breaking even, if at all. P/E was either undefined or extremely high. EV/EBITDA was meaninglessly high. Any earnings-based valuation methodology produced either no output or a deeply uninviting number.

The Fleet NAV analysis produced a completely different picture. VesselsValue assessments for Scorpio’s specific fleet of modern Ultramax and Supramax vessels — adjusted for their individual ages, fuel efficiencies, and survey statuses — produced a total Fleet Market Value that, after subtracting net debt, implied a per-share Fleet NAV materially above the $2.98 stock price. The market was pricing Scorpio’s equity at a significant discount to the immediate liquidation value of the fleet. This meant that a buyer purchasing 100% of Scorpio at $2.98 per share could theoretically sell every ship in the secondhand market at current broker prices, retire all the debt, and recover more than the purchase price. The operating franchise itself — the management team, the chartering relationships, the future earnings potential of the fleet — was being offered for free, or at negative implied value.

The thesis was not that freight rates were about to recover immediately. The thesis was strictly about the asset value. The ships were worth more, in a documented and verifiable sense, than the market implied by the stock price. Shipping markets eventually reflect the value of the underlying assets through one of two mechanisms: either freight rates recover, improving earnings and causing the market to re-rate the equity higher; or the assets are sold in the secondhand market, directly realising the Fleet NAV. In either case, the investor who owned the equity at a discount to Fleet NAV was positioned to benefit.

Within a year of the $2.98 analysis, Scorpio Bulkers had returned over 150% as freight markets recovered and vessel values moved higher. The return was not predicted by any earnings model, because the earnings recovery itself was not predicted — freight rates are notoriously difficult to forecast. The return was produced by buying an asset below its immediately observable market value and waiting for the market to recognise that value. The broker’s quote sheet was right. The earnings model was useless.

Subscribe now

2020: COVID Hits, VLCCs Spike, and the Asset Price Tells a Different Story From the Fear

In 2020, the COVID pandemic produced one of the strangest freight rate dynamics in modern shipping history, and it illustrates how the asset-price framework handles crisis conditions where the earnings model produces the most misleading signals.

When the pandemic began in early 2020, crude oil demand collapsed as the global economy shut down. OPEC+ production disputes, simultaneously, were pushing Saudi Arabia and Russia to flood the market with supply at exactly the moment demand was falling off a cliff. The result was a storage problem of historic proportions: the world had more crude oil than it had capacity to store onshore. The solution was floating storage — chartering Very Large Crude Carriers (VLCCs) to anchor offshore and hold crude oil until onshore storage became available. VLCC time charter rates, which had been approximately $30,000 to $40,000 per day in late 2019, spiked to over $200,000 per day in March and April 2020 — the highest rates in years, occurring simultaneously with the deepest economic contraction in decades.

For an investor running a conventional earnings model, this was completely uninterpretable. Earnings had spiked to extraordinary levels — but the reason was a distortion that would clearly not last once floating storage was released. A P/E multiple applied to peak-COVID VLCC earnings produced a valuation that no sane person would pay for a sustainable business. The spike in earnings was real but it was also obviously temporary, making the earnings metric actively misleading as a valuation input.

The Fleet NAV framework handled this correctly. A five-year-old VLCC in early 2020, before the storage spike, was worth approximately $80 to $90 million in the secondhand market. The storage spike caused some upward movement in vessel values as operators rushed to add capacity, but the structural valuation of the vessels was not changed fundamentally by a temporary rate spike. A Fleet NAV analysis in March 2020 — at the peak of the rate spike, when fear about the overall economy was simultaneously driving tanker equity prices down despite the high rates — would have shown VLCC owners trading at discounts to their Fleet NAVs because the equity market was pricing economic uncertainty while the asset market was pricing the vessel’s intrinsic value independently. The investors who used Fleet NAV rather than current earnings to evaluate tanker companies in April 2020 were buying the steel at a discount while everyone else was trying to decide whether the spike was real or fake.

The 10-year-old Capesize that was worth $19 million at the COVID trough in early 2020 and $33 million by 2022 is the simpler illustration of the same point. The fear that compressed vessel values at the COVID trough was real — demand uncertainty was genuinely high, and the secondhand market correctly reflected some of that uncertainty through lower prices. But the structural economics of the vessel — its ability to carry iron ore from Brazil to China at a daily rate that covers its operating costs with margin to spare — did not disappear because of COVID. The investor who looked at the $19 million vessel value in March 2020 and the enterprise value of the company holding it, and found that the market was pricing the equity below the fleet’s own assessment of what the ships could be liquidated for, was identifying a temporary mispricing — not a permanent impairment.

Subscribe now

CONCLUSION: THE ASSET HAS A PRICE TAG — USE IT

Every other sector in this series of white papers requires the investor to do work that the accounting system has declined to do: count the uranium reserves, measure the copper in the ground, calculate the CDI on the deposit franchise, estimate the replacement cost of the transmission grid. In each case, the investor is building a measurement framework from scratch because the accounting system provides no useful starting point.

Shipping is different. The asset prices are transparent (just less accessible). VesselsValue.com and Clarkson Research will tell you, to the nearest million dollars, what each specific named vessel in the fleet sold for most recently and what comparable vessels are worth today. The market for your primary asset is as transparent, as liquid, and as continuously quoted as the market for crude oil or copper. You do not need a model. You need a spreadsheet, a fleet list, and a broker’s data subscription (which AI can replace now — in terms of what we need). As a guideline:

The market’s persistent failure to use this straightforward tool — and its preference instead for earnings multiples that are volatile, cyclically misleading, and anchored to an accounting system that reflects none of the vessel market’s dynamics — is what creates the opportunity. Every time the dry bulk or tanker cycle turns down, every time freight rates fall and earnings disappear, the earnings model produces either a meaningless or deeply uninviting result. Analysts stop covering the companies. Investors sell. The P/B screens light up with apparently cheap stocks that are, by the earnings model, unanalysable.

The Diana Shipping balance sheet said $1.42 billion in 2016. The broker’s quote sheet said $675 million. The stock looked cheap on P/B. It was expensive on Fleet NAV. And the investor who could distinguish those two numbers was the one who avoided the trap.

The Scorpio Bulkers stock was $2.98 in 2016. The Fleet NAV said the ships were worth more than the price implied. Earnings were near zero. The earnings model offered nothing. The Fleet NAV offered a clear signal: the market is selling you steel below its current market price. A year later, the stock had returned over 150%.

The ten-year-old Capesize was worth $19 million in March 2020. It was worth $33 million in 2022.

This white paper is for educational and informational purposes only. Fleet valuation data is sourced from Baltic Exchange Sale and Purchase Assessments, VesselsValue.com broker methodology, Clarkson Research Services, and Xclusiv Shipbrokers as cited in Lloyd’s List and Seatrade Maritime reporting. Diana Shipping fleet book value data is drawn from the company’s Form 20-F filings with the SEC for the relevant years. The Fleet NAV methodology and specific transaction prices cited are consistent with documented sources including Value Investor’s Edge and contemporaneous shipping industry publications. The Mineral Shougang International transaction price is sourced from Lloyd’s List citing shipbroker reports for late 2025. Nothing herein constitutes investment advice.

If you like this read, check out my other reads below 👇

There is an executive I know (for real, and not a AI generated story) — no longer in aluminium, now an executive in a low-key biopharma startup — who spent a decade inside the industry before most analysts covering aluminium companies had visited a smelter. She is, in the old Millionaire Next Door sense, quietly comfortable. She does not appear on CNBC. She does not manage a public fund. She made her money the way people who genuinely understand industrial businesses make it: by identifying the variable that everybody else models incorrectly and betting on it with conviction.

Her insight, stated plainly, was this: in aluminium production, the only number that ultimately determines whether a company is extraordinary or ordinary is the price it pays for electricity. Not the aluminium price. Not the revenue per tonne. Not the EBITDA margin. Those are all downstream consequences of the energy cost. The company that gets its electricity cheapest wins. And the company that figured out how to get electricity from wind turbines — when wind was still considered exotic for heavy industrial applications — went to the moon.

She does not name the company publicly. She does not need to. The principle she identified is so clear, so mechanical, and so consequential that once you see it, you cannot unsee it. This paper will show you exactly what she saw, why the market chronically fails to price it correctly, and where — right now — the same gap exists between what accountants record, what financial models produce, and what a business owner paying for the real asset would calculate.

Subscribe now

PART I: THE PHYSICS BEFORE THE FINANCE

Why Electricity Is Not One Cost Among Many

Most manufacturing businesses have a range of meaningful input costs. Labour, raw materials, logistics, energy. None of them individually dominates the economics. In a typical automotive parts manufacturer, energy might represent 5 to 8 percent of total production cost. In a food processing plant, perhaps 3 to 5 percent. These are significant but not singular.

Aluminium smelting is categorically different. The Hall-Héroult process — invented independently in 1886 by Charles Martin Hall in Ohio and Paul Héroult in France, and unchanged in its fundamental physics ever since — converts aluminium oxide (alumina) into aluminium metal by passing enormous electrical current through a molten bath of cryolite at approximately 960 degrees Celsius. The electrolytic cells, called pots, each draw roughly 150,000 to 500,000 amperes of current. A modern smelter with 200,000 tonnes per year of capacity runs approximately 400 to 800 of these pots simultaneously, consuming roughly 14 to 16 megawatt-hours of electricity for every tonne of aluminium produced.

Pause on that number. Fourteen to sixteen MWh per tonne. At a global average electricity price of approximately $0.07 per kWh, that is $980 to $1,120 per tonne in electricity cost alone. Aluminium’s current market price is approximately $2,400 to $2,600 per tonne. This means electricity represents 38 to 47 percent of the total revenue of every tonne sold. Remove the electricity cost from the business and you have revealed a powerful, high-margin manufacturing operation. Leave electricity priced at global average rates and the business barely survives. Change the electricity price from $0.07 to $0.02 per kWh — which is exactly what Siberian hydropower and Icelandic geothermal deliver — and you have created a structural competitive advantage that no competitor can replicate without access to the same physical geography.

This is the axis around which the entire aluminium industry rotates. Not the aluminium price. Not the bauxite grade. Not the labour cost. The electricity cost. The executive who understood this — who looked at companies not through their earnings models but through their power contracts — was not making a financial insight. She was making a physics insight.

And the finance followed from the physics, exactly as it always does when you are looking at the right thing.

Subscribe now

PART II: THE ASSET, PRECISELY DEFINED

The Smelter and the Power Contract — Two Assets in One Package

The primary income-generating asset of an aluminium producer is a combination of two things that are legally and physically inseparable in practice but completely separable in valuation terms. The first is the smelter: the physical potroom, with its reduction cells, rectifiers, alumina handling systems, metal casting infrastructure, and permitted environmental controls. The second is the power contract: the long-term agreement that delivers electricity to the smelter at a specific price per kilowatt-hour, for a specific number of years, at a specific volume.

The smelter without the power contract is an expensive building with no economic purpose. The power contract without the smelter is an electricity purchase agreement with no industrial application. Together, they are the machine that converts electrical energy into aluminium metal at whatever cost structure the power contract specifies. A smelter with a $0.02/kWh power contract is a licence to print money when aluminium prices are anywhere above $2,000/tonne. A smelter with a $0.06/kWh power contract loses money at $2,200/tonne and survives only when prices spike above $2,600/tonne.

This is why the executive’s insight was so valuable. She was not looking at earnings models. She was asking a prior question: what does this company pay for its electricity? That single number, applied to the physics of the Hall-Héroult process, gives you the cost structure of every tonne the company will ever produce, for the duration of the power contract. The earnings model calculates the margin after the fact. The owner’s analysis starts from the power price and works forward.

Subscribe now

What a Business Owner Would Pay Right Now

The owner’s valuation of an aluminium smelter has three inputs, all of which can be calculated today without a single forecast.

Input one: the net margin per tonne at current aluminium prices and current power cost. At $2,500/tonne aluminium and $0.02/kWh power cost, the electricity component of production cost is approximately $280 to $320 per tonne. Add alumina cost of approximately $350 to $400/tonne, and other direct costs of approximately $200 to $250/tonne, and total cash production cost is approximately $830 to $970/tonne. Net operating margin: approximately $1,530 to $1,670 per tonne. At $0.05/kWh power cost, the electricity component is approximately $700 to $800/tonne, total cash cost rises to $1,250 to $1,450/tonne, and the net margin collapses to $1,050 to $1,250/tonne — still positive but already dramatically different. At $0.08/kWh — the approximate current European grid electricity cost for industrial users after the post-Ukraine energy crisis adjustments — cash production cost exceeds $1,600/tonne and the margin is essentially zero, or negative at times.

Input two: annual production capacity. A modern potline of 200,000 tonnes per year produces 200,000 tonnes of aluminium per year. This is a physical fact about the smelter that does not change materially with commodity cycles.

Input three: the remaining term of the power contract. A power contract with 20 years remaining at $0.02/kWh locks in the cost advantage for 20 years. This is the duration of the asset’s value. No forecast required. The contract exists. Its terms are known. The physical throughput is calculable from operating capacity data.

Multiply net margin by annual tonnage and you have the gross annual operating income from the asset. The owner then asks: what is the remaining life of this margin stream? And what would it cost to replicate it? The answer to the last question — what would it cost to replicate it — is the most important number in the analysis, and it is the number that GAAP entirely fails to capture.

Subscribe now

PART III: WHY YOU CANNOT BUILD A NEW ALUMINIUM SMELTER IN THE DEVELOPED WORLD

The Greenfield Problem

Century Aluminum’s Robards, Kentucky facility, completed in the late 1990s, was the last significant greenfield primary aluminium smelter built in the United States. In the roughly 25 years since its completion, not a single new primary aluminium smelter has been built in the US at commercial scale. In Europe, the story is similar: the last major new European smelter was Alcoa’s Fjarðaál facility in Reyðarfjörður, Iceland, which began production in 2007 — and Iceland is an outlier precisely because of its renewable energy abundance, not despite it.

The reason is simple and physical. A new aluminium smelter requires a long-term power supply contract at approximately $0.02 to $0.03 per kWh to be economically viable in today’s aluminium price environment. At $0.05/kWh or above, the smelter’s cash cost of production is structurally uncompetitive against the global cost curve. No US utility will offer a 20-year industrial power contract at $0.02/kWh today, because the utility’s own marginal cost of new generation — whether from wind, solar, gas, or nuclear — is approximately $0.05 to $0.08 per kWh at the levelised cost level. Offering industrial users a price below the cost of generating the electricity is not something a profit-motivated utility will do.

This creates a category of asset that has no parallel in most industries: a fully functioning, operating smelter with a legacy power contract at below-replacement-cost electricity prices is not merely cheap relative to its replacement cost — it is irreplaceable in practice. You cannot build a new one. The power contract that makes the existing one economically valuable cannot be reproduced on any timeline at any capital budget. The asset’s advantage is not a patent, not a customer relationship, not a network effect. It is physics and geography. The Bratsk hydropower dam on the Angara River in Siberia, commissioned in 1961 and producing 4,500 megawatts of electricity at a variable operating cost close to zero, is a geological and engineering fact that cannot be moved, replicated, or competed away.

Subscribe now

PART IV: THE GAAP PROBLEM — AN ASSET THAT GETS CHEAPER ON PAPER AS IT BECOMES MORE VALUABLE IN REALITY

Straight-Line Depreciation and the Fully-Amortised Potroom

Under US GAAP (ASC 360) and IFRS (IAS 16), property, plant, and equipment is carried at historical cost minus accumulated depreciation. The useful life assigned to a primary aluminium smelter in standard accounting practice is typically 20 to 25 years for the potline infrastructure — the reduction cells, the pot shells, the electrical busbars, and the rectifier equipment. A smelter built in 1995 on a 20-year accounting life was fully depreciated by 2015. If that smelter is still producing aluminium in 2026 — which many are — its net book value on the balance sheet is approximately zero, or whatever residual scrap value the accountants assigned at the outset.

This is where the gap between accounting value and economic value becomes most acute. A fully depreciated potroom that continues operating has a balance sheet value approaching zero. Its economic value — the value to an owner who understands what it produces and what it costs to produce it — is determined entirely by the power contract and the margin spread, neither of which appears on the balance sheet.

In Century Aluminum’s December 31, 2023 10-K, the company reported net PP&E of $1,004.2 million across all its operations, including smelting capacity in Kentucky, South Carolina, and Iceland. The annual depreciation charge was $74.7 million. The critical question is not what these assets cost to build or what their depreciated book value is — it is what the power contract attached to each of these facilities costs per kilowatt-hour and how many years of that contract remain. The balance sheet tells you nothing about either of those questions. The power contract is not an asset recognised on the balance sheet.

It is disclosed in operating lease commitments and contractual obligations tables, buried in the back half of the annual report, in language that requires industry-specific knowledge to interpret correctly.

The earnings model — which most analysts apply as the primary valuation tool — uses the reported EBITDA figure and applies a sector multiple. The EBITDA is a product of the current aluminium price and the current production cost, which includes the current power cost under the existing contracts. But the earnings model treats the power cost as a variable input — something that will fluctuate with energy markets — rather than as a fixed contractual right at a specific price for a specific remaining term. This is the core error. The power contract is not a variable. It is a signed legal commitment. It is as fixed and certain as a government bond, for the duration of its term.

Subscribe now

PART V: STORIES FROM THE POTROOM — HOW THE POWER CONTRACT CREATED AND DESTROYED FORTUNES

The Siberian Miracle: How RUSAL’s Founders Saw What Nobody Else Saw

In March 2000, two oligarchs — Oleg Deripaska of Sibirsky Aluminium and Roman Abramovich of Millhouse Capital — merged their Russian aluminium assets to create United Company RUSAL. The transaction was not primarily a bet on aluminium prices. It was a bet on a specific physical reality that Soviet planners had created decades earlier without fully understanding its long-term commercial implications.

Soviet industrial planners in the 1950s and 1960s, designing the electrification of Siberia, had a problem: the enormous hydroelectric stations they were building on the Angara and Yenisei river systems — dams of extraordinary scale, including the 4,500-megawatt Bratsk dam on the Angara and the 6,000-megawatt Krasnoyarsk dam on the Yenisei — were producing electricity far in excess of what the sparse Siberian population needed for residential and light industrial use. The solution was to build aluminium smelters directly adjacent to the dams, specifically to consume the surplus electricity that could not be transmitted efficiently to distant population centres.

The Bratsk aluminium smelter, commissioned in 1966, was built literally 20 kilometres from the Bratsk hydropower station. The smelter consumed approximately 75% of the power station’s total output. The Krasnoyarsk smelter, commissioned in 1964, was similarly co-located with the Krasnoyarsk hydropower dam. The electricity price the smelters paid to the power stations was set administratively by Soviet planners at levels designed to utilise the generation capacity — not at market-clearing rates. By the time Deripaska and Abramovich were assembling their aluminium empire in the late 1990s, these Soviet-era power agreements had evolved into commercial contracts that still reflected the fundamental reality: electricity generated by a fully amortised 1960s hydropower dam, with no fuel cost and minimal operating cost, is among the cheapest electricity available on earth.

A Canadian Mining Journal journalist who toured the Bratsk facility in 2004 documented the power price: 25.3 kopecks per kilowatt-hour — at then-prevailing exchange rates, approximately $0.009 per kilowatt-hour. Less than one US cent per kilowatt-hour. Apply the Hall-Héroult physics: at $0.009/kWh and 15 MWh per tonne of aluminium, the electricity cost per tonne of metal was approximately $135. At London Metal Exchange aluminium prices of $1,600 to $1,800 per tonne in the mid-2000s, and an all-in cash cost including alumina, labour, and overhead of perhaps $700 to $800 per tonne, the gross operating margin from Bratsk was approximately $800 to $1,000 per tonne — extraordinary by any industrial standard, sustained at scale across a million tonnes of annual capacity.

This is what Deripaska saw that the Western financial press, focused on the geopolitics of Russian oligarchs rather than the thermodynamics of Siberian hydro, missed entirely. The Bratsk and Krasnoyarsk smelters were not just industrial assets. They were machines attached to essentially free power sources, producing a commodity that trades on a global exchange at prices set by global supply and demand rather than by Russian production costs. The arbitrage between a $0.009/kWh electricity cost and global aluminium prices was structural, physical, and — for as long as the power agreements held — unassailable.

When RUSAL listed its shares on the Hong Kong Stock Exchange in January 2010, raising approximately $2.24 billion, it became possible for the first time to buy equity in a company whose primary competitive advantage was this power cost arbitrage. The listing prospectus disclosed that RUSAL obtained 93% of its electricity from hydroelectric sources at costs dramatically below global market rates. The balance sheet, per standard IFRS treatment, carried the Siberian smelters at historical Soviet-era construction cost less decades of depreciation — near zero for the oldest facilities. The economic value of those same facilities — powered by hydro electricity at $0.01/kWh or below, producing aluminium that sells at $2,400/tonne globally — was orders of magnitude above their accounting carrying value.

Subscribe now

Iceland: When Cheap Electricity Moves the Smelter to the Power Source

Around the same time the Siberian hydro story was playing out quietly, a parallel story was developing in Iceland that illustrated the same principle from a different angle — and that eventually produced the insight the former executive is known for, in a slightly different form.

Iceland’s electricity grid is powered entirely by renewable energy: approximately 72% hydropower from glacially-fed rivers, 25% geothermal energy from the volcanic geology beneath the island, and an emerging contribution from wind. The levelised cost of hydropower from Iceland’s geologically ideal river systems — high flow rates, large elevation drops, minimal silt — is among the lowest in the world. Iceland’s national power company, Landsvirkjun, had surplus generation capacity that far exceeded the small island nation’s domestic residential and commercial demand.

The solution was the same one Soviet planners had reached in Siberia: co-locate industrial electricity consumers directly at the generation source. Iceland had no indigenous aluminium ore, no particular industrial tradition in metalworking, and no geographic proximity to any major aluminium market. It had one thing: extraordinary surplus electricity at extraordinarily low cost, in a politically stable jurisdiction with a well-educated workforce and functional rule of law.

By 2017, Iceland — a nation of approximately 360,000 people — was producing more primary aluminium than the entire United States. The three Icelandic smelters — Norðurál (Century Aluminum), Fjarðaál (Alcoa), and ISAL (Rio Tinto) — between them produced approximately 800,000 tonnes of aluminium per year, powered entirely by renewable electricity at costs that US domestic smelters, paying commercial industrial rates for grid power, could not come close to matching.

Century Aluminum’s Norðurál smelter in Grundartangi, Iceland, which began production in 1998, signed power contracts with Landsvirkjun that have been structured and renegotiated multiple times. The disclosed power costs at Norðurál — while not stated as a precise kWh price in Century’s US public filings — are reflected in the smelter’s C1 cash cost of production, which has historically been among the lowest in Century’s portfolio. The smelter’s net book value on Century’s consolidated balance sheet — after 25+ years of straight-line depreciation — approaches a small fraction of its 1990s construction cost. Its economic value — anchored to the remaining term of its Iceland power agreement, to the renewable electricity cost advantage that makes it one of the most competitive smelters on earth, and to the physical impossibility of replicating either the Icelandic hydro resource or the smelter’s permitted operating status — is vastly higher than any balance sheet figure suggests.

Subscribe now

The Wind Energy Pivot: The Company That Went to the Moon

The executive’s story points to something more recent and more instructive than RUSAL’s Soviet hydro inheritance or Iceland’s geological fortune. Those were advantages that came from where the smelters happened to be built. Her insight was about a company that actively went and found its power advantage — that made a strategic decision to secure cheap, renewable electricity when the market did not yet understand why that decision would be worth many multiples of its apparent cost.

The principle is the same regardless of whether the energy source is Siberian hydropower, Icelandic geothermal, or industrial-scale wind. The variable that matters is dollars per kilowatt-hour paid for electricity at the smelter fence. The company that reduces that number by 30 to 50 percent relative to grid-connected competitors does not reduce its costs by 30 to 50 percent — it transforms its margin structure entirely, because electricity is 35 to 45 percent of total cost. A 40 percent reduction in electricity cost, applied to a 40 percent cost share, reduces total production cost by 16 percentage points and expands the gross margin per tonne by the same absolute dollar amount.

At $2,500/tonne aluminium, moving a smelter’s electricity cost from $0.05/kWh to $0.03/kWh saves approximately $280 to $320 per tonne in direct electricity cost. On a 200,000-tonne smelter, that is $56 to $64 million per year in additional operating income — every year, for the life of the power agreement. Capitalised at a modest 8 to 10 percent rate, this single cost reduction creates $560 to $800 million in net present value for a smelter whose balance sheet carrying value might be $100 million or less. The power contract is the asset. The balance sheet does not show it. The earnings model does not isolate it. Only the owner who asks the right question — what do you pay per kilowatt-hour and for how long? — sees the full picture.

Subscribe now

PART VI: THE ACCOUNTING TREATMENT AND THE GAP IN NUMBERS

What GAAP Records vs. What the Owner Calculates

For Century Aluminum specifically: the company’s December 31, 2023 balance sheet shows net PP&E of $1,004.2 million across its US and Iceland operations. Annual depreciation is $74.7 million. The PP&E figure has grown in recent years due to capital expenditures rather than reflecting any revaluation (which far outweighs the value of any CAPEX improvements) — there is no mechanism under US GAAP to write up the value of a fully depreciated smelter, even when its earning power has increased because aluminium prices have risen while its power contracts have locked in a fixed electricity cost.

Century’s US smelters — Hawesville, Kentucky (approximately 244,000 tonnes/year capacity) and Sebree, Kentucky (approximately 166,000 tonnes/year) — have operated since the 1960s and 1970s respectively. Their net book values are minimal after 50+ years of depreciation. The power contracts that allow Hawesville and Sebree to operate are disclosed in Century’s contractual obligations tables. These contracts run for specific terms at specific prices. They are not on the balance sheet. They are the primary determinant of whether the smelter earns $200/tonne or $800/tonne in gross operating margin. They are, in every meaningful economic sense, the asset — and they are invisible in the PP&E line.

The owner’s perspective for a Century Aluminum smelter with a legacy power contract at, say, $0.03/kWh:

• Annual production: 200,000 tonnes

• Revenue at $2,500/tonne: $500 million

• Electricity cost at $0.03/kWh × 15 MWh/tonne × 200,000 tonnes: $90 million

• Alumina and other direct costs: ~$550 million combined

• Gross operating income: ~$360 million per year

A business owner willing to buy this cash stream for 5 years of gross operating income would pay approximately $1.8 billion for a single smelter — a facility that might carry $150 to $200 million in net book value on the balance sheet, from a company with a total market capitalisation that has ranged from under $500 million to approximately $1.5 billion across the 2020 to 2024 period. But will the smelter only last 5 years? — The book value says one thing, the owner’s perspective tells us another.

Subscribe now

The RUSAL Book-to-Value Chasm

For RUSAL, the gap between accounting value and economic value is even more extreme, for three compounding reasons. First, the Siberian smelters were built with Soviet-era capital — accounted for at official exchange rates that bore no relationship to market prices — and have been depreciated under various accounting regimes over decades, leaving carrying values that reflect neither current construction costs nor the economic productivity of the assets. Second, the power contracts that underpin the Siberian smelters’ structural cost advantage are internal agreements between RUSAL and the hydro assets that RUSAL itself partly owns — they are eliminated on consolidation and do not appear in disclosed contractual obligations in the same way a third-party power contract does. Third, the Russian regulatory and political environment adds geopolitical risk discounts to RUSAL’s valuation that further suppress the market price relative to the physical asset value.

The Bratsk smelter alone, the world’s largest at approximately 1,020,000 tonnes per year of aluminium output, is connected directly to the Bratsk hydropower station with electricity costs at a fraction of global industrial rates. The gross operating income from Bratsk, at mid-cycle aluminium prices and Bratsk’s disclosed cost structure, is extraordinary relative to any accounting book value of the asset. The carrying value of Soviet-built industrial infrastructure on a post-1990 accounting basis is essentially a historical fiction. The economic value — tonnes per year times the net margin generated by the power cost arbitrage — is the real number.

Subscribe now

PART VII: THE WALL STREET PROBLEM — MODELS THAT MEASURE EVERYTHING EXCEPT THE RIGHT THING

How the Earnings Model Fails Aluminium Producers

The standard approach for covering an aluminium producer on Wall Street begins with the LME aluminium forward curve. The analyst takes the market’s forward price for aluminium — which embeds consensus views about Chinese demand, supply curtailments, and global inventory levels — and uses it as the revenue driver. They then model production volumes (capacity times utilisation rate), subtract estimated cash costs, arrive at EBITDA, and apply a sector multiple.

This model has several structural problems for primary aluminium producers with power contract advantages.

The first problem is that the aluminium price is the most visible and most volatile input, so it dominates the model. When LME aluminium is at $2,100/tonne, analysts produce bearish price targets. When it is at $2,700/tonne, they produce bullish ones. But the competitive advantage of a Siberian or Icelandic smelter exists at every point in the aluminium price cycle — the low power cost is a fixed structural advantage, not a cyclical variable. The earnings model treats cost structure as relatively fixed and price as the swing variable. In reality, for a power-advantaged smelter, the variable is the spread between the locked-in power cost and the aluminium price — a spread that widens every time aluminium rises and narrows only in extreme downturns when aluminium falls below total cash cost.

The second problem is that the multiple applied to EBITDA treats all EBITDA equivalently. A dollar of EBITDA from a US grid-connected smelter paying $0.06/kWh, which will disappear if power costs rise, is not the same as a dollar of EBITDA from a Siberian hydro-powered smelter paying $0.01/kWh, which will persist regardless of external energy market conditions. The first EBITDA stream is fragile and contingent. The second is structural and durable. The multiple that correctly values the second is higher than the multiple that correctly values the first — but the earnings model, applying a uniform sector EV/EBITDA multiple, treats them identically.

The third problem is that earnings models perform their worst exactly when the valuation opportunity is greatest. In periods when aluminium prices are depressed — the 2015 to 2016 trough, the COVID trough of 2020 — earnings models show low or negative EBITDA for marginal producers, and even power-advantaged smelters show reduced earnings as the aluminium price compresses the margin spread. The earnings model says: sell. The owner’s analysis says: the power contract is unchanged, the production capacity is unchanged, and I am now buying the durable cost advantage at a lower aluminium price than it requires to be profitable. Buy.

Subscribe now

PART VIII: STORIES FROM THE MARKET — THREE CASE STUDIES IN ALUMINIUM VALUE ARBITRAGE

Century Aluminum: The Smelter That Was Worth Nothing and Then Everything

In the aftermath of the 2008 global financial crisis, aluminium prices collapsed from over $3,200/tonne at their mid-2008 peak to below $1,300/tonne by early 2009. For aluminium producers without low-cost power, the collapse was existential. The Hawesville, Kentucky smelter — one of Century’s primary US assets — cut production and operated at reduced capacity. Century’s debt load, accumulated during earlier expansion, became a serious concern. The stock, which had been in the $60 to $80 range during 2007 and early 2008, fell to under $5 per share in early 2009.

At that stock price, Century’s market capitalisation was under $200 million. Its US smelter capacity — Hawesville and Sebree combined — was approximately 400,000 tonnes per year. The power contracts at those facilities, negotiated in prior years, had remaining terms and price structures that, at normalised aluminium prices, would generate hundreds of millions of dollars in annual operating income. The balance sheet showed minimal PP&E after decades of depreciation. The earnings model showed losses at $1,300/tonne aluminium. The owner’s analysis showed something different: these are the last remaining primary aluminium smelters in the United States with pre-negotiated, legacy power agreements, and I am buying 400,000 tonnes of capacity — infrastructure that cannot be replicated in the current US power market at any price — for approximately $500 per tonne of annual capacity in total market value. At normalised aluminium prices and the locked-in power cost, this capacity will generate $200 to $400 per tonne in annual gross operating income.

By 2011, as aluminium prices recovered and the macroeconomic situation stabilised, Century’s stock had recovered to $25 to $30 per share — a 5 to 6 times return from the 2009 trough. The smelters were the same smelters. The power contracts were the same power contracts. The aluminium production capacity was the same. Only the market’s willingness to price the asset at its owner’s value, rather than its current earnings, had changed.

The lesson is the same one the executive with the biopharma startup learned: the power contract is the asset. The earnings model at a cyclical trough does not reflect the asset’s value. An owner who calculates what the asset produces per year at normalised prices and costs — and pays for it based on that calculation, not on current earnings — earns extraordinary returns when the cycle turns.

Subscribe now

The European Smelter Catastrophe: When the Power Contract Expires and Nobody Has a Plan

The story of European aluminium smelters since 2021 is the mirror image of the Siberian and Icelandic stories — and it illustrates the power cost dependency from the disaster direction.

Germany was once home to a meaningful primary aluminium smelting industry. Aluminium Norf, operated by Novelis and Hydro, is the world’s largest aluminium rolling mill by volume — but it processes recycled aluminium, not primary metal. Germany’s primary aluminium capacity, relying on German grid electricity at industrial rates, was already under pressure before 2021. When Russia’s invasion of Ukraine in February 2022 triggered a spike in European natural gas prices — and because European electricity markets are marginal-cost priced off gas generation — German industrial electricity prices surged from approximately €50 to €60 per MWh in 2021 to over €300 to €400 per MWh at the height of the energy crisis in late 2022.

At €300/MWh electricity and the Hall-Héroult physics of 15 MWh per tonne of aluminium, the electricity cost per tonne of aluminium surged to €4,500 — nearly double the LME aluminium price of approximately €2,200/tonne equivalent at the time. German primary aluminium production became not merely unprofitable but catastrophically uneconomic. German primary aluminium output reportedly declined 30% in 2022 and a further 45% in 2023. One in four German aluminium smelters closed. The European Aluminium industry association warned of permanent capacity loss.

The closures were not temporary curtailments. When a potline is shut down completely — when the electrolytic cells are allowed to cool and the molten cryolite bath solidifies — the restart cost is enormous and the timeline is long. Pot relining, the process of rebuilding the insulating interior of each reduction cell, takes months and costs millions of dollars per cell across hundreds of cells. In many cases, the closed German smelters will not restart because the economics at European grid electricity prices do not support aluminium production at any foreseeable LME price.

This is the negative proof of the core thesis. The power contract — or the lack of one — is the entire business. Germany’s smelters, reliant on grid electricity with no long-term power price protection, were running a business that was structurally dependent on a cost input over which they had no control. When that input moved adversely, the business ceased to exist. The Bratsk smelter, powered by a hydroelectric dam on the Angara River at $0.009/kWh, was still producing aluminium profitably through the same period. The difference between the two outcomes is not management quality, not technology, not customer relationships. It is the price per kilowatt-hour.

Subscribe now

Metro Mining and Capral: The Upstream and Downstream Perspectives

Metro Mining Limited (ASX: MMI) operates the Bauxite Hills mine in Queensland, Australia, producing bauxite — the raw ore from which alumina is refined and ultimately aluminium is smelted. Metro’s position in the supply chain is upstream of the smelter — it provides the bauxite feedstock that flows to Chinese alumina refineries, which in turn supply Chinese smelters. The value arbitrage in Metro’s case is slightly different from the smelter-level story but illustrates the same accounting gap.

Bauxite is the primary ore of aluminium, with over 90% of global production utilized to produce alumina via the Bayer process, which is then refined into aluminium metal.

Metro’s Bauxite Hills deposit contains substantial proven reserves of bauxite carried at the historical cost of exploration and development. The bauxite trades in a market where Chinese demand is essentially price-inelastic within a reasonable range — Chinese aluminium production requires bauxite from external sources because China’s own bauxite deposits are lower quality than its aluminium production scale demands. The shipping cost from Queensland’s cape ports to Chinese refineries is a known, calculable variable. The net margin per tonne — bauxite sale price minus mining cost minus shipping minus royalties — is calculable from disclosed production data and current contract prices. The owner’s analysis of Metro’s reserve base applies the same methodology used throughout this paper series: tonnes in reserve times net margin per tonne at current prices, compared to the enterprise value. The balance sheet shows bauxite mineral rights at capitalised exploration and development cost. The owner sees raw material supply for a structurally inelastic industrial demand.

Subscribe now

PART IX: CONFIRMED CATALYSTS — WHAT IS ALREADY HAPPENING

Catalyst 1: Green Aluminium Premium — Already Priced Into Contracts, Invisible on Balance Sheets

The aluminium market has quietly bifurcated into two parallel products: conventional primary aluminium with a global average carbon footprint of approximately 16 to 17 tonnes of CO₂ per tonne of metal, and low-carbon or green aluminium produced from renewable energy sources with footprints of 2 to 4 tonnes CO₂ per tonne. RUSAL’s hydropower-produced “Allow” brand aluminium and Iceland’s geothermal and hydro-sourced metal command documented premiums in the market.

This premium has been growing as automotive, aerospace, and consumer electronics manufacturers — under pressure from their own customers and from regulatory bodies — commit to reducing the embedded carbon in their products. A Mercedes-Benz or BMW that uses Icelandic aluminium in its body panels has a lower Scope 3 emissions footprint than one using Chinese coal-powered aluminium. This is now a quantifiable, commercially relevant distinction that buyers are willing to pay for.

The premium for certified low-carbon aluminium has been reported in the range of $50 to $200 per tonne above the LME price, depending on certification standard and buyer requirements. For a power-advantaged smelter already producing at low cost — say, the Iceland or Siberian facilities — this is pure incremental margin on top of the existing cost advantage. The renewable energy infrastructure that creates the low power cost simultaneously creates the low-carbon product that commands the premium. One asset produces both advantages. Neither is fully reflected on the balance sheet.

RUSAL’s explicit strategy to market its Siberian hydro-powered aluminium as a premium product — evidenced by its corporate agreements with downstream manufacturers including the Budweiser Brewing Group for aluminium cans produced at the Krasnoyarsk hydropower smelter — is the beginning of a market segmentation that will structurally advantage renewable-powered producers over coal-powered ones. As the premium becomes more widely quoted and contracted, the value of the renewable power supply arrangement — which is not on any balance sheet — becomes more explicitly a component of the product’s commercial value.

Subscribe now

Catalyst 2: Carbon Border Adjustment Mechanism (CBAM) — Already Enacted, Phasing In

The European Union’s Carbon Border Adjustment Mechanism came into force on October 1, 2023, and moves into its full phase-in period from January 2026 through 2034, at which point it will impose a carbon price on imports of carbon-intensive products — including aluminium — based on the difference between the carbon price embedded in the imported product and the EU’s own carbon price under the Emissions Trading System.

The practical effect: Chinese aluminium produced with coal-fired electricity will face a carbon import tariff of approximately €50 to €80 per tonne or more, depending on the EU carbon price, when entering the European market. Icelandic aluminium produced with renewable electricity faces near-zero CBAM cost. Siberian aluminium — geopolitically complicated by Russia sanctions — faces its own structural access barriers. The net result is that renewable-powered Western producers gain a structural market access advantage in the EU market that is proportional to the carbon intensity difference between their production method and the global average.

CBAM is not a proposal. It is enacted law, in implementation. The aluminium industry is one of the six sectors explicitly covered in the initial CBAM scope. The mechanism will apply to aluminium ingot, alloyed aluminium, and aluminium products. Companies with renewable-powered smelters have a confirmed, legally mandated competitive advantage in the EU market from January 2026 forward.

Subscribe now

Catalyst 3: US Trade Policy and the Section 232 Tariff Stack on Aluminium

The United States reimposed and strengthened Section 232 tariffs on aluminium imports in 2025, adding to the existing tariff framework that had been in place since 2018. The 25% aluminium import tariff creates a domestic price floor that advantages US domestic producers — of which there are very few remaining — over imported metal. Century Aluminum, with its domestic US smelting capacity at Hawesville and Sebree, is one of a small number of US domestic producers with existing, operating capacity.

This is a direct policy confirmation of the irreplaceable value of existing domestic smelting capacity. The US government has now, across multiple administrations and political configurations, consistently chosen to protect domestic aluminium production capacity with tariffs — precisely because new capacity cannot be built economically, making the existing capacity a national industrial asset of strategic significance. The tariff makes the existing power contracts embedded in domestic US smelters worth more, because the market premium for tariff-protected domestic production adds to the spread between production cost and sale price.

Subscribe now

Catalyst 4: AI Data Centre and EV Battery Demand Increasing Electricity Demand, Tightening the Power Advantage Moat

A counterintuitive consequence of the AI data centre boom and EV charging infrastructure build-out is that they are consuming massive quantities of electricity — at market rates — that is tightening electricity availability for new industrial users across the US and Europe. Data centres are signing long-term power purchase agreements, locking in renewable electricity supply for 10 to 20 years at prices competitive with what new aluminium smelters would need. This is directly reducing the availability of cheap renewable power for hypothetical new smelter construction.

The effect is to make existing legacy power contracts — signed before the data centre boom when renewable electricity was available at lower prices — even more valuable relative to what a new entrant would face. The moat around existing power-advantaged smelters is widening, not narrowing, as data centres and EV infrastructure compete for the same renewable electricity supply. A power contract signed in 1998 or 2003 at $0.02 to $0.03/kWh, with 10 to 15 years remaining, is more precious today than it was five years ago — because the renewable electricity it locks in is now being bid for by data centre operators willing to pay $0.05 to $0.07/kWh. The existing contract holder owns something that cannot be replicated at anything close to the contracted price.

Subscribe now

CONCLUSION: THE WOMAN WHO SAW THE PHYSICS FIRST

The executive who quietly made her money in aluminium and then moved on to biopharma is not particularly famous. She did not write a book about it. She does not appear on conference panels discussing energy transition investment. She saw something simple and acted on it — which, in the history of industrial investing, is usually the entire story.

What she saw was that aluminium is an energy business masquerading as a metal business. The kilowatt-hours that flow into the potrooms are transformed into metal ingots through a process that has not fundamentally changed since 1886. The margin on each ingot is determined almost entirely by the price paid for those kilowatt-hours. And the price paid for those kilowatt-hours is determined by contracts and geography — locked-in agreements with power generators, located near hydroelectric dams or geothermal fields or wind farms — that the balance sheet does not record, that the earnings model does not isolate, and that the sector P/E multiple treats as no different from any other cost input.

The company she identified — the one that went to the moon because it pivoted to wind power for its aluminium production — did not go to the moon because of its management team or its customer relationships or its revenue growth. It went to the moon because it secured an electricity cost structure that made it one of the most profitable smelters per tonne on earth, from assets that its balance sheet valued at depreciated historical cost.

The owner who asks the right question — what do you pay per kilowatt-hour, for how many years, and how much aluminium does that produce at today’s price? — bypasses the earnings model entirely. They are calculating the value of the power contract embedded in the smelter, and comparing it to the enterprise value the market is assigning. When those two numbers diverge significantly — which they do, cyclically and structurally, at every point in the aluminium price cycle when the earnings model generates pessimism — the owner has an opportunity that the earnings model’s user cannot see.

The smelter does not need to be new to be valuable. It does not need a high book value. It does not need to show high earnings this quarter. It needs one thing: cheap electricity for a long time. Everything else follows from the physics.

This white paper is for educational and informational purposes only. All financial figures cited are drawn from publicly available sources including Century Aluminum’s Form 10-K filings as filed with the SEC, RUSAL’s Hong Kong Stock Exchange disclosures and corporate history materials, Landsvirkjun and Askja Energy reporting on Iceland’s aluminium power contracts, and industry publications as cited throughout. Nothing herein constitutes investment advice.

If you like this read, check out my other writings 👇

Yet here is the peculiarity: on their books, these licenses are carried at values that bear almost no relationship to what the open market has repeatedly demonstrated they are worth. Some licenses sit at precisely zero dollars. Others are recorded at purchase prices from acquisitions completed decades ago. None reflect the billions of dollars that wireless carriers and the U.S. government itself have paid for comparable spectrum rights in arms-length transactions over the past two decades.

This white paper explores the anatomy of that discrepancy — not through the lens of discounted cash flow models, price-to-earnings multiples, or Wall Street forecasting frameworks, but through the eyes of a business owner who understands what it would actually cost to replace these assets today. We will trace how a quirk of accounting rules, combined with the peculiar history of American broadcasting regulation, has created a genuine value arbitrage — one that several confirmed catalysts now appear poised to unlock.

Subscribe now

The Hidden Value In Plain Sight

Subscribe now

What a Broadcast License Actually Is

Imagine you own a plot of land at the busiest intersection in a major city — a piece of real estate that no one else can replicate. The government granted you that land decades ago for free, and under the rules of the system, you get to keep it indefinitely so long as you continue using it. You can build on it, generate income from it, sell it to someone else, or lease portions of it. But on your personal balance sheet, you carry it at either zero or whatever you paid for it when you last bought a neighboring plot in 1995.

That, in essence, is the position of every major television broadcaster in America with respect to its FCC spectrum licenses.

These licenses are not mere pieces of paper. They are exclusive rights to use a specific slice of the electromagnetic spectrum — a finite, government-allocated natural resource that underlies everything from television broadcasting to mobile phone service to military communications. Each license covers a defined geographic area and a defined frequency band. The standard industry metric for valuing these rights is dollars per megahertz of bandwidth per person covered, expressed as $/MHz-POP.

When the FCC conducted its landmark Broadcast Incentive Auction in 2017 — an event that would ultimately redistribute a significant portion of the television broadcast band to wireless broadband providers — television broadcasters who chose to sell their spectrum rights back to the government realized proceeds that, for many, exceeded the entire market value that public equity markets had assigned to their companies.

Subscribe now

The Auction That Changed Everything

The 2017 FCC Incentive Auction was a landmark event in telecommunications, effectively a massive “repacking” of the airwaves. To understand why it was so complex, we first need to look at the two gears that made the machine turn.

Reverse vs. Forward Auctions: The Basics

In a typical auction, you have one seller and many buyers competing to pay the highest price. The Incentive Auction was unique because it used two different directions to find a market equilibrium.

• Forward Auction (The Standard Way): This is what most people recognize. There is one seller (the FCC) and many buyers (wireless carriers like T-Mobile). Bidders compete against each other, driving the price up until only the highest bidders remain.

• Reverse Auction (The “Buy-Back”): Here, the roles are flipped. There is one buyer (the FCC) and many sellers (TV broadcasters). The FCC offers a high price to buy back spectrum rights, and then gradually lowers that price in rounds. Broadcasters must decide if they are willing to “sell” their airwaves at the lower price. If they say no, they keep their station but lose out on the payout. This allows the buyer to find the lowest price a seller is willing to accept.

The results were staggering by any measure.

In the reverse auction, 175 television stations collectively pocketed $10 billion for spectrum rights—assets many had used for decades under licenses originally granted for free. The payouts were transformative: the largest single check reached $304 million, while 36 stations (including 11 non-commercial, educational outlets) walked away with over $100 million each.

On the flip side, the forward auction proved just how hungry the market was for low-band “beachfront” spectrum. Wireless carriers and private bidders paid $19.8 billion for 70MHz of licensed airwaves, marking the second-largest haul in FCC history. Nationwide, the price averaged $0.93 per MHz-POP, though competition in the top 40 metropolitan markets pushed that figure to $1.31.

The spending was dominated by two major players: T-Mobile, which led the pack with an $8 billion investment, and DISH Network, which committed $6.2 billion to secure its future in the wireless landscape.

These are not theoretical numbers. They represent actual cash that changed hands between sophisticated, arms-length counterparties in a transparent auction process. They tell us, with precision, what the market is willing to pay for broadcast-quality spectrum when given the opportunity.

And yet, for the broadcasters who chose not to participate — who retained their licenses and continue operating their television stations — those same licenses remain on their books at values that are, in many cases, a tiny fraction of what the auction demonstrated they could fetch.

Subscribe now

Why the Books Tell a Different Story

To understand the discrepancy, one must understand the accounting rules that govern how these assets are treated under Generally Accepted Accounting Principles (GAAP).

Under ASC Topic 350, FCC broadcast licenses are classified as indefinite-lived intangible assets. They are not amortized — meaning their carrying value is not systematically reduced over time — but they are subject to annual impairment testing. If an impairment test determines that the fair value of a license has fallen below its carrying value, a write-down is recorded. But critically, if the fair value has risen — which, as the 2017 auction demonstrated, has been the case for many broadcast licenses — no upward adjustment is permitted.

Furthermore, the method by which licenses are initially recorded on the balance sheet depends entirely on how they were obtained. Original broadcast licenses granted by the FCC during the analog era — often issued at no cost to the broadcaster — are carried at zero. Licenses acquired through the purchase of a television station are recorded at their allocated purchase price within that specific transaction. The result is a patchwork: a license acquired in a 1995 station deal sits at its 1995 allocation; a license acquired in a 2015 deal sits at its 2015 allocation; a license granted for free in 1952 sits at zero.

None are marked to current market comparables.

This accounting treatment is perfectly legal and entirely consistent with GAAP. But for anyone trying to understand what a television station group is actually worth, it creates a profound distortion. The single most valuable asset on the enterprise’s economic balance sheet is either invisible or dramatically understated on its accounting balance sheet.

Subscribe now

How a Business Owner Values The Asset

Wall Street analysts and investment bankers have their own frameworks for valuing broadcasting companies. They build discounted cash flow models projecting advertising revenue, retransmission consent fees, and political advertising cycles. They apply enterprise value-to-EBITDA multiples derived from comparable public company trading ranges. They stress-test assumptions about cord-cutting, audience fragmentation, and the secular decline of linear television. They produce target prices and issue buy, hold, or sell recommendations.

A business owner — someone who actually operates television stations, understands the regulatory landscape, and has watched the spectrum market evolve over decades — approaches valuation from an entirely different direction.

The business owner asks a simpler, more fundamental question: What would it cost to replace this asset today?

This is not an academic exercise. It is the most practical question one can ask about any asset. If you own a warehouse, what would it cost to build an equivalent warehouse on equivalent land today? If you own a patent, what would a competitor pay to develop an alternative technology? If you own an FCC broadcast license covering 6 MHz of spectrum reaching 10 million people in a top-20 market, what would it cost to acquire equivalent spectrum rights in that market through any available channel?

The answer, based on the best available evidence from actual transactions, is clear: far more than the carrying value on the books.

Subscribe now

The Replacement Cost Framework

Let us work through a simplified example to illustrate the business owner’s lens.

Consider a hypothetical television station in a market like Denver, Colorado — a top-20 designated market area with approximately 1.8 million television households, translating to roughly 4.5 million people in the coverage area. The station operates on a 6 MHz channel in the UHF band. Its broadcast license was originally granted by the FCC decades ago and is carried on the books at zero, or perhaps at a few million dollars allocated from a station acquisition in the 1990s.

To value this license on a replacement cost basis, a business owner would look at the MHz-POP math: 6 MHz of bandwidth multiplied by approximately 4.5 million people covered equals 27 million MHz-POPs. At the 2017 auction’s national average of 0.93 per MHz−POP, the implied value would be approximately $25 million. At the top 40 market average of 1.31 per MHz−POP, the value would be approximately $35 million.

And that is for a single station. Nexstar Media Group, the largest television station group in the United States, owns, operates, and provides services to more than 200 stations across 116 markets, reaching approximately 39% of all U.S. television households.

Subscribe now

What the Carriers Have Actually Paid

The 2017 auction is not the only data point. The secondary market for spectrum licenses — private transactions between wireless carriers, or between carriers and speculators who acquired spectrum at auction — provides an ongoing record of what sophisticated buyers are willing to pay.

In September 2023, T-Mobile entered into an agreement to acquire 600 MHz spectrum licenses from Comcast in a deal valued at up to $3.3 billion. T-Mobile had earlier acquired 600 MHz spectrum from Columbia Capital for $3.5 billion, covering 108 million people — roughly one-third of the U.S. population — in a deal that further validated the pricing established at auction.

Looking further back, the 2008 auction of 700 MHz spectrum — which had been vacated by television broadcasters as part of the digital television transition — saw Verizon pay an average of 1.10 per MHz-POP for its winnings, with AT&T paying 3.15 per MHz-POP for its more selectively targeted licenses. The Chicago license alone sold for $892 million, translating to 9.19 per MHz-POP — a reminder that in the most densely populated markets, spectrum values can reach multiples of the national average.

None of this is reflected on the balance sheets of broadcasters who hold comparable — and in many cases superior — low-band spectrum that propagates further and penetrates buildings better than the mid-band frequencies that have commanded such high prices in recent auctions.

Subscribe now

Case Study: Gray Television — The $2 Billion Question

The story of Gray Television offers perhaps the most vivid illustration of the value arbitrage embedded in broadcast spectrum — and the painful consequences when debt, rather than assets, dictates a company’s fate.

Development: Building an Empire of Airwaves

Gray Television, founded in 1946 by James Harrison Gray as Gray Communications Systems, grew over decades from a single station into one of America’s largest television broadcasters. By 2024, the company owned or operated 180 stations across 113 markets, generating $3.6 billion in annual revenue.

Like its peers, Gray had been an active consolidator during an era of industry roll-up. Each acquisition brought new stations, new markets, and new FCC licenses onto the balance sheet — recorded at whatever portion of the purchase price was allocated to the licenses at the time of each deal. The cumulative carrying value of these licenses on Gray’s books reflected a mixture of acquisition-date allocations stretching back years, bearing no relationship to what comparable spectrum had fetched in the 2017 auction or subsequent secondary-market transactions.

Subscribe now

When the Market Lost Sight of the Assets

By 2023, Gray Television found itself in a difficult position. The company had taken on significant debt to fund its acquisition strategy. As interest rates rose and advertising markets softened — particularly in the off-cycle years between elections, when political advertising spending plummets — the company’s equity value came under severe pressure.

At one point, Gray Television’s market capitalization fell to approximately 400–400–500 million. That figure represented the collective judgment of the stock market about what the entire enterprise — all 180 stations, all their FCC licenses, all their real estate, all their operating businesses, all their retransmission consent agreements, all their network affiliations — was worth.

Yet analysts who took the time to value Gray’s spectrum portfolio on a replacement-cost basis, using the MHz-POP pricing established in the 2017 auction and subsequent private transactions, arrived at a dramatically different number. By some estimates, the spectrum portfolio alone — just the FCC licenses, setting aside all operating assets — was worth in excess of $2 billion.

The company was trading at a 75% to 80% discount to the replacement value of its spectrum. The operating business — $3.6 billion in annual revenue, a portfolio of network affiliations with NBC, CBS, ABC, and FOX, and a growing digital platform — was effectively being assigned a negative value by the market.

Subscribe now

Outcome: The Debt Overhang

The tragedy of this situation, from a value-realization perspective, is that the discount persisted not because the spectrum was worth less than the analysis suggested, but because Gray’s balance sheet was burdened with debt that constrained its strategic options.

By 2025, Gray was actively restructuring its obligations. The company issued $775 million of senior secured first lien notes due 2033 and $900 million of second lien notes due 2032, extending maturities and reducing near-term refinancing risk. These moves bought time but did not, in themselves, unlock the value embedded in the spectrum portfolio.

The critical lesson from Gray Television is that asset value alone is not sufficient. To realize that value, a company needs either the financial flexibility to pursue monetization opportunities at its own pace, or a catalyst that forces the market to recognize what the assets are worth. Gray had the assets; what it lacked was the breathing room to maximize them.

The situation also underscores a fundamental asymmetry in how different constituencies value the same assets. The bond market was focused on Gray’s leverage ratios, interest coverage, and refinancing risk. The equity market was focused on declining linear television viewership, the lumpiness of political advertising revenue, and cord-cutting headwinds. Neither was focused on the simple question: what would it cost to replicate this spectrum portfolio today? The answer to that question — $2 billion or more — was hiding in plain sight, buried in the gap between historical-cost accounting and current market reality.

Subscribe now

The Lesson

A valuable asset encumbered by debt is still valuable — but only when it can unlock its value. The Gray Television experience demonstrates why the business owner’s replacement-cost framework, while essential for understanding what an asset is worth, must be paired with an analysis of “will the undervalued spectrum portfolio remain trap?”.

Subscribe now

Case Study: Sinclair And The Edgebeam Catalyst

If Gray Television illustrates the risk of asset value trapped by balance sheet constraints, Sinclair Broadcast Group illustrates the potential for confirmed catalysts to unlock that value.

Development: The Second-Largest Station Group

Sinclair Broadcast Group, founded in 1986 and headquartered in Hunt Valley, Maryland, operates the second-largest portfolio of television stations in the United States, with 185 full-power stations in 86 markets covering approximately 40% of U.S. television households. Like Gray and Nexstar, Sinclair’s FCC licenses sit on its balance sheet at historical acquisition-accounting values that significantly understate their current market worth.

In 2024, Sinclair delivered record financial results: $405 million in political advertising revenue — $876 million in Adjusted EBITDA, representing a 57% jump from 2023. The company’s operational performance was strong. But the stock market, fixated on the secular challenges facing linear television, continued to price Sinclair as a declining business rather than as the holder of a scarce and increasingly valuable spectrum portfolio.

Subscribe now

The Birth of EdgeBeam Wireless

In early 2025, Sinclair, together with Nexstar, Gray Television, and E.W. Scripps — collectively the four largest television station groups in the United States — announced the formation of EdgeBeam Wireless, a joint venture focused on delivering data services using ATSC 3.0 broadcast signals.

This was not a speculative concept paper or a vague expression of interest. It was a concrete business entity, backed by the combined spectrum resources of the four largest broadcasters in the country, with a clear commercial mandate. EdgeBeam’s proposition is straightforward: use the broadcasters’ existing spectrum infrastructure to deliver data — not television programming, but data — to devices, vehicles, and enterprise customers across the United States.

ATSC 3.0, the next-generation broadcast standard, enables capabilities that go well beyond traditional television. It supports targeted advertising, enterprise datacasting, software updates for connected vehicles, and even a broadcast positioning system that could serve as a backup to GPS. The technology transforms broadcast spectrum from a one-way video delivery mechanism into a two-way data platform with applications that the wireless industry has historically dominated.

Subscribe now

The joint venture structure is significant for several reasons. First, it pools spectrum resources across multiple station groups, creating a nationwide footprint that no single broadcaster could offer on its own — precisely the kind of coverage that wireless carriers and enterprise customers demand. Second, it creates a dedicated vehicle for spectrum monetization that is separate from the broadcasters’ core television operations, allowing investors to value the spectrum business on its own merits. Third, and most importantly, it signals that the broadcasters themselves — the people who understand these assets better than anyone — are actively working to unlock their value. To put it simply:

Imagine the internet is like delivering pizzas:

1. The Wireless Carriers (Verizon, T-Mobile, AT&T)

• What they do: They have a fleet of scooters (Cell Towers) that deliver one pizza to one house at a time.

• The Problem: This is great for regular orders. But imagine if everyone in the city ordered a pizza at the exact same second (like when everyone downloads the same iPhone update or watches the Super Bowl). The roads get jammed, the scooters get stuck, and nobody gets their pizza on time.

2. Sinclair’s Spectrum (Before EdgeBeam)

• What it is: Sinclair owns giant blimps (TV Towers).

• Why Carriers didn’t want it:

  • Wrong Language: The blimps could only drop “TV signals,” not “Internet pizzas.” Carriers couldn’t use them.

  • Too Small: Sinclair only had blimps in some cities. Carriers need a delivery service that covers the whole country, not just random spots.

  • Result: The spectrum was useless to carriers because it didn’t fit their system.

3. Enter EdgeBeam (The Magic Translator)
EdgeBeam changed the rules in two ways:

• The Translator (ATSC 3.0): EdgeBeam taught the blimps to carry “Internet data” instead of just TV shows. Now, the blimps can drop digital files.

• The Team-Up: EdgeBeam got all the other blimp owners (Sinclair, Nexstar, Gray, and Scripps) to join forces. Now, instead of random spots, they have a fleet of blimps covering the entire country.

4. Why Carriers Love It Now
Now, when a carrier needs to send a huge file to millions of people (like a software update for every Tesla), they don’t clog up their scooter fleet.

• They pay EdgeBeam to put the file on the blimps.

• The blimps fly over the city and drop the file to everyone at once.

• The scooters stay free for regular calls and texts.

  Subscribe now

Why is EdgeBeam a Catalyst?

EdgeBeam is the catalyst because it turned a local, one-way TV signal into a national, data-delivery network. It unlocked value by solving the “congestion” problem that 5G towers physically cannot solve cheaply. It proved that old TV airwaves could do modern internet work.

Will Other Spectrum Owners Get Rich?

Likelihood: High for participation, Low for “selling out” alone.

• The “Big Club” Problem: EdgeBeam is already a massive team-up of the four biggest spectrum owners (Sinclair, Nexstar, Gray, Scripps). They control the vast majority of the useful spectrum.

• Small Players: If you are a smaller owner of TV spectrum, you likely won’t be able to start your own competitor to EdgeBeam because carriers won’t hire you for just one city.

• The Opportunity: Instead of selling their license to a carrier (which is legally difficult), smaller owners will likely lease their spectrum to EdgeBeam. They will get paid to let EdgeBeam use their “blimp” to complete the national network.

  Subscribe now

Outcome: A Dual-Path Monetization Strategy

Sinclair’s approach to spectrum monetization has evolved into a dual-path strategy. On one path, EdgeBeam Wireless generates recurring revenue from data delivery services, creating an ongoing income stream from spectrum that was previously used solely for broadcast television. On the other path, the possibility of outright spectrum sales — either through future FCC auctions or through private transactions with wireless carriers — remains a powerful option.

A conservative estimate of Sinclair’s spectrum holdings, based on historical auction pricing, suggests its license portfolio could be worth $1 billion or more.

The EdgeBeam development is particularly noteworthy because it represents a confirmed catalyst — not a speculative possibility, not a regulatory proposal that may or may not materialize, but an actual operating entity with committed resources and a defined business plan. This is precisely the kind of development that can narrow, and eventually eliminate, the gap between the replacement value of spectrum assets and the market’s valuation of the companies that hold them.

Subscribe now

The Lesson

Spectrum value does not need to be realized through a single transaction. EdgeBeam Wireless demonstrates that broadcasters can unlock the value of their spectrum incrementally — through operating businesses that leverage the spectrum’s data-carrying capacity — while preserving the option for more transformative monetization through future sales or auctions. The key insight is that the spectrum is not an inert asset sitting idle on a balance sheet; it is productive infrastructure capable of generating returns that are not captured by traditional television broadcasting metrics.

Subscribe now

The Catalysts

The value arbitrage described in this paper exists because of an accounting framework that systematically understates the economic worth of broadcast spectrum licenses. But value discrepancies can persist indefinitely unless specific catalysts force the market to recognize them. Several such catalysts are now visible on the horizon.

Catalyst 1: ATSC 3.0 and the Data Monetization Paradigm

ATSC 3.0 represents the most significant technological upgrade to broadcast television since the transition from analog to digital. But its implications extend far beyond sharper pictures and better sound. ATSC 3.0 transforms a broadcast tower from a video delivery system into a data distribution platform.

The technology supports IP-based architectures that enable datacasting, personalized content delivery, and advanced metadata that powers dynamic ad insertion. Multi-modal AI can generate rich, scene-level metadata enabling hyper-contextual advertising with measurable increases in effective cost-per-thousand (eCPM) and return on investment. These capabilities allow broadcasters to compete directly with digital platforms for advertising dollars that have migrated away from traditional television.

More significantly, ATSC 3.0 enables entirely new use cases that have nothing to do with television. The standard can support critical infrastructure communications, software updates for autonomous vehicles, public safety paging, and SCADA (Supervisory Control and Data Acquisition) systems for industrial applications. Each of these applications represents a potential revenue stream that leverages existing spectrum assets without requiring additional capital investment in spectrum acquisition.

The EdgeBeam Wireless joint venture — backed by Sinclair, Nexstar, Gray, and Scripps — is the most concrete expression of this catalyst. It is an operational business, not a concept, and its success or failure will be measurable in quarterly revenue numbers within the next several years.

Subscribe now

Catalyst 2: The Private Incentive Auction Proposal

In April 2026, a significant development emerged that could dramatically accelerate spectrum monetization. Landover Wireless proposed a privately financed version of the FCC’s incentive auction — a framework that would be “privately financed and could be completed within approximately 24-36 months” and return more than $15 billion to the U.S. Treasury.

Under this proposal, broadcasters would voluntarily monetize underutilized spectrum in channels 28 through 36 while continuing to deliver their existing television programming. The private financing structure would bypass the cumbersome congressional authorization process that has historically slowed FCC spectrum auctions, potentially compressing the timeline from years to months.

Critically, this proposal targets spectrum that broadcasters are already using less intensively than their primary channels — meaning the monetization would not require broadcasters to cease operations or surrender their core business. It represents a pure value-extraction opportunity from assets that are currently generating little to no incremental return.

Subscribe now

Catalyst 3: The EchoStar Precedent

In 2025, EchoStar — the satellite communications company controlled by Charlie Ergen — entered into definitive agreements to sell spectrum licenses to AT&T for $23 billion and to SpaceX (Starlink) for $17 billion — a combined $40 billion in spectrum transactions. The FCC approved these deals in May 2026.

The EchoStar transactions are significant not merely for their scale but for what they demonstrate about the spectrum market. EchoStar’s licenses were not acquired through the 2017 broadcast incentive auction; they were assembled over years through a combination of auction purchases and secondary-market acquisitions. The prices AT&T and SpaceX were willing to pay — $40 billion in aggregate — reflect the intense and growing demand for spectrum rights from companies building next-generation wireless networks.

For television broadcasters, the EchoStar precedent establishes that spectrum does not need to be sold back to the government through an FCC auction to realize its value. Private transactions with strategic buyers — wireless carriers, satellite operators, technology companies — can achieve comparable or superior pricing, with faster timelines and greater certainty of execution.

Subscribe now

Catalyst 4: Regulatory Tailwinds Under the Current FCC

The regulatory environment for broadcast spectrum monetization has shifted meaningfully. Under Chairman Brendan Carr, the FCC has signaled openness to policies that would facilitate spectrum transactions and reduce barriers to broadcast ownership consolidation.

Recent FCC actions include approval of the $40 billion EchoStar spectrum sales, advancement of ATSC 3.0 rulemaking that could accelerate the transition to next-generation broadcasting, and consideration of changes to national broadcast ownership caps that would allow station groups to achieve greater scale and operational efficiency.

Commissioner Anna Gomez has publicly acknowledged that broadcasters will derive greater benefit from ATSC 3.0 if the agency raises or eliminates the national ownership cap — a recognition that the regulatory framework needs to evolve to accommodate the technology-driven transformation of the broadcast industry.

The direction of regulatory travel is clear: toward greater flexibility for broadcasters to monetize, consolidate, and repurpose their spectrum assets. This does not guarantee that every broadcaster will maximize the value of its spectrum portfolio, but it removes obstacles that have historically prevented value realization.

Subscribe now

Catalyst 5: Industry Consolidation and the Nexstar-TEGNA Transaction

Consolidation within the broadcasting industry serves as both a catalyst for value realization and a direct demonstration of the value that sophisticated buyers place on spectrum assets.

In 2025, Nexstar Media Group — already the largest television station group in the United States — filed applications with the FCC seeking approval to acquire TEGNA Inc. in a transaction valued at approximately $6.2 billion. The acquisition would further expand Nexstar’s already-substantial spectrum portfolio while creating operational efficiencies through scale.

Nexstar’s willingness to pay $6.2 billion for TEGNA — a price that reflects the value of TEGNA’s FCC licenses, its station operations, its network affiliations, and its retransmission consent agreements — provides a real-world benchmark for what a knowledgeable industry buyer believes these assets are worth. Nexstar management, who understand the spectrum market as well as anyone, are effectively voting with their shareholders’ capital that the public market is undervaluing broadcast assets.

Subscribe now

The Lesson: Value Is What Someone Will Pay

The story of broadcast spectrum in America is, at its core, a story about the difference between price and value — between what the accounting rules say something is worth and what a willing buyer would actually pay for it.

The 175 television stations that sold their spectrum in the 2017 incentive auction did not need to convince anyone of this difference. They simply participated in an auction, accepted bids from willing buyers, and collected their proceeds.

The broadcasters who chose not to sell — Nexstar, Gray, Sinclair, TEGNA, Scripps, and others — still hold spectrum that the 2017 auction demonstrated is worth multiples of its carrying value. The question is not whether that value exists; the auction results, the secondary-market transactions, and the EchoStar precedent all confirm that it does. The question is when and how it will be realized.

Several mechanisms for value realization are now in motion. ATSC 3.0 and the EdgeBeam joint venture offer a path to incremental monetization through data delivery services. The Landover private auction proposal offers a path to accelerated spectrum sales without the years-long delays of traditional FCC processes. Industry consolidation — exemplified by Nexstar’s proposed acquisition of TEGNA — offers a path to value realization through strategic transactions at prices that reflect the true worth of the underlying spectrum assets.

The broadcast spectrum that carries local news, network programming, and Sunday football to American households is the same spectrum that carries mobile data to smartphones, connects autonomous vehicles to the cloud, and supports the next generation of wireless applications. The market for that spectrum — as demonstrated by the 2017 auction, the T-Mobile transactions, the C-band auction, and the EchoStar deals — is deep, liquid, and willing to pay prices that bear no relationship to the historical-cost figures on broadcasters’ balance sheets.

The gap will not persist indefinitely. Catalysts are accumulating. The only question is who will own these assets when the gap finally closes.

This white paper is provided for informational and educational purposes only. It does not constitute investment advice, a recommendation, or an offer to buy or sell any security. The author may hold positions in securities discussed. All readers should conduct their own due diligence and consult with qualified financial advisors before making investment decisions. Past auction results and historical transaction data are not guarantees of future pricing. Spectrum values can fluctuate based on regulatory changes, technological developments, and market conditions.

If you like what you’re reading, check out my other writings 👇

A Note on Method Before We Begin

The discipline applied in this paper is identical across every sector where it has been applied before. We are not building a discounted cash flow model. We are not applying an earnings multiple. We are not forecasting commodity prices three years into the future. All of those approaches require you to first forecast the thing you are trying to value — and then apply a multiplier to that forecast — layering compounding uncertainty on top of compounding uncertainty until the output has very little to do with economic reality.

Instead, we do this: identify the primary income-generating asset of the business. Then calculate what a business owner would pay for that asset right now, today, at current prices and current costs, in the same way a farmer walking a plot of land calculates what that land is worth — not based on what rent it might fetch in 2028 under an optimistic scenario, but based on what it yields right now per acre, multiplied by what that crop sells for today, minus what it costs to farm it. That number is the asset’s owner value.

Then we look at what the accountant says the same asset is worth on the balance sheet. We measure the gap. And we explain why the gap exists, why it persists, and what forces will eventually force the market to recognize it.

In phosphate and potash, this gap is unusually large, structurally persistent, and underpinned by a physical reality that is more defensible than almost any other commodity sector: phosphorus cannot be manufactured. Potassium cannot be synthesized. They come from the ground, in specific places, in finite quantities, and without them, crops do not grow. That is not a forecasting assumption. It is biochemistry.

Subscribe now

PART I: THE ASSET, FROM FIRST PRINCIPLES

What Phosphorus and Potassium Actually Are to a Plant

To understand why phosphate and potash reserves are worth what they are — and why the market chronically misprice them — you must start with the biology, not the financial statements.

Every plant that grows on earth requires three primary macronutrients to survive. Nitrogen, which plants can receive from multiple sources including fixation from the atmosphere (and which can be manufactured industrially via the Haber-Bosch process from natural gas and air). Phosphorus, which must come from the soil and cannot be synthesized. And potassium, which must also come from the soil and cannot be synthesized. There is no industrial pathway to manufacture either phosphorus or potassium. They must be mined. This is the foundational fact upon which everything else in this paper rests.

Phosphorus in the soil exists primarily as phosphate ions. When a farmer applies diammonium phosphate (DAP) fertilizer — the world’s most widely traded phosphate product — they are delivering concentrated phosphate ions directly to the root zone of the crop, accelerating the plant’s ability to develop roots, transfer energy, and form seeds. A cereal crop like wheat that does not receive adequate phosphate will produce 20–40% lower yields. There is no substitute input. You cannot give wheat more nitrogen to compensate for phosphorus deficiency. The nutrient is specific, the deficiency is specific, and the remedy is specific.

Potassium — applied primarily as potassium chloride (KCl), which is the end product of potash mining — governs the plant’s water regulation, disease resistance, stem strength, and grain filling. A crop deficient in potassium produces weaker stems, higher disease susceptibility, and reduced grain weight. Again: there is no substitute.

This means the demand for phosphate and potash is not a function of price — it is a function of whether you want food. Farmers do not stop applying fertilizers when prices are high; they cut back only at the margin, and even then, the soil depletion that results creates a payback demand in subsequent years. Global food consumption grows at roughly the rate of population growth — 1% per year — compounded by the shift of developing populations toward meat-intensive diets, which require 5–8x more grain per calorie than direct grain consumption. The demand floor for phosphate and potash is essentially equivalent to the demand floor for food. There is no technology that changes this.

Subscribe now

Where Phosphate and Potash Come From — And Why the Geography Matters

Phosphate rock globally is extraordinarily concentrated. Morocco’s OCP controls approximately 70% of the world’s proven phosphate rock reserves — an estimated 50 billion tonnes, per the US Geological Survey’s 2024 Mineral Commodity Summaries. Morocco’s deposits in the Khouribga, Youssoufia, and Ben Guerir regions have ore concentrations averaging 30–35% P₂O₅ in mineable deposits, at production costs that are structurally lower than any other major supplier. In North America, Florida’s Bone Valley region accounts for approximately 80% of domestic US phosphate capacity — and there is no comparable domestic alternative. The Florida deposits are not merely large; they sit adjacent to existing processing infrastructure, port facilities, and the transportation networks that serve North American agriculture. Building an equivalent complex from scratch would take 10–15 years and billions of dollars, and you still could not move the Florida deposits’ geographic advantage relative to US agricultural heartland.

Potash is similarly concentrated. The Devonian evaporite deposits of Saskatchewan, Canada — the ore bodies that underlie Nutrien’s and Mosaic’s mines — contain the world’s largest and highest-quality potash reserves. Saskatchewan’s Esterhazy mine alone, operated by Mosaic, is the largest single potash mine in the world by reserve base and production capacity. The geological conditions that created Saskatchewan’s potash — ancient inland seas evaporating over millions of years under specific temperature and pressure conditions, leaving behind thick beds of sylvinite ore containing 20–30% potassium chloride — are not replicated elsewhere with the same combination of grade, depth, and accessibility.

Belarus and Russia hold the other major global potash reserves. Belaruskali in Salihorsk and Uralkali in Perm together control substantial portions of global production capacity. But Russian and Belarusian supply has been progressively removed from Western markets since 2022: EU and US sanctions on Belarus following the Lukashenko government’s post-election crackdown and Russia’s subsequent invasion of Ukraine effectively constrained the volume of Eastern European potash available to Western agricultural markets. This geopolitical supply disruption is not temporary. It is structural, and it has permanently altered the sourcing calculus for North American and European farmers.

Subscribe now

PART II: THE OWNER’S VALUATION — FOUR NUMBERS AND AN HONEST CALCULATION

An owner assessing a phosphate or potash mining company does not need a model. They need three inputs.

Input 1: Certified tonnes of reserve in the ground. Under GAAP (SEC Industry Guide 7 for US companies) and international equivalents (JORC, NI 43-101), proven and probable mineral reserves are certified by independent qualified persons and published in annual reports. These are not estimates or aspirations — they are statistically confident calculations of extractable ore based on drill hole data, geological mapping, and mine engineering. They represent what the owner legally owns below the surface.

Input 2: Average historical market price per tonne of finished product (including lowest quartile cost). Historically, market prices for finished products are benchmarked against liquid, publicly quoted indices. For phosphate, the historical average is based on Diammonium Phosphate (DAP) at US Gulf FOB; as of 2024, prices traded between $400–$450 per tonne, a significant normalization from the 2022 peak of approximately $900 per tonne. Similarly, potash prices—benchmarked as Potassium Chloride (MOP) at Vancouver or Tampa—stabilized in the $280–$320 per tonne range in 2024, down from historical highs of $950 per tonne in 2022. These benchmarks are tracked weekly by institutions like the World Bank to provide a consistent record of long-term market trends.

Input 3: Cash operating cost per tonne of finished product. This is the actual cost of mining the ore, processing it to finished fertilizer, and delivering it to port. For Mosaic’s Florida phosphate operations, their disclosed cost of production for finished phosphate fertilizer is approximately $280–$320 per tonne all-in (production cost plus processing plus transportation to port), per segment disclosures in their 10-K filings. The resulting gross margin at $420/tonne DAP is approximately $100–$140 per tonne. For Saskatchewan potash, Mosaic’s Esterhazy mine has a disclosed cash operating cost of approximately $80–$120 per tonne KCl — among the lowest in the world given its scale and ore quality. At $300/tonne MOP, the margin per tonne is approximately $180–$220. These margins are not theoretical — they are what the mine actually earned in recent operating periods.

Subscribe now

The Owner’s Calculation:

An owner buying this business is not buying a forecast. They are buying assets that exist today, are currently producing, are currently generating measurable margins at current prices, and cannot be replicated by a competitor at any reasonable cost or timeline. The book value says $5.4 billion. The owner’s value — calculated simply as what these assets are capable of generating in profits on average for over a 10 year period — says something fundamentally different, and the gap between the two numbers is the arbitrage.

Subscribe now

PART III: THE WALL STREET PROBLEM — AND WHY EARNINGS METRICS FAIL HERE

Commodity Cycles Are Not Business Cycles

When a Wall Street analyst covers Mosaic, the model invariably centers on earnings forecasts. What will DAP prices be in 2025? What will potash volumes do? What is the consensus on nitrogen? What multiple does the sector deserve? These questions are not unreasonable in isolation — they are the standard toolkit for industrial companies. But they are particularly ill-suited to a business whose primary asset is a mineral reserve, because of one characteristic that almost every financial model ignores: the earnings collapse during price troughs dramatically understates asset value, and the earnings spike during price peaks dramatically overstates it.

Consider what happens to a fertilizer company’s reported earnings when commodity prices fall 40–50% from peak. The ore body is unchanged. The tonnes in the ground are unchanged. The reserves certified by the geological engineers are unchanged. The infrastructure — the chemical plants, the port facilities, the rail access — are unchanged. But the reported P/E ratio, applied to depressed earnings, suggests that the company is expensive or unprofitable, which triggers selling from investors who anchor on the earnings metric. The asset owner, who looks at the same company and says “the mine still has 40 years of reserve life at current production rates, and it still costs me $100/tonne to produce something I can sell for $350/tonne — that is a $250/tonne margin on 8 million tonnes — this asset is worth approximately X,” sees a completely different business than the analyst running earnings models.

The inverse is equally misleading. When DAP prices spiked to $900/tonne in 2022, phosphate companies reported extraordinary earnings. P/E multiples contracted because earnings expanded so rapidly that the stock price could not keep pace. Investors who bought on the basis of “cheap P/E” during a commodity price spike were buying at the peak of the earnings cycle — the worst possible time — because the earnings themselves were not sustainable at those price levels. The ore body, again, had not changed. The mine had not become more valuable. The market price had moved temporarily, and the earnings followed it, and the financial metrics followed the earnings.

The correct instrument for measuring a mining company’s intrinsic value is not the earnings multiple. It is the ore body. How many tonnes are in the ground? What does it cost to extract them? What does the market pay for them? That arithmetic, at mid-cycle prices, is the business owner’s answer — and it is immune to the short-term commodity volatility that makes financial metrics so misleading in this sector.

Subscribe now

PART IV: THE ACCOUNTING GAP — WHY THE BALANCE SHEET TELLS YOU ALMOST NOTHING

What GAAP Says About Mineral Reserves

Under US GAAP (ASC 360 for PP&E, with specific guidance for extractive industries under ASC 932 and SEC Industry Guide 7), mineral reserves are not recognized as assets on the balance sheet. The underlying ore body — the billions of tonnes of phosphate rock or potash ore certified by independent geological engineers — carries a balance sheet value of exactly zero. What appears instead is the cost of acquiring the mineral rights and developing the mine, capitalized as Property, Plant, and Equipment, and then depleted against production using the Units of Production (UOP) method.

This creates a systematic and predictable gap. The depleted book value of a mine has three characteristics that distinguish it from economic value: it reflects what someone paid in the past, not what the ore is worth today; it declines as production continues (depletion reduces the carrying value), even as the mine may generate increasing value if commodity prices rise; and it bears no relationship to the replacement cost of an equivalent ore body, which would require 10–15 years of exploration and development capital.

Mosaic’s Florida phosphate operations were assembled primarily through the 2004 merger of IMC Global and Cargill’s crop nutrition business, at prices reflecting early-2000s commodity economics and early-2000s capital costs. The Florida phosphate ore body in 2004 was worth what it was worth in 2004 — DAP prices at the time were approximately $150–$170 per tonne. Today, DAP prices are $400–$450 per tonne even after the 2022 spike correction. The ore body produces the same tonnes at the same cost. The market pays more than twice as much for the finished product. Under GAAP, the carrying value of the asset has declined — because depletion systematically reduces the historical cost figure — while its economic value has roughly doubled.

This is the central absurdity. GAAP requires the balance sheet to show an asset that is worth less every year — because depletion charges reduce the historical cost figure — even as the asset’s economic value may be growing, because the commodity it produces is selling for more money. The accountant says: “This mine is worth less than it was five years ago, because we have depleted part of the historical cost.” The owner says: “This mine is worth more than it was five years ago, because DAP prices are higher, we still have 35 years of reserve life, and reserves barely decreased” Both statements are consistent with their respective measurement frameworks. Only one of them measures the actual economic value.

Subscribe now

PART V: THREE CASE STUDIES IN PHOSPHATE AND POTASH VALUE ARBITRAGE

Case Study 1: The BHP Bid That Revealed Everything (2010)

In the summer of 2010, BHP Billiton’s $38.6 billion unsolicited bid for PotashCorp—at $130 per share—forced a stark realization upon the market. At the time, the industry was valuing PotashCorp using Discounted Cash Flow (DCF) models heavily weighted toward the prevailing spot prices. With potash trading near historic highs, these models produced valuations that mirrored the peak of the cycle. To the average analyst, PotashCorp was “fairly valued” because the math worked—as long as $800/tonne potash was the input.

The fundamental risk of this approach is that a DCF model is only as good as its price assumptions. In a cyclical commodity market, relying on current prices means:

• During Booms: Valuations are inflated, making acquisitions seem prohibitively expensive or justifying “buying at the top.”

• During Slumps: Valuations collapse, making world-class assets appear “broken,” leading investors to stay on the sidelines when they should be buying.

The Nutrien Reality Check

The flaw in spot-price DCF modeling is most evident when comparing the 2010 bid to the 2018 formation of Nutrien. By 2018, the industry was in an “ongoing slump,” with prices plummeting below $300/tonne. Consequently, the market’s DCF-driven valuation for the Agrium-PotashCorp merger was approximately $36 billion.

Think about the math: In 2010, the market accepted a $38.6 billion valuation for PotashCorp alone. Eight years later, the market valued that same asset base—plus Agrium’s massive retail network—at $2.6 billion less.

The geology hadn’t changed. The reserves hadn’t moved. Only the price input in the spreadsheets had shifted.

Some context on price: The primary reason the Nutrien merger valuation ($36 billion in 2018) appeared lower than BHP’s historical bid ($38.6–$43 billion in 2010) is a fundamental shift in the potash market and the differing nature of the two deals — BHP's bid occurred when potash prices were significantly higher, fueled by post-recession optimism and high global demand in 2010. By the time Agrium and PotashCorp merged, the industry was in an "ongoing slump". Over-supply and the breakdown of major potash cartels (like the 2013 Uralkali-Belaruskali split) had caused prices to plummet from over $800/tonne to below $300/tonne

Missing the Opportunity

If you value a company solely through a DCF lens tied to the current curve, you will almost always miss the biggest opportunities. A DCF makes an asset look least attractive exactly when it is most valuable to own: at the bottom of the cycle.

BHP’s bid represented a different logic—one that recognized a “floor” on intrinsic value. While the industry’s DCF models swung wildly between 2010 and 2018 based on the whims of potash cartels and global supply, the actual utility and irreplaceability of the Saskatchewan ore bodies remained constant. To capture true value in this sector, one must average out prices over the long term. Valuing an asset based on its mid-cycle earning power provides a tether to reality that a spot-price DCF simply cannot offer.

Subscribe now

Case Study 2: The Cartel Crack and the $900 to $300 Collapse (2008–2015)

The potash market from roughly 2005 to 2013 operated under what was, in practical terms, a duopoly. On one side: Canpotex, the North American export marketing company jointly owned by PotashCorp, Mosaic, and Agrium, which collectively controlled roughly 35% of global potash supply. On the other side: the Belarusian Potash Company (BPC), which coordinated the export marketing of Belaruskali of Belarus and Uralkali of Russia, controlling another 35–40% of global supply. Together, these two marketing consortiums controlled approximately 70% of traded potash, and they managed that supply with the explicit goal of maintaining high prices by restricting output — a structure the financial press and economists correctly identified as a de facto cartel, referred to at the time as “the OPEC of potash.”

Market watchers and investors were broadly aware of this structure. The celebrated value investor Jeremy Grantham, in his July 2012 quarterly letter, warned explicitly about the potential for duopolistic price gouging in potash markets and the food security implications. Investment banks wrote reports about how Canpotex and BPC would sustain prices above $400/tonne indefinitely by coordinating their production volumes. The assumption was structural, embedded in models, and widely believed.

In 2008, driven partly by the food crisis and partly by cartel discipline, potash prices reached approximately $875–$1,000 per metric tonne — up from roughly $100/tonne in 2004. A 10x price increase in four years. The financial models extrapolated this trend; the earnings models for potash companies looked spectacular; and the market priced potash miners as if $500–$700/tonne potash was the new permanent equilibrium. Mosaic’s stock price reached approximately $160 per share in 2008.

On July 30, 2013, Uralkali — one of the world’s largest potash producers — announced that it was unilaterally withdrawing from the Belarusian Potash Company marketing consortium. The reason was a dispute with its Belarusian partners over accusations that Belaruskali had been selling volumes outside the agreed cartel framework. Regardless of the cause, the effect was immediate and dramatic: the informal pricing discipline that had held potash prices above $400/tonne for years evaporated overnight. Uralkali’s CEO, Vladislav Baumgertner, stated publicly that the company’s strategy was now to “increase volumes” — the explicit declaration that the cartel was over and price competition would follow. Potash prices fell from approximately $400/tonne to $300/tonne within months. By 2016, prices had reached approximately $200/tonne.

The financial models that had built $400–$500/tonne potash into their assumptions were catastrophically wrong. Mosaic’s earnings per share collapsed. Earnings multiples that had looked attractive suddenly became unattractive as earnings fell more than prices, due to operating leverage. The stock fell from its 2008 peak of $160 to approximately $20–$25 per share by 2015–2016. Analysts reduced price targets. Institutional investors sold. The market consensus shifted from “potash is a strategic asset with pricing power” to “potash is a cyclical commodity with competitive supply.”

The collapse in potash prices revealed something important about the distinction between earnings-based valuation and owner’s valuation. At $200/tonne potash and Mosaic’s cash operating cost of $80–$110/tonne for Esterhazy production, the mine was still generating a net margin of $90–$120/tonne. It was less than the $300–$400/tonne margin at peak prices, but it was real, positive, and ongoing. Esterhazy’s ore body had not changed. The certified reserves were unchanged. The mine’s physical capacity was unchanged. The financial models said the business was worth $20–$25 per share; the ore body said the asset was worth considerably more.

The confirmation came in 2021–2022, when Russia’s invasion of Ukraine triggered sanctions on Belaruskali — which had been the supply that broke the cartel — and simultaneously tightened Russian potash export capacity. With Belarusian and Russian supply constrained by sanctions, potash prices returned to approximately $950/tonne by April 2022. The same Esterhazy ore body, the same Saskatchewan geology, the same mine infrastructure — now worth dramatically more. Mosaic’s stock price recovered from its 2016 lows to above $65 per share in 2022. The owner who had looked at the ore body in 2016, calculated the margin at $200/tonne potash and $100/tonne cost, and understood that the asset would still be there when prices recovered — that investor earned a 200–250% return while the financial models told everyone else to sell.

The cartel collapse taught the market to price potash as a competitive commodity. The owner’s framework always treated it as something different: a certified geological reserve in an irreplaceable location, generating positive cash margins at any potash price above $100/tonne (Esterhazy’s approximate floor cash cost), with decades of reserve life remaining. The earnings collapsed. The asset did not. Confusing the earnings with the asset is the fundamental error that cyclical commodity investing requires you to avoid. The owner never confuses them.

Subscribe now

Case Study 3: Intrepid Potash and the Market That Forgot the Mine (2022–2024)

Intrepid Potash, Inc. (NYSE: IPI) is the only US-domiciled potash producer of any meaningful scale. It operates two conventional potash mines in New Mexico — the IPNM facility near Carlsbad, which uses solution mining (dissolving underground potash deposits and recovering the brine) and the Moab facility in Utah — plus a langbeinite operation producing a specialty potash product called langbeinite (K-Mag), which contains potassium, magnesium, and sulfur in a single mineral and has no commercial equivalent produced by any other company in North America.

This last fact is important enough to repeat: Intrepid Potash is the only commercial producer of langbeinite in the world. Langbeinite is particularly valuable for specialty agriculture — wine grapes, tree nuts, and organic crops where chloride sensitivity makes standard potassium chloride unsuitable. The premium for langbeinite versus standard potash historically ranges from 30–50%. There is no substitute product from any other supplier.

Per its SEC filings, Intrepid Potash’s New Mexico mineral reserves as of December 31, 2022 included: 5.4 million metric tonnes of proven potash reserves plus 4.6 million metric tonnes of proven langbeinite reserves at the IPNM facility, with net book values of the New Mexico property at approximately $172.3 million, Moab at $62.6 million, and Wendover at $41.0 million — a combined book value of approximately $275.9 million across all mineral properties.

When potash prices were at $950/tonne in early 2022, Intrepid’s revenue per tonne was approximately $700–$800 (blended across products). The company’s market capitalization at the 2022 peak was approximately $1.2 billion. Applying the owner’s arithmetic to the potash reserve: 4.5 million product tonnes of potash plus 4.6 million product tonnes of langbeinite, at margins of approximately $400–$500/tonne at peak prices, implied an extraordinary embedded operating income.

By late 2023 and into 2024, potash prices had corrected from their $950/tonne peak to approximately $280–$310/tonne standard MOP. Langbeinite prices also softened, though less severely given their specialty nature and Intrepid’s monopoly position. Intrepid’s quarterly revenue fell from peak levels. Reported earnings compressed. The stock fell from its 2022 peak of approximately $75 per share to approximately $18–$22 per share by 2023–2024.

At $22 per share and approximately 19 million diluted shares, the market capitalization was approximately $418 million. Set against this: the New Mexico mineral properties on the balance sheet at $172.3 million book value (conventional GAAP, historical cost, depleted). The owner’s arithmetic at trough prices: 4.5 million product tonnes of potash at $290/tonne current price, less approximately $160/tonne cash operating cost (per company disclosures), equals a $130/tonne margin on potash alone. 4.6 million product tonnes of langbeinite at $350/tonne price (approximate specialty premium), less $160/tonne operating cost, equals a $190/tonne margin. Total gross operating income embedded in the New Mexico reserves, at current trough prices: approximately $585 million from potash and $874 million from langbeinite — a combined $1.46 billion in embedded operating income from certified reserves, against a market capitalization of $418 million. The market capitalization was less than 30% of the embedded operating income of the mine at trough prices. The balance sheet said $275 million in mineral properties. The owner’s arithmetic said approximately $1.5 billion at conservative current pricing.

The specialty premium in langbeinite is particularly striking. Because Intrepid is the only producer, it has no comparable on the supply side. The market cannot replace Intrepid’s langbeinite with anything from any other company. The mine, the ore body, and the US specialty agriculture customer relationships Intrepid has built over decades are a bundled asset with genuine monopoly characteristics. This asset is carried on the balance sheet at historical development cost. The balance sheet treats it the same as any other mining property. It is categorically not the same as any other mining property.

The case remains in progress as of mid-2026, which makes it particularly instructive. The ore bodies have not been impaired. The langbeinite monopoly has not been challenged. Intrepid’s cash operating costs have not materially changed. The only thing that changed was the commodity price cycle. An owner looking at this situation sees a company priced at less than one-third of the embedded operating income of its certified reserves, at trough prices, with a genuinely irreplaceable specialty product that no competitor can supply. The thesis is the same thesis the owner always applies: the asset exists, the cash flows from it are real and positive, and the market is pricing the stock as if the business is impaired when the business is merely operating at a trough price point in a cycle that has historically turned.

The langbeinite monopoly in Intrepid’s case is the clearest possible illustration of the gap between accounting treatment and economic reality. GAAP says: this is a mining property at historical cost, depleted over time. The owner says: this is the only commercial langbeinite mine on earth, serving a premium specialty market with no alternative supplier, and the certified ore body runs for decades. Those are not the same asset. GAAP cannot capture what “only” means in a competitive market context.

Subscribe now

PART VI: PURE-PLAY CASE STUDIES — ICL GROUP AND THE DEAD SEA

ICL Group: When the Mine Is a Geologic Miracle

ICL Group (NYSE: ICL) operates its primary potash production from what may be the most operationally unusual mining asset in the world: Dead Sea Works, located on the southern shores of the Dead Sea in Israel. This is not a conventional underground mine. It is a solar evaporation operation. ICL pumps brine from the northern basin of the Dead Sea into a series of enormous, shallow evaporation ponds — 150 km² in total — where the Israeli desert sun does the work of concentrating the minerals. Carnallite (a potassium-magnesium mineral) precipitates from the concentrated brine and is harvested, processed, and refined into potassium chloride fertilizer at approximately 62% K₂O purity — among the highest concentrations in global potash production.

The economics of this process are unlike any underground mine. There is no conventional mining — no drilling, no blasting, no shaft sinking, no rock handling. The primary energy input is solar radiation, which is free. The primary infrastructure input is the evaporation ponds themselves, which once constructed have minimal maintenance requirements and effectively unlimited operational life as long as the Dead Sea itself continues to exist. ICL’s Dead Sea Works has been operating continuously since 1931 — nearly a century — and its concession from the Israeli government runs through 2030, with anticipated renewal.

Annual potash production capacity at Dead Sea Works is approximately 4.0 million metric tonnes, contributing to ICL’s position as the world’s sixth-largest potash producer. ICL also operates underground potash mines in Spain (ICL Iberia, with two mines near Barcelona producing approximately 1.5 million tonnes per year) and phosphate mining operations in Israel’s Negev Desert through ICL Rotem, which holds a long-term government concession for phosphate extraction.

The Accounting and Owner’s Gap for ICL

ICL’s evaporation ponds at Dead Sea Works are capitalized on the balance sheet at historical construction cost — the cost of building the pond infrastructure, roads, pumping stations, and processing facilities — minus accumulated depreciation over decades of operation. The older sections of the Dead Sea Works evaporation infrastructure, built in the 1930s–1950s, are almost entirely depreciated. Their net book value approaches zero.

Yet they produce 4 million tonnes of potash per year. At a current MOP price of approximately $300/tonne and an all-in operating cost that ICL has historically disclosed to be among the lowest in the industry — the solar evaporation process and the proximity of the brine source to the processing facility create a structural cost advantage — ICL’s Dead Sea potash segment operates at margins that are largely invariant to the capital cost of the evaporation infrastructure, because that infrastructure is already built and already paid for.

The owner’s calculation: 4 million tonnes per year, at approximately $180–$200/tonne margin (reflecting the cost-competitive solar evaporation process versus conventional underground mining), generates approximately $720–$800 million per year in gross operating income from the Dead Sea Works potash operation alone — from infrastructure with a near-zero net book value on the balance sheet. ICL’s total enterprise value as of 2023–2024 ranged from approximately $7–$9 billion. The Dead Sea Works potash operation, on a standalone owner’s metric, was generating the equivalent of 8–11% of the total enterprise value in annual operating income from near-zero-book-value assets.

ICL also disclosed a meaningful sensitivity: every $20 change in average potash CIF price produces approximately $100 million of annual EBITDA impact (disclosed in their Q4 2023 earnings release). This linear price-to-EBITDA relationship confirms the owner’s arithmetic in another form: the leverage to commodity prices is direct and predictable, because the operating cost structure of the Dead Sea is largely fixed and the marginal revenue from each additional tonne flows nearly entirely to the operating line.

Subscribe now

PART VII: THE PHOSPHORUS SCARCITY ARGUMENT — THE LONG-TERM FLOOR NO MODEL CAPTURES

There is a layer of value in phosphate reserves that no financial model adequately captures because it operates on a timescale that exceeds any analyst’s forecast horizon. It is worth stating clearly because it is the single most important structural fact about phosphate as a resource.

Phosphorus is not recyclable in the conventional sense. It leaves the farm in grain, travels through the food system, is consumed by humans and animals, and ultimately ends up in wastewater, rivers, and oceans — where it is effectively inaccessible. The US Geological Survey estimates that at current rates of global consumption, identified world phosphate rock reserves would last approximately 300 years. This sounds comfortable until you account for the concentration of those reserves: Morocco holds approximately 70%, and the second-largest holder is also an unstable jurisdiction from a Western supply-chain security perspective. Accessible, affordable, Western-controlled phosphate reserves are significantly more limited.

More importantly, there is no substitute. You cannot replace phosphorus in plant nutrition with any other element. You cannot synthesize it. You cannot extract it from the atmosphere. Every tonne of phosphate rock extracted from the ground and converted to fertilizer is a tonne permanently removed from the extractable reserve base.

Unlike potash, which at very high prices might eventually be substituted to a limited degree by different potassium sources, phosphorus has no substitute pathway at any price.

The Florida phosphate district — which accounts for 80% of US domestic phosphate capacity and roughly 15% of global phosphate rock production — is an irreplaceable national resource. Mosaic’s Florida reserves are not merely a balance sheet asset. They are a strategic agricultural infrastructure asset with no domestic alternative. The US cannot replace Florida phosphate rock with imports from Morocco at stable prices, because a shock to Moroccan supply (political instability, export restrictions, or OCP’s market pricing decisions) would immediately affect every American farmer who applies phosphate fertilizer to their fields.

The countervailing duty that the US Department of Commerce has imposed on Moroccan phosphate fertilizer imports — reached 16.81% as of November 2024 — is itself evidence of this dynamic. The US government has explicitly chosen to protect domestic phosphate production capacity precisely because the strategic value of owning the only significant domestic phosphate reserves exceeds the short-term consumer cost savings from importing cheaper Moroccan product.

Subscribe now

PART VIII: CONFIRMED CATALYSTS — EVENTS ALREADY IN MOTION

A catalyst, in this context, is not a forecast or a hope. It is a confirmed event that is already in motion and that structurally tightens either the supply side or the demand side of the phosphate and potash market — increasing the net margin per tonne that flows to the owners of certified reserves.

Subscribe now

Catalyst 1: Belarusian and Russian Supply Constrained by Sanctions — Structural, Not Cyclical

Belarus is the world’s third-largest potash producer by historical capacity, with Belaruskali controlling roughly 15–17% of global production capacity. Russia’s Uralkali and EuroChem collectively control another 20%+ of global capacity. Since the onset of European and US sanctions on Belarus following 2021’s disputed election crackdown, and since expanded sanctions on Russia following the February 2022 invasion of Ukraine, the ability of Belarusian and Russian potash to reach Western markets has been structurally constrained.

This is not a short-term disruption. Sanctions on Belarus are embedded in the political positions of the EU, UK, and US. The Lukashenko government shows no signs of changing its political alignment or behavior that might lead to sanctions relief. Russian sanctions are contingent on a geopolitical resolution to the Ukraine conflict that is unpredictable in timing and outcome. For Western agricultural markets — Europe and North America — the effective removal of Belarusian and Russian supply from the accessible supply pool is a permanent structural shift until geopolitical conditions change, and geopolitical conditions are moving slowly in the wrong direction for Minsk and Moscow.

The beneficiaries are the companies holding Western-controlled, sanctions-immune potash reserves: Nutrien in Saskatchewan, Mosaic at Esterhazy, and ICL Group at Dead Sea and Spain. Every tonne of potash demand that previously came from Belarusian or Russian supply and now must come from Western sources directly increases the utilization rate and pricing power of Esterhazy, Dead Sea Works, and Iberia mines. This is not a forecast. It is current market reality.

Subscribe now

Catalyst 2: China’s Phosphate Export Restrictions — Tightening Global Supply

China is the world’s largest producer of phosphate fertilizers by volume, historically exporting approximately 9–10 million metric tonnes of ammoniated phosphates per year (DAP and MAP). Since spring 2021, China has intermittently restricted phosphate fertilizer exports, citing domestic food security concerns. In 2024, Chinese phosphate exports declined to approximately 6.6 million metric tonnes, per IFPRI data — a 34% reduction from the 2021 volume. These restrictions have been implemented unilaterally and without advance notice to global markets.

The second driver of Chinese export restriction is less discussed but structurally more important for the long term: the rapid growth of domestic Chinese demand for phosphate rock as a feedstock for lithium iron phosphate (LFP) battery cathode materials for electric vehicles. LFP batteries require high-purity iron phosphate as a key input, and China — which manufactures the vast majority of the world’s LFP batteries — is consuming increasing quantities of domestic phosphate rock to feed this industrial demand. Domestic industrial demand for phosphate competes directly with export supply. As Chinese EV production scales up, the domestic consumption of phosphate rock for non-agricultural purposes will increase, further constraining the export volume available for global agricultural markets.

This catalyst is already active. It is also self-reinforcing: as Chinese exports decline, global phosphate prices rise, which makes Mosaic’s Florida operations and ICL Rotem’s Negev Desert operations more valuable, because they are producing into a tighter market. The reserves in the ground are worth more per tonne when global supply is constrained and their marginal revenue increases.

Subscribe now

Catalyst 3: Growing Population, Dietary Transition, and the Iron Demand Floor

Global population is expected to reach approximately 9.7 billion by 2050, per the UN World Population Prospects. The caloric demand associated with feeding this population, particularly as Asian and African living standards rise and diets shift toward higher protein consumption (which requires 5–8x the agricultural land per calorie compared to grain-based diets), translates directly into demand for fertilizer.

The International Fertilizer Association (IFA) in its November 2024 report noted a downward revision of near-term phosphate demand growth to account for affordability constraints — some farmers in price-sensitive developing markets cut application rates when prices are high. But the same report acknowledged that structural underuse of fertilizers in sub-Saharan Africa, Southeast Asia, and South Asia means that long-term demand growth is not constrained by soil saturation — it is constrained only by affordability, which is a temporary cycle, not a structural ceiling.

When affordability recovers — which it does at every price cycle — deferred fertilizer applications come back as catch-up demand, often concentrating in a single agricultural season. The farmers who skipped a phosphate application in 2023 due to high prices will have measurably lower soil phosphorus levels in 2024–2025, requiring larger-than-normal applications to restore productivity. This deferred demand effect is not captured in any quarterly earnings model. It is the background physics of soil chemistry, operating on a longer timescale than any analyst’s price forecast.

Subscribe now

Catalyst 4: LFP Battery Demand Creating a New, Non-Agricultural Market for Phosphate

ICL Group has been pursuing a strategic pivot that, if successful, would create a second demand base for its phosphate chemical assets that is entirely disconnected from agricultural markets: lithium iron phosphate (LFP) cathode active materials for electric vehicle batteries.

LFP batteries are structurally simpler, cheaper, and more thermally stable than nickel-manganese-cobalt (NMC) batteries, making them preferred for lower-cost EVs and stationary energy storage. Their one disadvantage — lower energy density — is acceptable for short-range vehicles and fixed storage applications, and ongoing cell chemistry improvements are reducing this gap. LFP battery demand is growing rapidly, and LFP requires high-purity iron phosphate as a core input.

ICL had begun constructing a facility in St. Louis, Missouri to produce LFP cathode active materials — having received a US Department of Energy loan guarantee commitment for approximately $380 million. The project was cancelled in October 2025 after the Trump administration rescinded the federal funding commitment as part of broader cuts to green energy subsidies. A parallel European JV with Chinese partner Dynanonic was also restructured.

The LFP pivot represents a technology catalyst with uncertain near-term execution risk but confirmed long-term structural logic: a company that mines phosphate and has spent decades developing high-purity phosphate chemistry has a natural advantage in producing the downstream phosphate chemicals required for battery manufacturing. The value of ICL’s phosphate mining infrastructure is higher if it can serve both agricultural and industrial demand than if it serves only agriculture. The exact timing of this realization depends on battery market economics, policy certainty, and execution — which is why this catalyst carries more conditionality than the geopolitical supply catalysts above.

Subscribe now

CONCLUSION: THE ARITHMETIC OF IRREPLACEABILITY

The core insight of this paper is simple enough to state in two sentences.

Phosphorus and potassium must be mined. The companies that own the best mines in the best locations — Florida, Saskatchewan, the Dead Sea, the Negev Desert — own assets that are certified in tonnes by geological engineers, that produce measurable margins at today’s prices, that have decades of remaining reserve life, and that cannot be replicated by any competitor at any capital budget within any commercially relevant timeframe.

GAAP records these assets at what someone paid for them 20 years ago, minus accumulated depletion. The assets have not declined in economic value since then. In most cases, they have increased in value, because the commodity prices they produce have risen and the scarcity of equivalent replacement assets has increased. The gap between the balance sheet’s answer and the owner’s answer is not an artifact of market irrationality. It is a systematic consequence of an accounting standard designed to track historical cost, applied to an asset whose value is determined by current geology and current commodity economics — neither of which has any relationship to what someone paid in 2004.

The earnings will fluctuate. They always do in commodity businesses. The ore body will not. And the ore body — irreplaceable, certified, producing positive margins at today’s prices — is what you are actually buying.

This white paper is for educational and informational purposes only. All reserve figures, cost data, transaction prices, and market statistics are drawn from publicly available company annual reports (10-Ks, 20-Fs), SEC filings, regulatory submissions, USGS mineral commodity summaries, World Bank commodity data, and industry sources as cited throughout. Nothing herein constitutes investment advice.

If you like this read, check out my other writings 👇

Imagine you own a bakery. Every morning, your baker arrives at 4am, mixes flour, water, and yeast, and four hours later you are selling croissants for $6 apiece. Now imagine your accountant tells you that your bakery’s most important asset — the baker’s twenty-year mastery of laminating dough — is worth exactly zero on your balance sheet. And that the recipe book he spent fifteen years perfecting, the one that makes your croissants different from every other bakery on the block, is also worth zero. The oven and the marble countertop, on the other hand, are listed as assets because you can touch them.

This is not a hypothetical. This is, almost exactly, what happens to every pharmaceutical and biotechnology company operating under US Generally Accepted Accounting Principles (GAAP).

The drug a company spent a decade and $800 million developing? Worth zero on the balance sheet — the entire cost was expensed, year by year, as it was incurred. The late-stage cancer drug sitting in a Phase III clinical trial, where preliminary data shows a 47% tumor response rate in patients who previously had zero options? Zero. The Breakthrough Therapy Designation the FDA handed out — its formal signal that the drug shows substantial improvement over anything that currently exists? Not an asset. Not even a footnote in the asset section.

What is on the balance sheet: the laboratory equipment, the office furniture, some cash, and whatever intellectual property the company acquired when it bought another firm.

The gap between what a building’s balance sheet says and what the building is actually worth has created some of the most extraordinary investment opportunities in the history of public markets. But to find them, you have to stop thinking like an accountant and start thinking like the person who owns the bakery — the person who knows, viscerally, what that recipe book is actually worth.

Subscribe now

Part One: The Accounting Paradox — The Harder You Work, the Poorer You Look

Let us start with the foundational absurdity that creates the entire opportunity.

Under ASC 730 — the US accounting rule that governs Research and Development — all R&D spending is expensed in the year it is incurred. This means every dollar a pharmaceutical company spends running clinical trials, hiring scientists, and manufacturing drug substance for regulatory purposes flows directly through the income statement as an operating expense in the year it is spent. None of it accumulates on the balance sheet as an asset.

The practical consequence of this rule is dramatic. Consider what Merck’s 10-K filings show: in fiscal years 2012, 2013, and 2014, the company spent $8.2 billion, $7.5 billion, and $7.2 billion respectively on research and development. That is $22.9 billion over three years, vanished from the balance sheet as if it had been burned. The accounting treatment creates a paradox: a company that invests nothing in its future will have a larger book value than one that is furiously building the next generation of medicine.

This inversion is important because Wall Street and the financial modeling community have built their entire valuation framework on top of book value, earnings multiples, and cash flow projections. When a drug company’s earnings look suppressed because $7 billion in R&D is flowing through the income statement, the earnings-per-share number is low. When the book value of assets is artificially deflated because twenty years of R&D spending has been expensed rather than capitalised, the price-to-book ratio looks expensive. The financial model therefore concludes: overvalued. Move on.

The business owner standing in the lab thinks something entirely different. They see a drug that took a decade to build, that has cleared the most difficult regulatory hurdles in the world, and that has no close competitors. They are not looking at the earnings-per-share number. They are thinking about what a competitor would have to spend to build the same thing — and whether they even could.

Subscribe now

Part Two: The Drug Pipeline as a Manufacturing Asset — A Different Mental Model

Here is a mental model that rewires how you think about a pharmaceutical pipeline.

Think of a drug in Phase III clinical trials not as a speculative bet, but as a partially constructed factory — one that, once completed, will generate a stream of high-margin output for the next ten to fifteen years. The factory is not yet producing. It requires final inspections and government certification before it can open. But the equipment has been installed, the process has been validated, and the engineers know it works.

Under GAAP, the cost of building this factory has been expensed. The factory itself sits at zero on the balance sheet. But from the perspective of someone who understands manufacturing economics, what matters is this: what would it cost a competitor to build the same factory? And how long would it take them?

In pharmaceuticals, the answer is typically $800 million to $2.5 billion and ten to fifteen years — if they succeed at all. The Phase I to approval success rate for new drugs, based on Norstella’s analysis of the Citeline dataset covering 2014 to 2023, is 6.7%. That is the rate for the average drug across all disease areas. For drugs that have already cleared Phase II and entered Phase III, the implied probability of approval rises to roughly 55%. For drugs with a validated mechanism of action, a biomarker-selected patient population, strong Phase II endpoint data, and FDA Breakthrough Therapy Designation, the implied probability climbs higher still — into the range of 60 to 80% for best-in-class programs.

A Phase III asset with a 65% implied probability of approval, carried at zero on the balance sheet, is not speculative in any ordinary sense of the word. It is a partially inspected factory that the government is about to certify. The question is not whether you should assign it value. The question is what the right value is — and whether the market has assigned any value at all.

Note: It is important to make an assessment of existing substitutes or solutions in the market (if any). And ask ourselves, what is the price that the drug is selling for? Can the company justify the R&D cost (that is capitalised) based on what the market is already paying? If it is, the capitalised cost would usually be the floor.

Subscribe now

Part Three: The Story of the Drug That Wasn’t There — Merck, 2012–2014

In the summer of 2012, Merck & Co. was a large, moderately valued pharmaceutical company with a stock price trading around $40 to $44 per share — roughly 12 to 14 times trailing twelve-month earnings. The market saw Merck as a mature, dividend-paying pharmaceutical company whose blockbuster drugs (Januvia for diabetes, Singulair for asthma) generated reliable cash flows but faced patent expirations. The narrative was simple: Merck’s best days were behind it. The pipeline was uncertain. The stock was a bond substitute, not a growth asset.

What the financial statements did not show — what they were structurally incapable of showing — was a small clinical program internally designated MK-3475, which would later be named pembrolizumab and eventually marketed as Keytruda.

The science behind Keytruda was not a mystery in 2012. The PD-1 pathway — the biological mechanism that Keytruda targets — had been described in peer-reviewed literature for years. PD-1 is a protein on the surface of immune cells that acts as a kind of off switch: when activated, it prevents the immune system from attacking nearby cells. Tumor cells, which are clever adversaries, had learned to exploit this switch by expressing a ligand called PD-L1, which binds to PD-1 on T-cells and essentially tells the immune system: nothing to see here, keep moving. The result is a kind of biological camouflage — the tumor hides from the immune system in plain sight.

The scientific hypothesis was straightforward: if you block the PD-1 receptor with an antibody, you pull away the camouflage. The immune system sees the tumor again and attacks it.

In June 2012, the New England Journal of Medicine published Phase I data on Bristol-Myers Squibb’s anti-PD-1 antibody, nivolumab, showing cumulative response rates of 28% in patients with advanced melanoma — a disease where the five-year survival rate at that time was under 15%. These were patients who had exhausted all other options. The responses, crucially, were durable: many lasted more than a year. The data was published for anyone to read.

Merck had its own anti-PD-1 antibody — pembrolizumab — in a Phase I trial called KEYNOTE-001 (NCT01295827, publicly registered and visible on ClinicalTrials.gov). In late 2012, the FDA granted pembrolizumab orphan drug designation for advanced melanoma. In 2013 — the same year Merck’s 10-K showed $7.5 billion in R&D expenses flowing through the income statement and zero dollars in pipeline value on the balance sheet — the FDA granted pembrolizumab its Breakthrough Therapy Designation for advanced melanoma.

This was the first Breakthrough Therapy Designation ever granted for a cancer drug.

It was all public information. ClinicalTrials.gov listed the trial. The FDA’s public designations database listed the Breakthrough designation. The BMS nivolumab data, validating the entire mechanism of action, was published in one of the most widely read medical journals in the world.

The stock market largely did not react. Merck continued to trade at 12 to 16 times earnings — pricing that reflected only the existing drug portfolio, with essentially no premium for the MK-3475 program. The balance sheet, which showed zero value for the pipeline, told a story the market believed.

On September 4, 2014, the FDA granted accelerated approval to pembrolizumab — now branded Keytruda — making it the first approved anti-PD-1 therapy in the United States. By 2024, Keytruda was generating over $25 billion in annual revenue, representing roughly 40% of Merck’s total sales. It is the best-selling drug in the world.

The asset was always there. It simply was not visible to anyone looking at the balance sheet.

Subscribe now

Part Four: Learning to See the Invisible — A Practical Framework

The Merck/Keytruda story is clarifying, but only if you can draw lessons from it that are operational — things you can actually do the next time you are looking at a pharmaceutical company. Here is a structured approach to what a business owner, rather than a financial analyst, would examine.

Subscribe now

The First Question: What Has the Company Already Spent, and Where Did It Go?

The starting point is the R&D expense line in the income statement, cross-referenced with the pipeline disclosure in the 10-K annual report. Every major pharmaceutical company publishes a pipeline table — typically in the “Business” section of its annual filing — listing each drug candidate, its therapeutic area, and its current clinical phase. This table contains the actual assets of the business. The income statement contains only the cost of building them.

A company that has spent $3 billion cumulatively on R&D over the past five years, whose balance sheet shows zero in internally developed drug assets, and whose pipeline table lists three Phase III programs and five Phase II programs is not what the income statement makes it look like. The $3 billion was not destroyed. It was converted into those eight clinical programs — assets that GAAP refuses to recognize.

Compare the cumulative R&D expense over five to seven years (found by summing the R&D line from successive 10-Ks) to the current enterprise value of the company. If the cumulative spend is comparable to or exceeds the current enterprise value, you are potentially buying those pipeline assets for free. The existing commercial products — the ones already generating revenue — are being priced as if they are all that exists.

The catch: Only if those product gets commercialised — Which brings us to the next question below.

Subscribe now

The Second Question: Does the Mechanism of Action Actually Work?

This is where most non-specialist investors stop, and where the real work begins.

A drug’s mechanism of action (MOA) is the biological pathway through which it produces its effect. Understanding whether an MOA is validated is not about whether the drug works in theory — it is about whether the FDA and the scientific community have already confirmed that drugs operating on the same pathway can produce clinical benefit in humans.

Here is the distinction that matters. A drug with a first-in-class MOA is asking the FDA and the scientific community to take a genuinely novel leap of faith: we believe this pathway is relevant to disease X, we believe blocking/activating it will produce benefit, and here is our preliminary evidence. Even with strong data, this involves the highest level of regulatory and scientific uncertainty.

A drug with a best-in-class MOA operating on an already-validated target is asking something much simpler: the pathway is proven to matter. Other drugs working on the same target have already been approved. Our drug works on the same mechanism, and our data suggests we do it better, faster, or with fewer side effects. The FDA is not being asked to validate the mechanism — only to confirm the drug’s clinical performance.

This distinction has a measurable impact on success rates. A 2021 analysis published in Drug Discovery Today found that follow-on drugs targeting validated mechanisms had meaningfully higher Phase III success rates than first-in-class compounds, because the fundamental scientific hypothesis — that the target matters in the disease — had already been independently confirmed.

The practical question to ask is: Has any drug operating on the same target or pathway already received FDA approval? If yes, for what disease? And does the company’s drug share enough of that mechanism that approval in the same disease, or an adjacent disease, is scientifically plausible?

In the case of pembrolizumab in 2013, this question had a clean answer. The anti-PD-1 mechanism was validated by BMS’s nivolumab data, published in 2012. The pathway — blocking PD-1 to restore immune recognition of tumors — was proven to produce durable responses in melanoma patients. Merck’s drug worked on the identical mechanism. The question was not whether anti-PD-1 works. The question was whether pembrolizumab was a good enough version of the mechanism to achieve regulatory approval.

How to do this in practice: Search ClinicalTrials.gov for the drug’s target (e.g., “PD-1 inhibitor,” “KRAS G12C,” “BTK inhibitor,” “JAK1/2”) and filter by status “Approved” or “Completed.” Look for drugs in the same or similar disease area with positive results. If prior approvals exist, the mechanism is validated. Then look at the published Phase I and Phase II data for the company’s specific drug and ask: does the response rate, the durability of response, and the safety profile compare favorably to the drug that was already approved?

Subscribe now

The Third Question: What Does the Clinical Data Actually Say?

Every clinical trial registered with the FDA is publicly listed on ClinicalTrials.gov. Every company is required to post results there within a specified time after trial completion. Published Phase I and Phase II data are almost always available in peer-reviewed journals — the New England Journal of Medicine, the Journal of Clinical Oncology, The Lancet Oncology, and Nature Medicine publish the major oncology trial results.

For a non-specialist reading this data, the key numbers to look for are:

Overall Response Rate (ORR): The percentage of patients whose tumors shrank or disappeared by a predefined threshold. A 30% ORR in a tumor type where the standard of care produces 12% is a meaningful signal. A 7% ORR in the same context is not.

Duration of Response: How long the responses lasted. A drug that produces a 40% response rate but all responses disappear within three months is different from one where responses last two years. Durability signals that the drug is producing genuine biological change, not just temporary tumor suppression.

Progression-Free Survival (PFS): The median time patients go without their disease getting worse. Compare this to published benchmarks for the current standard of care in the same tumor type.

Safety Profile: What proportion of patients experienced grade 3 or 4 adverse events (serious or life-threatening side effects). A drug with a 60% ORR but 40% grade 3/4 adverse events is a different risk profile than one with 60% ORR and 8% grade 3/4 events.

These numbers are not hidden. They are published, peer-reviewed, and discussed at public oncology conferences (ASCO, ESMO, ASH) every year. The conference abstracts are freely accessible online. The question is simply whether anyone looking at the company’s stock price is reading them.

Subscribe now

The Fourth Question: What is the delivery mechanism?

A drug can be a scientific miracle in a test tube, but if the “user interface”—how the patient actually receives the medicine—is a logistical nightmare, its commercial value will be severely capped. In the pharmaceutical world, the delivery mechanism is the bridge between clinical efficacy and market share.

When evaluating a company’s pipeline, you must distinguish between the payload (the active chemical or biological agent) and the vehicle (how it enters the body). The market’s “acceptance” of a drug is often a direct function of how much it disrupts a patient’s daily life.

Generally, the market values delivery mechanisms in a specific order of “friction”:

• Oral (The Gold Standard): A pill or capsule that can be taken at home. This offers the highest patient adherence and the lowest cost to the healthcare system. It requires no specialized equipment or medical supervision.

• Subcutaneous (SubQ) Injection: A “shot” typically administered into the fatty tissue. Many modern drugs (like the GLP-1 weight-loss medications or insulin) use “autoinjector pens” that allow patients to treat themselves at home.

• Intravenous (IV) Infusion: The most invasive. This requires a patient to travel to a clinic, sit in a chair for thirty minutes to several hours, and have a healthcare professional administer the drug via a vein.

A common value-creation strategy for savvy pharma companies is reformulation. For example, a company might have an oncology drug that is currently delivered via a grueling two-hour IV infusion. If that company successfully develops a Subcutaneous version of the same drug that can be administered in five minutes, they have just created a massive market.

As a business owner, you must ask: Does the delivery mechanism match the disease?

• If you are treating a life-threatening, late-stage cancer, a patient will tolerate a four-hour IV infusion.

• If you are treating a chronic, manageable condition like high blood pressure or mild asthma, an IV requirement is a commercial death sentence. No one is going to an infusion center once a month for a condition they could treat with a daily pill.

How to do this in practice: Look at the “Method of Administration” in the clinical trial description. Then, look at the competitors. If the “Standard of Care” is a pill and the new company’s drug is an injection, the new drug’s clinical data doesn’t just have to be as good as the pill—it has to be exponentially better to convince doctors and insurance companies to switch to a more invasive, expensive delivery method.

Subscribe now

The Fifth Question: What Has the FDA Formally Said?

This is where many investors are most confused, because the FDA designation system is a series of overlapping signals that mean different things at different stages of development. Let us go through each one carefully, as a business owner evaluating how close a factory is to receiving its government operating license.

Part Five: A Complete Guide to FDA Designations — What the Government Is Telling You

Think of the FDA not as a black box that either says yes or no at the end of a decade of waiting, but as an ongoing scientific dialogue between the regulator and the drug company. The FDA issues formal designations at different points in that dialogue. Each designation is a signal about what the FDA has seen and what it is willing to do in response.

The Phases in Summary

Subscribe now

Fast Track Designation

Fast Track is the most commonly granted designation — and the least informative as a signal of quality. The FDA grants Fast Track to drugs intended to treat serious conditions where there is an unmet medical need, essentially meaning the existing treatments are inadequate. Fast Track does not require any clinical data from human trials. A company can apply for Fast Track based on preclinical (animal or cell culture) evidence and a reasonable scientific hypothesis.

What Fast Track does provide is practical: more frequent meetings with FDA staff, the ability to submit sections of the NDA (new drug application) on a rolling basis (rather than all at once), and eligibility for Priority Review if the drug shows significant improvement in later trials.

NDA is usually submitted after Phase 3

Fast Track is not a signal that the drug works in humans. It is a signal that the FDA is willing to talk to you more frequently because the target disease is serious and the current treatments are inadequate.

What to do with Fast Track: It tells you the FDA is interested in the disease area and that preliminary science is plausible. By itself, it is a necessary but not sufficient condition for the kind of pipeline value assessment we are discussing. Most pipeline drugs eventually get Fast Track. What matters is what comes next.

Orphan Drug Designation

Orphan Drug Designation is granted to drugs targeting diseases affecting fewer than 200,000 people in the United States. It comes with significant economic incentives: seven years of market exclusivity upon approval (even if patent protection would otherwise be shorter), 50% tax credits on qualified clinical trial costs, and waived FDA filing fees.

For investors, Orphan designation carries two distinct signals. First, it confirms that the FDA has reviewed the scientific rationale for using the drug in the specific disease and accepted that there is a plausible pathway to therapeutic benefit. Second — and more commercially important — it signals that if approved, the company will face a seven-year window without generic competition regardless of patent life.

In the Keytruda story, Merck received Orphan Drug Designation for pembrolizumab in advanced melanoma in late 2012. This was a factual, public signal that the FDA had reviewed the preliminary science and found it credible enough to grant a formal designation. It was publicly listed in the FDA’s Orphan Drug Products database. It was one piece of the visible evidence that the drug had scientific and regulatory traction — visible to anyone who chose to look.

Subscribe now

Breakthrough Therapy Designation (BTD)

This is the most important designation for the kind of pipeline valuation assessment we are building. It is the FDA’s formal statement that a drug has shown preliminary clinical evidence — meaning actual data from human patients — of substantial improvement over existing therapies on a clinically meaningful endpoint.

To receive BTD, a company must submit Phase I or Phase II human data showing that the drug is meaningfully better than the standard of care in a disease with serious morbidity or mortality. The FDA evaluates that data and makes a judgment. The bar is real: as of mid-2024, the FDA had received 1,516 requests for BTD and granted 587 — a 38.7% grant rate. The FDA is turning down roughly 60% of requests.

What BTD provides in practice: intensive guidance from senior FDA staff, the ability to use the FDA’s rolling review process, and priority review of the final application. The clinical guidance is particularly valuable — companies with BTD essentially get ongoing scientific feedback from the FDA about what data will be sufficient for approval, reducing the risk of producing a beautiful trial that fails on a technicality.

The approval statistics for BTD drugs are striking. A Jefferies analysis of 599 Breakthrough Therapy Designations granted between 2013 and 2022 found that 72% of designated drugs ultimately received FDA approval, with an additional 13% still in development at the time of the analysis. Compare this to the baseline Phase I-to-approval rate of 6.7%. A drug with BTD is not just a drug — it is a drug that the FDA has already looked at in human patients and said, formally: this appears to be substantially better than what we have.

That statement carries enormous weight and no corresponding weight on the balance sheet.

How to use BTD in your assessment: Treat BTD as a meaningful upward revision to the implied probability of approval, but not a guarantee. A drug with BTD in a validated MOA, positive Phase II data, and a biomarker-selected patient population is operating with a very different risk profile from the average pipeline drug. The relevant base rate is not the 6.7% overall Phase I-to-approval number. The relevant base rate is closer to the 72% eventual approval rate for BTD drugs — with the important caveat that some of those drugs had many years of additional trials left when the designation was granted.

Keytruda received BTD in 2013, before its Phase III confirmatory trials. The designation was the first BTD ever granted for a cancer drug. It was announced in a press release, listed on the FDA’s website, and discussed in Merck’s annual reports. It was visible.

Subscribe now

Priority Review

Priority Review is granted at the time a company submits its New Drug Application or Biologics License Application (BLA) to the FDA — the formal application for market approval. It shortens the FDA’s review period from the standard ten months to six months, for drugs that offer significant improvements in safety or effectiveness over available therapy.

Priority Review does not reflect judgment about whether the drug will be approved — it reflects a judgment about whether the FDA should move faster given the potential benefit. It is granted after Phase III trials are complete and the full data package is in hand.

For investors, Priority Review is important as a confirmation signal. It tells you the FDA reviewed the Phase III package and immediately classified the drug as offering meaningful improvement. Combined with BTD and strong clinical data, Priority Review is one of the last checkpoints before approval. A drug in Priority Review with 4-6 months until a decision date (the PDUFA date) is extremely close to either becoming a commercial product or confirming it has failed.

Subscribe now

The PDUFA Date: The Confirmed Catalyst

The Prescription Drug User Fee Act (PDUFA) date is the FDA’s committed deadline to complete its review of a drug application. It is public, it is specific, and it is almost always met. When a company submits its NDA or BLA, the FDA sets a PDUFA date — typically ten months out for standard review, six months for Priority Review — and publishes it in its calendar.

The PDUFA date is the closest thing to a confirmed, dated catalyst that exists in pharmaceutical investing. It is not a guarantee of approval. But it is the date on which the FDA has committed to making a decision, and for drugs with strong clinical data, validated mechanisms, and favorable designations, it converts an abstract probability into a specific binary event on a specific date.

Unlike “we think this will generate revenue in three years,” the PDUFA date is published fact. The FDA’s public calendar lists upcoming PDUFA dates for drugs under review. This is the anchor point for assessing when the invisible factory opens — or confirms that it will not.

Subscribe now

Part Six: Biomarkers — The Single Most Important Variable in Clinical Success Rates

Of all the factors that determine whether a drug will succeed in Phase III trials, none has had more impact on outcomes in the past decade than the use of biomarkers for patient selection.

A biomarker, in this context, is a measurable biological characteristic — a gene mutation, a protein expression level, a chromosomal rearrangement — that identifies patients whose tumors are likely to respond to a specific drug. Rather than testing a drug in everyone with, say, breast cancer, a biomarker-selected trial tests the drug only in patients whose tumors carry the specific feature the drug is designed to target.

The impact on success rates is not marginal. It is transformative. A 2016 analysis of over 10,000 clinical trials drawn from ClinicalTrials.gov, published in the Journal of Health Economics, found that trials using biomarkers for patient selection had nearly fivefold higher likelihood of drug approval compared to trials without biomarker selection. A 2015 study published in Biostatistics analyzing 406,038 clinical trial entries confirmed that “trials using biomarkers in patient-selection have higher overall success probabilities than trials without biomarkers.” And a targeted example from the HER2 mutation in breast cancer showed response rates improving from 12% in unselected patients to 32% in biomarker-selected patients — a nearly threefold improvement.

The intuition behind this is straightforward. If a drug works by blocking a specific protein that only 25% of tumors produce, then testing it in an unselected population means 75% of patients in the trial will have no chance of responding. The drug will look like it does not work — because in that population, it mostly does not. A biomarker test identifies the 25% who have that protein and tests the drug only in them. The response rates are dramatically higher, the statistical signal is cleaner, and the trial is smaller and faster.

What to look for in a Phase II or Phase III program:

First, look at the trial design on ClinicalTrials.gov. Is the trial requiring a biomarker test as an eligibility criterion? If yes, which biomarker? Is that biomarker scientifically connected to the drug’s mechanism of action, or is it coincidental? A drug targeting KRAS G12C mutations that requires a KRAS G12C test for enrollment is doing this correctly. A drug with a diffuse mechanism of action using a loosely correlated biomarker is doing this for appearances.

Second, look at whether the company has developed or partnered on a companion diagnostic — an FDA-approved test that identifies patients who should receive the drug. Companion diagnostics are required by the FDA for drugs where the biomarker is essential for identifying the appropriate patient population. If a company has already submitted or received approval for a companion diagnostic, the biomarker relationship is so strong that the FDA has required it as a label condition. That is a high signal of scientific rigor.

Third, compare the Phase II response rates to the historical rates in that tumor type for both biomarker-selected and unselected trials. A response rate that looks impressive in absolute terms may be unremarkable if other drugs targeting the same biomarker have already achieved similar results. But a response rate that is meaningfully superior to the best available biomarker-selected therapy, in a patient population where the biomarker prevalence is high, is a signal worth taking seriously.

Subscribe now

Part Seven: The Screening Arbitrage — When the Market Forgets the Factory Exists

All of the above analysis is only useful if the market is pricing the company in a way that assigns limited or no value to the pipeline. There is no point to purchase a company that is selling for a premium, or more than the value of which you expect it to give.

An example of a good indicator that the market has forgotten the invisible factory is a Price-to-Sales ratio below 2.0. At that multiple, the market is essentially saying: this company’s future revenue is worth less than twice its current annual sales. For a pharmaceutical company with a productive pipeline, that pricing implies that the pipeline will produce either nothing or something close to it. However, for a pure R&D company, this has no relation.

Let us make this concrete. A drug company with $3 billion in annual revenue from existing commercial products, trading at 1.8 times sales, has an enterprise value of approximately $5.4 billion. Its 10-K lists two Phase III programs and three Phase II programs. Its cumulative R&D expense over the past six years is $4.2 billion — all expensed, all invisible on the balance sheet. The market is pricing the company as if the R&D produced nothing worth acknowledging. The business owner — the person who understands that the $4.2 billion was converted into five clinical programs, not incinerated — is asking a very different question.

The P/S ratio below 2.0 is not a sufficient condition for an investment conclusion. It is a signal that the market may be undervaluing the pipeline. The next steps are the ones described in Parts Four through Six: examine the pipeline for mechanism validation, clinical data quality, biomarker selection, and FDA designations. The P/S ratio identifies the pool. The clinical assessment narrows it.

Note: Having a drug pipeline increases the chance of the company succeeding (because they could pre-license the drug out for working capital.etc) and flexibility in structuring creative deals which does not dilute shareholders equity. However, if the company (1) does not have a drug pipeline, and (2) does not have sufficient cash or earnings, it is very likely that the company would have to raise additional capital to push the drugs through approval. In most cases where revenues are negligible, and asset values are negligible, raising debt poses a hurdle whilst an equity raise would make the most sense.

Subscribe now

Part Eight: The Cautionary Counterweight — Why This Is Not as Easy as It Sounds

Honesty requires a counterpoint to everything above.

The 6.7% Phase I-to-approval success rate is a real number. So is the 28% Phase II-to-Phase III transition rate. The cemetery of pharmaceutical stocks is full of companies that had validated mechanisms, strong Phase II data, and Breakthrough Therapy Designations and still failed in Phase III. Sometimes the patient population was too small for the statistical signal to hold. Sometimes the side effects that appeared acceptable in 200 Phase II patients became disqualifying in 800 Phase III patients. Sometimes a drug works beautifully in the biomarker-positive 30% of patients but the Phase III trial design required efficacy across the full population.

Phase III failure is not rare. It is the standard outcome for most drugs. What the framework above does is narrow the distribution — it identifies programs where the background conditions are most favorable — but it does not eliminate the binary nature of the outcome.

The honest acknowledgment, as noted in the source materials, is that you cannot put an immediate liquidation value on a pipeline drug. You cannot sell the Phase III asset tomorrow and walk away with cash. The value is probabilistic, illiquid, and time-dependent. A Phase III drug is worth much more than zero, but it is not worth as much as an approved commercial product generating current revenue.

What you can do is recognize when the market has priced a company as if its pipeline is worth zero or close to it, when the publicly available evidence suggests that the pipeline contains programs with the specific profile — validated mechanism, biomarker selection, strong early data, favorable FDA designations, imminent PDUFA dates — that meaningfully narrows the distribution toward success. The gap between “zero” and “non-trivially positive probability” is where the value lives.

Subscribe now

Conclusion: The Map Was Always Public

Every piece of evidence that made pembrolizumab’s development look promising in 2012 and 2013 was publicly available. The PD-1 mechanism was described in academic journals. BMS’s nivolumab Phase I data was published in the New England Journal of Medicine. Merck’s KEYNOTE-001 trial was registered on ClinicalTrials.gov with its design and eligibility criteria visible. The FDA’s Orphan Drug Designation and Breakthrough Therapy Designation for pembrolizumab were posted on public FDA databases and announced in press releases. Merck’s 10-K filings disclosed the pipeline with sufficient detail to identify the program as late-stage and prioritized.

The financial models did not capture any of this, because financial models price earnings, and Keytruda had no earnings. The balance sheet did not capture any of this, because GAAP expenses R&D and carries internal drug development at zero. The analysts who covered Merck were focused on the patent cliff on Januvia and the declining pipeline of marketed products.

The business owner — the person thinking about what that factory was actually worth — was reading different documents. They were reading the New England Journal of Medicine article on nivolumab and asking: does Merck’s drug work on the same mechanism? They were checking the ClinicalTrials.gov page for KEYNOTE-001 and asking: how are the response rates? They were reading the FDA’s designation database and asking: what does it mean that the FDA just called this the first Breakthrough Therapy for cancer?

None of those questions required a financial model. They required an understanding of what the asset actually was, what it would be worth if it worked, and what the public evidence said about whether it would work.

That gap — between the financial statement that shows zero and the economic reality that shows a high-probability, high-value manufacturing asset nearing operational certification — is the hidden value. It does not disappear simply because the accountants cannot see it.

Note: This whitepaper references data from clinical trial databases, FDA public records, peer-reviewed medical literature, and public company filings including Merck’s 10-K annual reports for fiscal years 2012, 2013, 2014, and 2015 (SEC filings). Clinical success rate data draws on the MIT/Boston University biostatistics study by Wong, Siah, and Lo (2019), the Norstella/Citeline dataset analysis (2014–2023), and the Jefferies analysis of FDA Breakthrough Therapy Designations (2013–2022). All FDA designation statistics reflect publicly available data from the FDA’s designation databases and peer-reviewed analysis of that data.

If you like what you read, check out my other writings 👇

A Deep Reflection on a Week of Investment Research — May 2026

“An investor needs to do very few things right as long as he or she avoids big mistakes.” — Warren Buffett

Most people think investing is about finding the right stock. It isn’t. It’s about building the right mind. A mind that can look at the same data everyone else sees and extract something others miss — not because of superior information, but because of superior translation.

This week, working across five business analyses — Gran Tierra Energy (GTE), Proficient Auto Logistics (PAL), Gumtree Australia Markets (GUM), Goodyear Tire (GT), and Red Robin (RRGB) — I didn’t just collect facts. I practiced a more difficult and more valuable skill: turning raw financial and operational data into durable mental models.

This article is a rigorous unpacking of what I actually learned, why it matters, and how I plan to compound it.

Subscribe now

Part I: The GAAP Illusion — Learning to Read Two Sets of Books

“Accounting is the language of business, but it is a language with dialects — some of which obscure more than they reveal.”

The Problem With Reported Numbers

One of the most persistent traps in equity research is letting GAAP net income anchor your entire view of a business. This week, two companies forced me to confront that trap directly.

Gran Tierra Energy reported a GAAP net loss of $119 million in Q1 2026. On the surface, that looks like a catastrophe for a company with a market cap in the hundreds of millions. But strip away the noise:

• The loss was dominated by unrealized hedging paper losses — mark-to-market charges on oil price protection contracts that hadn’t settled yet. These are not cash outflows. They reverse if oil moves the other direction.

• Depletion and asset valuation adjustments — also non-cash, driven by accounting conventions on reserve drawdowns, not by the physical business deteriorating.

The cash story told a completely different picture: funds flow from operations jumped 60% quarter-over-quarter. Full-year 2026 EBITDA guidance was set at $345–395 million.

Proficient Auto Logistics did the same thing. A $27.8M goodwill impairment in Q4 2025 felt alarming at first glance — a signal that an acquisition was mispriced, perhaps a sign of broader misjudgement by management. But goodwill impairments are backward-looking accounting entries, not forward-looking cash charges. They tell you what a prior deal was worth, not what the current network generates. The underlying adjusted operating metrics told a far more constructive story about customer pull and competitive positioning.

Subscribe now

The Translation Framework

What I’m building — almost unconsciously — is a three-step translation discipline whenever I look at a P&L:

Step 1: Separate recurring from non-recurring. Is this charge something that will appear every quarter, or is it a one-time accounting or structural event? Goodwill impairments, restructuring charges, and unrealized derivatives are frequently one-time.

Step 2: Separate cash from non-cash. Depreciation, amortization, depletion, unrealized gains/losses — these don’t move the bank account. Start with operating cash flow or funds flow, and reconcile backward.

Step 3: Ask what changes the long-term cash power. The only thing that matters long-term is whether the business can generate more cash, at better margins, on a growing base. Everything else is noise you have to learn to mute.

Subscribe now

Why This Is a Compounding Skill

The reason this matters isn’t just for these two companies. It’s that this gap — between GAAP headlines and economic reality — is structurally persistent in several sectors: oil & gas, industrials with heavy capex, turnarounds with restructuring charges, businesses using derivatives for hedging. These sectors produce large, systematic mispricings precisely because most market participants anchor on reported numbers. Investors who can read through the accounting earn a durable informational edge.

Principle I’m internalizing: The bigger the gap between GAAP earnings and free cash flow / funds flow, the more important it is to understand the character of that gap. Sometimes it signals fraud or value destruction. More often, it signals opportunity.

Part II: The Anatomy of a Real Catalyst

“Everyone has a plan until they get punched in the mouth.” — Mike Tyson (applicable to investment theses)

Why Most Catalysts Aren’t Real Catalysts

In investment circles, the word “catalyst” gets thrown around loosely. It often just means: I believe something good will happen. That’s not a catalyst. That’s a hope. Real catalysts are specific, observable, mechanism-driven changes that cause measurable improvement in cash flow, margins, volumes, or valuation multiples — and they are falsifiable: if they don’t show up in data within a defined timeframe, the thesis is wrong.

This week, I started to feel the difference.

Subscribe now

Gran Tierra Energy: Mechanisms, Not Stories

GTE’s three catalysts aren’t stories — they’re operational mechanics:

Catalyst 1 — Oil Tailwind Operating Leverage. GTE’s cost structure is largely fixed. At a $48/barrel price baseline and $26/barrel cost, every dollar of Brent price increase above $48 flows almost entirely to the bottom line. In Q1 2026, a 24% quarter-over-quarter increase in realized Brent prices plus a 12% volume increase drove a 32% sequential jump in oil sales revenue and a 42% surge in adjusted EBITDA. This is not a narrative — it’s operating leverage doing what operating leverage does. The key monitor: track the relationship between Brent moves and EBITDA flow-through. If that ratio degrades, something structural has changed.

Catalyst 2 — Volume Growth via Waterflooding. GTE’s Acordionero field in Colombia is being waterflooded — injecting water to repressurize reservoirs and extend productive life of mature wells. This is proven, low-risk reservoir engineering, not exploration risk. The result: production is being held at 40,000–45,000 BOE/day rather than declining naturally. The key monitor: watch quarterly production reports. If waterflooding isn’t holding the curve, the thesis degrades.

Catalyst 3 — Infrastructure Ownership and Netback. GTE owns its own pipeline and loading terminal. This eliminates third-party transit fees, directly expanding the Operating Netback (profit per barrel after field costs and transport). Colombia field operating costs are frequently below $15–16/barrel — one of the lowest cost structures among Latin American independents. The key monitor: watch field operating costs per BOE and compare to peers.

Subscribe now

Goodyear Tire: Already-Triggered Catalysts

Goodyear presented a different challenge: the catalysts had already triggered. The question was whether the market had correctly priced in the benefits, or whether the sequential improvements in margins, deleveraging, and cost savings were still being undervalued.

The clean trigger list:

1. Sale of chemical division completed (Q3 2025) → removes capital-intensive, volatile drag → deleveraging enabled

2. Corrective pricing in Expedited Freight completed (February 2025) → revenue per hundredweight rising for two consecutive quarters

3. $250M+ of “Goodyear Forward” cost savings expected to flow through in 2026

4. ~$85M raw materials tailwind in Q1 2026 alone; ~$200M annualized benefit

But the risks are real and symmetrical: low-cost Asian tire imports flooding key markets and tariff-driven raw material cost headwinds of $150–160M. This is the correct way to think about a catalyst map — it’s not just upside levers, it’s levers in both directions, and the question is which force is larger.

Subscribe now

The Catalyst Checklist I’m Building

From this week, I’m formalizing a three-part catalyst evaluation:

1. Mechanism: What is the specific operational or financial lever being pulled? Can I describe it without using the words “improve,” “grow,” or “optimize”?

2. Timeline: When should this show up in the reported numbers? Quarters? Years? If it’s “eventually,” that’s not a catalyst.

3. Falsification: What would I expect to see in the data if this catalyst isn’t working? What’s my trip wire?

If I can’t answer all three cleanly, I don’t have a catalyst. I have a hope.

Subscribe now

Part III: Saying No — And Winning Anyway

“The most important thing to do if you find yourself in a hole is to stop digging.” — Warren Buffett

Red Robin and the Discipline of the Honest Pass

The Red Robin writeup was, counterintuitively, one of the most valuable pieces of work this week — and I passed on the stock entirely.

Red Robin is a casual dining operator in a structurally difficult business. But the reason I studied it carefully was that their tactical execution is genuinely interesting:

• Appointment Dining via Red Robin Royalty®: Instead of generic discounting (which destroys margins), they engineer urgency around specific calendar events — birthdays, graduations, seasonal moments. They create occasions rather than just transactions. A birthday-burger upgrade when you hit a yearly spend milestone isn’t a coupon. It’s behavioral engineering that increases customer lifetime value.

• Big Yummm Tiered Value Platform: Rather than random discounting, they created a structured entry price point ($9.99–$16.99 tiers) that now represents 13% of total sales. This is smart: it gives price-sensitive customers an on-ramp without collapsing check averages for everyone else. Combined with a 3.2% menu price increase, average guest check grew 1.0% — modest, but directionally correct and margin-accretive.

• Labor Efficiency via Accountability: Total Q1 2026 labor cost fell to 35.7% — the lowest in three years. This was achieved by pushing real P&L accountability to local restaurant managing partners, not just central mandates.

• Refranchising and Portfolio Cleaning: They closed 6 underperforming restaurants in Q1 2026 and are on track for ~20 closures for the year. Refranchising converts fixed corporate costs into variable franchise fees. These are the right moves. But 20 closures out of 500+ is still incremental, not transformative.

Subscribe now

The Honest Assessment

All of this is fine. New CEO David Pace (formerly Dine Brands) knows the turnaround playbook. But casual dining is a business fighting multiple simultaneous headwinds:

• Secular shift away from full-service, sit-down dining toward fast-casual and delivery

• Structurally high and rising labor costs (even at 35.7%, this is among the highest cost lines in any business)

• Dependence on foot traffic, which doesn’t compound

• No pricing power in a world where Chipotle and Five Guys exist

• $175.7 million credit facility to pay down with thin cash generation

Conclusion: No. But here is where the real learning lives.

The Meta-Lesson: Transferable Structural Insights

The appointment dining model — engineering occasion-based urgency rather than relying on passive discounting — is a transferable framework. It works at Red Robin only as a defensive survival tactic. But the same mechanism, applied in a business with genuine pricing power, could be extraordinarily powerful. Think of how luxury brands create urgency around limited releases. Or how SaaS companies tie renewals to usage milestones. The mechanism is universal; the context determines whether it creates value or just delays decline.

This is what it means to study “No” businesses. The stock was a pass. The mental model was a keep.

Principle: Every company I study, even ones I reject, should leave me with at least one reusable principle. If I can’t articulate it, I didn’t study hard enough.

Subscribe now

Part IV: The Power of Subtraction — Focused Businesses and Structural Clarity

“The difference between successful people and really successful people is that really successful people say no to almost everything.” — Warren Buffett

Gumtree Australia Markets — A Case Study in Deliberate Simplification

Gumtree Australia Markets (GUM) is not glamorous. It’s an online classifieds marketplace — think Australian Craigslist — focused on Motors, Pets, Jobs, and Real Estate. But what caught my attention was a deliberate, multi-year process of strategic subtraction that is now beginning to show up cleanly in the numbers.

In FY25, GUM completed a pivotal move: it sold its Capital Markets division (HotCopper, Stockhouse, The Market Online) to ADVFN for $6.8M. This sounds like a routine divestiture. But the effect was transformative. That division was lower-margin, capital-intensive, and structurally unrelated to classifieds. Removing it meant that the remaining business — pure-play Australian classifieds — could be evaluated on its own economics.

And the economics, once visible, looked quite different:

• Core classifieds EBITDA: $9.3M, up 32% year-on-year post-divestiture

• Net profit nearly doubled to $3.4M (classifieds only)

• Group net loss narrowed to approximately $100K — essentially breakeven

• H1 FY26: EBITDA up 23% to $3.9M, pre-tax profit returned at $1.4M, operating cash flow $3.4M

• $2.2M in loan repayments made; balance sheet actively cleaning up

Revenue declined 7% year-on-year — but this is explained entirely by the wind-down of Capital Markets. The core unit economics are improving. The CEO who executed the turnaround (Tommy Logtenberg) was formally appointed Managing Director, signaling board confidence. A takeover bid at $0.132/share was rejected by major shareholders who believed the standalone story would deliver more.

The Adyen integration for Gumtree Pay is launching, introducing monetization beyond listings. Monthly users have grown to 7M via a partnership with Australian Community Media.

Subscribe now

Why Simplification Is Underrated as an Investment Signal

Most investors focus on addition — new products, new markets, new segments. But some of the most reliable early signals of an impending re-rating come from subtraction: selling divisions that obscure core economics, closing loss-making units, narrowing focus.

The pattern looks like this:

1. A conglomerate or diversified business is undervalued partly because the market can’t cleanly value its core.

2. Management or activists force a divestiture.

3. The remaining core business is now clearly visible — and its economics are better than the blended group suggested.

4. The market re-rates to reflect the clean core.

GUM is in step 3, potentially moving toward step 4. This is a structurally interesting risk/reward if the classifieds EBITDA can sustain its growth trajectory.

Principle: When a business has recently simplified — sold a division, exited a geography, shut down a segment — read the pre and post numbers carefully. The “before” may have been obscuring a clean “after” that is now visible for the first time.

Subscribe now

Part V: The Deeper Meta-Skill — Systematic Translation

If there is one overarching theme from this week, it is this: the most valuable investor skill is not finding information. It’s translating it.

Everyone has access to the same 10-Q, the same earnings call transcript, the same financial data. The edge is not in finding data others don’t have — it’s in seeing what the data means when others only see what it says.

This week I practiced several forms of translation:

Translation 1: Accounting → Economics. Strip out non-cash, non-recurring, and derivative items. What is the business actually generating in cash? (GTE, PAL)

Translation 2: Tactical detail → Transferable principle. The Red Robin loyalty mechanics are interesting not because Red Robin is a buy, but because the mechanism of occasion-based behavioral urgency can be applied anywhere. (RRGB)

Translation 3: Complexity → Clarity. When a business sheds divisions, the complexity was hiding economics. Learn to read the “after” balance sheet, not just the “before.” (GUM)

Translation 4: Catalyst narrative → Measurable mechanism. Don’t let a thesis live as a story. Convert it to a specific, observable, falsifiable lever — and attach a timeline and a trip wire. (GTE, GT)

Subscribe now

Closing: The Compounding Commitment

Charlie Munger famously said that in his entire life, he has known no wise person who didn’t read all the time. The implicit follow-up is: reading is necessary but not sufficient. The value is in the synthesis — in pulling the thread across disparate observations and building something durable.

This week I read five businesses. Across them, I extracted five first-order principles. But more importantly, I’m beginning to extract meta-principles — principles about how to think, not just what to think.

The commitment going forward is to close the loop more explicitly:

• For each company studied, define the single most important metric to track going forward.

• For each thesis formed, define the disconfirming evidence that would kill it.

• For each “No,” articulate the transferable principle in one sentence.

Because this work doesn’t just compound in a portfolio. It compounds in a mind. And that’s the only kind of compounding that can’t be taken away.

Subscribe now

📌 Week’s Core Principles — Consolidated

1. GAAP ≠ economics. Always reconcile to cash; the biggest opportunities hide in the largest gaps.

2. A catalyst must be a mechanism. If you can’t measure it, you don’t have a catalyst.

3. Operating leverage cuts both ways. Fixed costs are a weapon in upturns and a vulnerability in downturns — size your conviction accordingly.

4. Study “No” businesses rigorously. The ROI is the mental model extracted, not the trade.

5. Subtraction reveals truth. Divestitures and focus can make visible what complexity was hiding.

6. Every thesis needs a trip wire. What would I need to see — or not see — in the next two quarters to know I’m wrong?

   Thanks for reading Compounding Zero! Subscribe for free to receive new posts and support my work.

If you like what you read, check out my other writings 👇

“It was awfully easy money. It was like having God just opening a chest and just pouring money into it.” — Charlie Munger, Acquired Podcast, November 2023

Preface: On the Nature of This Paper

This paper does not forecast earnings. It does not build discounted cash flow models or assign price targets. It does not speculate about what these businesses will be worth in ten years. What it does instead is something both simpler and harder: it reconstructs, from primary sources — SEC-filed bond prospectuses, annual reports, and the words of the men who made the trade — the exact mechanical architecture of what may be the most structurally elegant large-scale investment of the last fifty years. The goal is understanding of the thought process of a genius, not the prediction of a business’ potential.

Subscribe now

Part One: The Book

In the summer of 2019, Warren Buffett — then 88 years old, running a company with a market capitalisation approaching half a trillion dollars, and supposedly incapable of finding anything worth buying at scale — sat down with a copy of the Japanese Company Handbook, a semi-annual directory published by Toyo Keizai. The handbook is not elegant. It contains no narrative, no analysis, no recommendation. It is a reference book in the truest sense: roughly 3,000 companies compressed into a single volume, one page per company, each page containing revenue, earnings, net assets, dividends, and a recent price range. It is the kind of document that most professional investors in 2019 had never heard of, because most professional investors in 2019 had dismantled their Japan desks and redirected their attention to China.

Buffett has described his fundamental research methodology as “turning every page.” He meant it as a young man combing through Moody’s manuals in the 1950s, and he meant it literally in 2019 with the Japanese handbook. At the 2025 Berkshire Hathaway annual meeting, his last as CEO, he recalled the discovery plainly: “I was just going through a little handbook that probably had two or three thousand Japanese companies in it. There were these five trading companies selling at ridiculously low prices. So I spent about a year acquiring them. Turning every page is one important ingredient to bring to the investment field. And very few people who turn every page are going to tell you what they find, so you’ve got to do a little of it yourself.”

The five companies he found — Mitsubishi Corporation, Mitsui & Co., Itochu Corporation, Sumitomo Corporation, and Marubeni Corporation — are collectively known in Japan as sogo shosha, or general trading companies. They are among the oldest and most globally extensive commercial enterprises on earth. Itochu traces its roots to 1858. Mitsubishi was founded in 1870. Sumitomo’s origins in copper trading extend to the seventeenth century. By 2019, each was generating tens of billions of dollars in annual revenue, operating in more than 60 countries, holding equity stakes in hundreds of businesses across every major sector of the global economy. And each was trading at between 0.5 and 0.8 times the stated accounting value of its net assets, on earnings multiples in the single digits, with dividend yields between 4 and 7 percent.

The Western investment community had collectively decided these companies were not worth understanding. They were labelled “trading companies” — a description that conjured images of commodity middlemen operating on thin margins — and filed away under the category of things that sophisticated modern capital allocators did not need to think about. This misclassification was not an accident. It was the product of genuine analytical difficulty: the sogo shosha are, as Moody’s formally recognises, a category unto themselves with no Western comparable. They are simultaneously trading houses, private equity firms, infrastructure operators, food distributors, retail operators, energy developers, and financial intermediaries, all in one entity, operating across every industry and continent at once. Most analysts, confronted with something they cannot categorise, simply move on. Buffett, confronted with something he cannot categorise, reads more carefully.

Subscribe now

Part Two: What He Was Actually Looking At

Before we can understand the capital structure that Buffett built around these companies, we have to understand what the sogo shosha actually are — because this is where the intellectual foundation of the trade lives.

The common description — “general trading company” — is misleading in a specific and important way. In the 1980s and early 1990s, the sogo shosha did earn a significant portion of their income from acting as commercial intermediaries: buying from producers, selling to buyers, and earning a thin spread on the transaction. That model was always capital-light but scale-dependent, and it generated enormous volumes with modest margins. The criticism that accompanied it was fair: why would any large manufacturer continue paying a middleman when it could sell directly? Disintermediation, the logic went, would eventually hollow these businesses out.

What the analysts who wrote that narrative missed — and what Buffett saw when he read the handbook numbers carefully — is that the sogo shosha had been quietly transforming themselves for thirty years. Beginning in the 1990s, they began deploying their balance sheets not as intermediaries but as principals. Instead of earning a commission on a copper trade, Mitsubishi took an equity stake in the copper mine itself. Instead of financing a grain shipment for a fee, Marubeni bought grain farming operations outright. Instead of distributing food products to convenience stores for a margin, Itochu acquired FamilyMart, Japan’s second-largest convenience store chain, outright — via a takeover bid in 2020 — while simultaneously owning Nippon Access, the wholesaler that delivered inventory to FamilyMart’s stores three times a day.

This is the shift that matters. The trading commission revenue is high-volume and thin. The equity investment income — dividends, retained earnings, capital appreciation from hundreds of owned businesses — is the engine that has driven these companies’ actual earnings for the past two decades. When Buffett read the handbook and saw single-digit earnings multiples on businesses with 150-year operating histories, what he was actually observing was the market pricing a transformation it had not yet recognised. The companies were still being valued as commodity middlemen. The earnings being generated were from a permanent, diversified portfolio of owned businesses spanning every sector of the global economy.

Mitsui & Co.’s 2019 annual report — the document that would have corresponded to the numbers Buffett was reading in the handbook that year — lists ownership stakes in IHH Healthcare (33%), the largest hospital group in Asia; Penske Automotive (17%), one of North America’s largest automotive retailers; Roy Hill iron ore operations in Australia; LNG terminals across Asia-Pacific; and food businesses spanning grain trading, animal feed, protein production, and retail distribution. These are not the assets of a trading company. They are the assets of a diversified permanent capital vehicle — something that looks, in structure and philosophy, almost exactly like Berkshire Hathaway itself. Buffett noted this directly in his 2022 shareholder letter, writing that these companies “very successfully operate in a manner somewhat similar to Berkshire itself.”

Subscribe now

Part Three: The Architecture of the Moat

To appreciate why the carry trade Buffett was about to construct was not speculative but structural, you first have to understand why the cash flows underpinning it are not fragile. That requires understanding the barriers that protect them — and these barriers are unlike anything taught in a standard framework of competitive advantage.

Subscribe now

The First Wall: One Hundred and Fifty Years Cannot Be Purchased

There is a category of competitive advantage that capital cannot replicate, and it is the one that most investors systematically underweight: the advantage of accumulated operational tenure. The sogo shosha have been physically present in their core markets — not as salespeople, not as financial investors, but as operators — for between 100 and 165 years. That tenure has produced something that has no balance sheet entry and no market price: the trust of governments, communities, and commercial counterparts who have watched these organisations show up, deliver, and stay across multiple generations.

When Mitsui first signed iron ore offtake agreements with Australian miners in the early 1960s, it was not executing a financial transaction. It was beginning a relationship that now spans six decades of joint venture operations, shared infrastructure development, and government-level engagement in Canberra. The Pilbara iron ore operations that Mitsui participates in today were not built by writing a cheque. They were built by being there for sixty years, through commodity cycles, through political transitions, through the conversion of a remote Western Australian wilderness into the world’s largest iron ore province. No amount of capital can compress that history into a shorter timeline.

The United States government understood this problem when it attempted to engineer American equivalents of the sogo shosha. Congress passed the Export Trading Company Act of 1982 specifically to enable US companies to build Japanese-style general trading organisations. General Electric formed GE Trading Company. Sears formed Sears World Trade Organization. Both were defunct by the mid-1980s — not because they lacked capital, but because they could not purchase the one input the model requires most: time. You cannot buy a 150-year relationship. You have to show up every day for 150 years.

Subscribe now

The Second Wall: A Network That Gets Stronger as It Gets Larger

The second layer of the moat is structurally different from the first, and it is the one that makes the competitive position genuinely compound over time rather than merely persist. The sogo shosha generate what can be called primary information — intelligence that flows from operating simultaneously across twenty sectors and sixty countries, in real time, at commercial scale. This is categorically different from the secondary information that most market participants rely on: analyst reports, news wires, commodity indices, and consultant assessments.

When Mitsui’s grain desk in Iowa observes an unusual drought pattern forming across the US Midwest, that information does not stay on the grain desk. It flows — through internal systems, through regular communication between business units, through the natural conversations that happen when 45,000 employees work inside a single integrated organisation — to Mitsui’s LNG trading operations in Tokyo, to its protein businesses in Southeast Asia, and to its food retail operations in Japan. The grain desk is not charging the LNG desk for this intelligence. There is no internal transfer price. The information is a free byproduct of scale, and it creates immediate, actionable competitive advantage across apparently unrelated businesses.

Sumitomo Corporation described this dynamic directly in its own investor materials: “What makes this possible is our combination of diversity and scale. Since our companies operate in such a broad range of fields, we can diversify our exposure. And the scale of our business strengthens our ability to cope with even substantial risks. To manage these risks, we rely heavily on our worldwide network of offices, which provide a rich supply of primary information. This is a business model that obviously cannot be built overnight, and so it can be said that the barriers to new entrants into our industry is very high.”

The phrase “cannot be built overnight” understates the case considerably. It cannot be built in a decade. The information advantage is not a function of having offices in sixty countries — it is a function of those offices having operated in their markets for long enough that they have become embedded nodes in local commercial networks. The Mitsui office in Lagos is not merely an outpost. It is a known entity with established relationships with Nigerian government ministries, local businesses, logistics operators, and banking institutions, accumulated across decades of continuous presence. A new entrant opening an office in Lagos today would spend years building what Mitsui already has. By the time they had built it, Mitsui would have deepened it further. The gap does not close — it widens.

This is the property of network effects applied to physical, relationship-based commercial intelligence, and it is one of the most durable forms of competitive advantage in existence. Unlike a technology platform, it cannot be copied by writing code. Unlike a brand, it cannot be built by advertising spend. Unlike a patent, it does not expire. It accumulates silently, year after year, as a byproduct of simply continuing to operate.

Subscribe now

The Third Wall: Regulatory Positions That Were Granted Under Regimes That No Longer Exist

The third layer is the most permanent of all, and it is the one that modern markets are most systematically bad at pricing. The sogo shosha hold resource concessions, trading licences, and operating permits across dozens of jurisdictions — many of which were awarded under regulatory frameworks that no longer exist in the same form and that cannot be replicated under current conditions.

A liquefied natural gas concession awarded to Mitsui in Brunei in 1972 was granted under the regulatory expectations of 1972: different environmental standards, different community consultation requirements, different government priorities, and a geopolitical environment in which Japan’s need for energy security made it a particularly welcome partner for resource-rich Southeast Asian nations. Applying for an equivalent concession today would require navigating environmental impact assessment processes, community approval mechanisms, government procurement rules, and international partnership requirements that bear no resemblance to the conditions of 1972. In practice, for the most valuable resources in the most sensitive geographies, the regulatory path to a new concession is not merely difficult — it is functionally closed.

The sogo shosha’s position in these markets is therefore not just competitive. It is structurally exclusive. The assets they hold cannot be replicated by new entrants because the conditions under which those assets were obtained no longer exist. This is the commercial equivalent of a building permit issued before a historic preservation district was established: the holder has rights that the regulatory system will never issue again.

When you add all three walls together — the unreplicable operational tenure, the self-reinforcing information network, and the regulatory positions granted under extinct frameworks — the resulting structure is a moat that does not merely protect existing cash flows. It actively deepens over time, becoming more durable with every passing year of continued operation. This is what Buffett was holding in his mind when he told Nikkei Asia in 2023 that he expected Berkshire to hold these stakes “for the next 50 years.” He was not expressing sentiment. He was describing the logical conclusion of a structural analysis: barriers this deep and this compounding do not erode.

Subscribe now

Part Four: The Capital Structure — Where the Genius Lives

In September 2019, one month before Berkshire began quietly accumulating shares in the five trading houses, Berkshire Hathaway’s treasury team executed a bond issuance in the Japanese domestic market that would become the financing instrument for the most elegant capital structure Buffett had ever assembled.

The issuance, documented in SEC filings, was ¥430 billion — approximately $4.04 billion at prevailing exchange rates — structured across six tranches with maturities of 5, 7, 10, 15, 20, and 30 years. The coupons, as filed with the SEC, ranged from 0.17% on the 5-year tranche to 1.10% on the 30-year tranche. This was, at the time, a record-sized yen bond issuance by any foreign company. It was also, in retrospect, the opening move of a trade whose architecture would not be fully understood by the market for several years.

S&P Global’s capital markets commentary noted the precise reason why Japanese institutional investors devoured these bonds: Berkshire’s 10-year tranche offered a yield of approximately +0.44%, compared to roughly -0.25% on Japanese Government Bonds of the same maturity. Japanese insurance companies and pension funds, trapped in a negative yield environment by the Bank of Japan’s extraordinary monetary policy, needed positive yield from names they trusted. Berkshire, with its AA credit rating and global recognition, was exactly that name. The supply and demand aligned so cleanly that the deal was oversubscribed many times over.

Berkshire’s yen bond programme continued through subsequent years, with documented issuances of ¥196 billion in April 2020 — the 5-year tranche at 0.674%, the 10-year at 1.002% — and ¥160 billion in April 2021, with the 5-year tranche at a remarkable 0.173% and the 10-year at 0.437%. By 2023, Berkshire had accumulated approximately ¥1.9 trillion in yen-denominated bonds outstanding, representing roughly $13 billion at prevailing exchange rates, at a blended average coupon that can be conservatively estimated at approximately 0.5% per annum.

Now set that number beside the starting point of the equity investment.

By the time Berkshire disclosed its positions in August 2020, it had accumulated approximately 5% stakes in each of the five trading houses for a total cost of approximately $6.7 billion. The yen bond proceeds covered approximately $4.0 to $4.5 billion of that initial cost. Berkshire’s own equity capital deployed — the residual between the total position and the bond financing — was approximately $2.7 billion.

The capital structure at the moment of initial disclosure was therefore:

• Bond-financed portion: approximately $4.0 billion, or 60% of total investment

• Berkshire’s own equity: approximately $2.7 billion, or 40% of total investment

• Blended cost of the bond financing: approximately 0.5% per annum

• Dividend yield on the equity position at purchase price: approximately 5–6% per annum

From this structure, a specific and calculable arithmetic follows. Annual dividend income from the $6.7 billion position at a 5% yield: approximately $335 million. Annual interest cost on $4.0 billion of bonds at 0.5%: approximately $20 million. Net annual cash income to Berkshire after debt service: approximately $315 million. Return on Berkshire’s own $2.7 billion of equity, from the income stream alone, in Year 1: approximately 29% per annum.

This number requires no assumptions about future growth, no forecast of commodity prices, no view on the macroeconomy. It is the arithmetic product of the capital structure as it existed on the day the position was disclosed. The bonds were outstanding. The dividends were being paid. The spread was locked. The 29% was already there.

Subscribe now

Part Five: The Compounding Engine — When a Fixed Cost Meets a Growing Income

Twenty-nine percent in Year 1 is not the end of this analysis. It is the beginning of a separate and more powerful phenomenon — one that Buffett had seen operate over thirty years in a different context and recognised instantly when he saw the conditions for it reappearing in Japan.

The sogo shosha, between fiscal years 2021 and 2025, grew their dividends per share at an average annual rate of 19.2%. This rate is partially elevated by the post-COVID commodity supercycle, and a long-run sustainable rate is more conservatively estimated at 8 to 12 percent. But it establishes the direction and magnitude of the underlying compounding dynamic.

Now apply that dividend growth to the capital structure described above, holding the bond cost constant — because the bond coupons, fixed at issuance, do not participate in any growth. They are permanent liabilities with permanent, fixed costs. The equity income above them grows freely:

(Net cash income uses the earnings yield on the full position, not dividend yield alone, to reflect the full economic return Berkshire’s ownership stake generates. Bond interest at 0.5% on $4.0B bond principal held constant.)

Note that 58.6% is on 1 year, without including previous year dividend payouts along with the capital growth of the equity stake. This investment case is a rare one where the compounding economics is created by financial engineering instead of through a compounding machine (business) which Berkshire have traditionally done so.

The mechanism at work here is what engineers call a fixed-cost leverage structure. The bond interest is a fixed denominator. The earnings flowing from the equity position — growing at 19% annually — have no obligation to share any of their growth with the bondholders. Every dollar of dividend growth above the fixed bond cost flows directly to Berkshire’s equity holders, compounding the return on Berkshire’s own $2.7 billion at an accelerating rate. By Year 5, Berkshire is earning $1.58 billion annually from a $2.7 billion equity investment — a yield on original cost of 58.6%.

Buffett had seen exactly this dynamic before. In his 2022 shareholder letter, he described the arithmetic of Berkshire’s Coca-Cola investment with characteristic plainness. Berkshire completed its purchase of Coke shares in 1994 for a total cost of $1.3 billion. In that first year, Coke paid Berkshire $75 million in cash dividends — a yield of approximately 5.8% on cost. Coke grew its dividend every year thereafter. By 2022, the annual cash dividend check had grown to $704 million — a yield of 54% on Berkshire’s original $1.3 billion investment, received annually, in perpetuity.

The structural parallel is precise. In both cases, Buffett identified a durable cash flow stream attached to a business with deep competitive protection, acquired at a price that gave him an attractive initial yield, and then simply waited while the yield on his original cost grew toward something that bears no relationship to any standard measure of investment return.

The difference between the two trades is the timeline. Berkshire held Coca-Cola for nearly thirty years before the yield on original cost crossed 50%. In Japan, because Buffett introduced fixed-rate yen bond leverage into the capital structure, the mathematics of yield-on-original-cost acceleration are compressed dramatically. The 58% yield on equity materialises in five years rather than thirty — not because the underlying business is growing faster than Coke did, but because Berkshire’s own equity denominator is smaller relative to the total position, so each dollar of income growth translates into a larger percentage return on that smaller equity base.

Subscribe now

Charlie Munger understood this compression immediately. In his November 2023 interview with the Acquired podcast, seven weeks before his death at age 99, he described the trade with characteristic directness: “Something like that — if you’re as smart as Warren Buffett, maybe two, three times a century, you had an idea like that. It was awfully easy money. It was like having God just opening a chest and just pouring money into it.” He then identified the precise source of the advantage: “These trading companies were really entrenched, old companies, and they had all these cheap copper mines and rubber plantations, and so you could borrow easily. But Berkshire, with its credit, could. The only way you could get it was to be very patient and just pick away at little pieces at a time. It took forever.”

The phrase “took forever” matters. The position required approximately a year of quiet accumulation — buying in small amounts, in the open market, in yen, without causing visible price movements that would alert the market and push prices higher before Berkshire had built the full position. This patience is not a personality trait. It is a structural requirement of the trade: the returns are only available to someone capable of accumulating billions of dollars of shares over twelve months without advertising what they are doing.

Subscribe now

Part Six: The Yen — The Dimension the Market Ignored

There is a third layer to this trade that is almost never discussed in Western financial commentary, and it is the layer that makes the entire structure genuinely elegant rather than merely clever.

Berkshire borrowed in yen. Berkshire received dividends in yen. Berkshire’s liabilities and its income are denominated in the same currency. This is not a hedging strategy. It is a structural fact that eliminates foreign exchange risk from the spread — the gap between the dividend yield and the bond cost — entirely. Whatever happens to the yen relative to the dollar has equal and opposite effects on the income side and the liability side of the trade. They cancel.

The yen depreciated significantly from approximately ¥107 per dollar in 2019 to approximately ¥150 per dollar by 2023 — a weakening of roughly 40 percent. In dollar terms, a yen-denominated income stream that stayed flat in yen would have lost 40% of its dollar value over that period. But the yen-denominated bond liability declined by exactly the same proportion. Berkshire owed yen to its Japanese bondholders, and those yen were worth less in dollars. The liability shrank in dollar terms at exactly the same rate as the income shrank. The net dollar spread — income minus debt service — remained stable.

More than stable, in fact. Berkshire’s 2024 annual letter disclosed that by the end of that year, Berkshire had accumulated $850 million in after-tax gains from the dollar’s appreciation against the yen. This is the accounting recognition of the net effect: Berkshire’s yen liabilities, translated back to dollars for reporting purposes, had declined in value by more than Berkshire’s yen-denominated equity investments had declined, because the equity investments had simultaneously been appreciating in yen terms as the sogo shosha’s earnings and book values grew. The currency movement, instead of hurting the trade, added to it.

Inflation = Asset value increases whilst currency value decreases — that is what happened to Japan, whereby in 2019-2024, currency depreciated 40%. It makes much more sense to compound that capital in a Yen-denominated business than hold cash (Yen)

This outcome was not guaranteed when Buffett entered in 2019. He has said explicitly that he does not take views on foreign exchange. What he did instead was construct a structure in which he did not need a view: by matching the currency of his liabilities to the currency of his income, he removed the foreign exchange question from the equation entirely and let the fundamental economics of the trade — the spread between a sub-one-percent borrowing cost and a compounding five-to-seven percent equity yield — drive the outcome. The currency became neutral by design, not by luck.

Subscribe now

Part Seven: The Complete Picture — Three Layers Stacked

The income arithmetic — $784 million in net annual cash in Year 1, growing toward $1.58 billion by Year 5, against $2.7 billion of Berkshire’s own equity — captures the operating engine of the trade. But it is only one of three layers that constitute the total return.

The second layer is capital appreciation. Berkshire’s initial $6.7 billion position, accumulated at prices averaging 0.6 to 0.8 times book value, was worth approximately $23.5 billion by the end of 2024. The $16.8 billion in unrealised capital gains represents the market’s slow recognition of what Buffett observed in the handbook in 2019: that these businesses were being priced as commodity middlemen when they were, in fact, permanent capital vehicles with global reach and 150-year operating histories. The Tokyo Stock Exchange’s governance reform — which began pressuring companies trading below book value to improve their returns or return capital — provided the mechanism through which this recognition was forced. Buybacks accelerated. Dividends grew. The P/B multiple expanded from below one to above one as the market processed what was always there.

The Lindey effect: the future life expectancy of non-perishable things (like ideas, technologies, or books) is proportional to their current age. Simply put — the longer something has already survived, the longer it is expected to endure in the future.

The third layer has no dollar value yet, and may be the most valuable of all: the strategic option embedded in the relationship itself. Buffett told Nikkei Asia in 2023 that Berkshire wanted to be “the first port of call” for any of the five companies thinking of making a large acquisition or seeking a co-investment partner. This is not a casual expression of goodwill. It is a description of Berkshire’s position as the preferred permanent-capital co-investor for five of the most globally connected commercial organisations on earth — organisations that collectively operate in more than 60 countries, see primary deal flow across every sector of the global economy, and have established the kind of trust with governments, resource owners, and local businesses that takes 150 years to accumulate.

Every co-investment opportunity that flows from those relationships in the next fifty years comes to Berkshire first, with no competitive process, at terms negotiated between parties who have already demonstrated mutual trust. The option has no cost basis. It was acquired as a consequence of buying equity at prices that were already attractive on their own terms. This is what Buffett meant when he wrote, in his 2024 shareholder letter, that “Berkshire will find other ways to work productively with the five companies in the future” — and why he specified that Greg Abel and his eventual successors would be holding these positions for “many decades.”

Subscribe now

Part Eight: Why Only Berkshire

Every element of this trade was, in theory, visible to every institutional investor in the world in 2019. The handbook was publicly available. The financial statements were filed. The yen bond market was open. The arbitrage between near-zero Japanese interest rates and 5-7% dividend yields from century-old conglomerates trading below book value required no proprietary information to identify. And yet no one executed it. Understanding why illuminates something important about what this trade actually was.

To execute the yen bond component of this arbitrage at the interest rates Berkshire obtained, you needed Berkshire’s AA credit rating. A lower-rated issuer in the Japanese market would have paid 200 to 300 basis points more — collapsing the spread from 4.5% to something between 1.5% and 2.5%, which is respectable but not the once-in-a-century opportunity Munger described. The AA rating is itself the product of sixty years of Berkshire’s capital discipline, its permanent equity capital base, its reputation for honouring obligations, and its insurance float — none of which can be manufactured quickly by an institution that wants to replicate the trade.

To accumulate $13 billion in illiquid Japanese equities over multiple years without announcing your position or moving prices materially against yourself, you needed Berkshire’s patience and its ability to deploy capital slowly without quarterly performance pressure. A hedge fund building the same position would face redemption pressure, performance benchmarking, and the structural incentive to concentrate quickly rather than accumulate gradually. By the time a hedge fund had deployed enough capital for the position to matter to its returns, it would have already moved the market against itself.

And to sit for five, ten, or fifty years without selling — which is what makes the yield-on-original-cost mathematics compound to its most extraordinary values — you needed Berkshire’s institutional commitment to permanent capital. Berkshire does not manage money for clients who can withdraw it. It does not answer to a limited partner advisory board that reviews the portfolio annually. It answers to shareholders who understand, because Buffett has explained it in every annual letter for sixty years, that Berkshire’s preferred holding period is forever. This institutional structure is itself a competitive advantage in the execution of long-duration, compounding trades.

Munger put the constraint precisely: “Other companies couldn’t get it. But Berkshire, with its credit, could.” The trade was available only to an institution that was simultaneously trusted by Japanese institutional investors as a bond issuer, trusted by markets as a long-term holder of equity, and willing to spend a year accumulating a position quietly before revealing it. That is a very short list. It may be a list of one.

Subscribe now

Part Nine: The Lesson

There is a tendency, when studying great investment trades, to focus on the outcome — the return figures, the capital appreciation, the validation by subsequent events — and to extract from that outcome a lesson that sounds profound but is practically useless: find cheap assets. Be patient. Think long-term.

The lesson from this particular trade is more specific, and more useful.

What Buffett identified in 2019 was not merely a cheap asset or an overlooked market. He identified a convergence of three conditions that independently were interesting and jointly were extraordinary: a structural moat so deep that the cash flows it protected could be treated as near-certain; a financing opportunity available only to an institution with his specific combination of credit rating, capital base, and patience; and an accounting regime that was systematically understating the value of the assets behind that moat. Any one of these conditions is worth investigating. All three simultaneously — in a market that Western professional investors had collectively abandoned — is what produces the outcome that Munger described as a gift from God.

Subscribe now

The sogo shosha’s competitive position had been building for 150 years. The yen bond arbitrage had been available, in principle, since the Bank of Japan adopted its near-zero rate policy in the 1990s. The accounting understatement of these companies’ assets — their fully depreciated resource concessions, their cross-holdings carried at historical cost, their relationship networks priced at zero — had been sitting in plain sight in annual reports for decades. None of these conditions was secret. All of them required the same thing to be acted upon: someone willing to read carefully, think patiently, and understand that the most powerful investment structures are usually not discovered through sophisticated modelling but through the simple, disciplined act of turning every page.

At 88 years old, with $700 billion under management and the resources to hire any analyst in the world, Warren Buffett found a $23 billion idea in a handbook. He found it because he was the only person of his scale who was still willing to read the handbook.

This paper was prepared as an analytical study of a completed investment structure. All data cited refers to historical figures. Nothing in this paper constitutes investment advice or a recommendation regarding any security. The bond coupon and tranche data cited in this paper is drawn from SEC-filed Form 424B5 prospectuses for Berkshire Hathaway’s yen bond issuances in September 2019, April 2020, April 2021, and January 2022, all filed under CIK 0001067983. The September 2019 inaugural issuance of ¥430 billion is confirmed by S&P Global Capital Markets Weekly commentary dated September 2019. The $23.5 billion valuation of Berkshire’s sogo shosha stakes is from Berkshire’s 2024 annual report. The 19.2% average annual dividend per share growth rate for 2021–2025 is sourced from public dividend disclosures across all five companies. The $850 million after-tax foreign exchange gain is from Berkshire’s 2024 shareholder letter. ROE data for each company is drawn from their respective IFRS annual reports, with Mitsui’s five-year average ROE of 14.4% confirmed by Investing.com data citing fiscal years ending March 2021–2025. Charlie Munger’s quotations are from the Acquired podcast interview, November 2023. Warren Buffett’s quotations are from the 2025 Berkshire annual meeting, the 2024 Berkshire shareholder letter, and the April 2023 Nikkei Asia interview conducted during his Tokyo visit.

If you like what you read, check out my other writings 👇

PART I: WHAT COPPER ACTUALLY IS, AND WHY IT CANNOT BE SUBSTITUTED AWAY

The Invisible Infrastructure of Civilisation

Before any valuation, you need to understand why copper is not a commodity in the same sense as soybeans or newsprint. Soybeans can be substituted with other oil seeds; newsprint has been replaced by digital media. Copper has not been replaced by anything in two thousand years of commercial use, and there is no credible candidate to replace it in the next two decades — precisely the period when demand is projected to grow most aggressively.

Copper is the second-best electrical conductor among practical metals, behind only silver. Silver is 70 times rarer and priced accordingly. Every wire that carries electricity in your home, your car, your phone, and the power grid connecting your city uses copper because nothing else performs the same function at a comparable cost. Copper is also the best thermal conductor in the price range that makes it useful for heat exchangers, motors, and industrial machinery. It does not corrode in water the way iron does. It can be drawn into extremely thin wire without breaking. It can be alloyed with zinc to make brass or with tin to make bronze. It has natural antimicrobial properties that make it useful in medical and food environments. The list of applications is not a marketing document — it is an engineering reality that has been tested and refined over two millennia.

An electric vehicle contains approximately 83 kilograms of copper, compared to roughly 23 kilograms in a conventional internal combustion vehicle. The difference — 60 kilograms per vehicle — is driven by the copper in the electric motor windings, the battery pack connections, the onboard charger, the power electronics, and the wiring harness. A single offshore wind turbine requires approximately 4 to 15 tonnes of copper in its windings, cables, and transformer connections. A solar farm requires 5 to 6 tonnes of copper per megawatt of generating capacity. A grid-scale battery storage facility requires copper in its busbars, cables, and control systems. Artificial intelligence data centers, which require dense, reliable power delivery to thousands of processors, are estimated by S&P Global to need 250,000 to 550,000 tonnes of copper per year by 2030 — entirely additional demand that did not exist five years ago.

None of this is a forecast about the future. It is a description of specifications that are already locked into products being manufactured today. The EV that rolls off a factory line tomorrow contains its 83 kilograms of copper already. The offshore wind turbine ordered last year is being built with its 15 tonnes of copper now. The demand exists. The question is only whether the supply will be there to meet it.

Subscribe now

The Discovery Problem Nobody Talks About

Here is the single most important structural fact about copper that financial analysts consistently under-weight, and that business owners who have spent time in the mining industry understand immediately: copper deposits are extraordinarily hard to find, and the industry has effectively stopped finding them.

S&P Global Market Intelligence has tracked major copper discoveries for decades. Between 1990 and 2013 — a period of 23 years — the global mining industry made 225 major copper discoveries. In the decade from 2013 to 2023, the same industry made exactly 14 major discoveries. Not 14 per year. Fourteen total, in a decade. This is not a temporary statistical fluctuation. It reflects a fundamental reality: the accessible, high-grade porphyry copper deposits at shallow depths in politically stable jurisdictions were found and developed in the 20th century. What remains is harder to find, deeper, lower-grade, and more likely to sit under contested land or within environmentally sensitive terrain. The IEA’s Global Critical Minerals Outlook 2025 projects a 30 percent supply shortfall by 2035 under the Stated Policies Scenario — meaning the most conservative, business-as-usual demand case still produces a 30 percent deficit. The IEA’s executive director Fatih Birol stated plainly at the UK Critical Minerals Summit in late 2025: “Copper is the one we’re really concerned about. This is a really challenging mineral to ramp up supply quickly.”

Subscribe now

And from discovery to production, a new copper mine now takes an average of 17 years. That is not a guess — it is the documented average calculated from IEA data and cited separately by S&P Global, Sprott, and multiple independent analyses. A deposit found today does not produce a single pound of commercial copper until 2042. Every pound of copper from an existing, operating mine is therefore more valuable today than it would be in a world where new mines could be brought online quickly. It cannot be replaced quickly. The ore bodies that already exist — already drilled, already permitted, already producing — are, in a very precise sense, irreplaceable within any commercially relevant time horizon.

Subscribe now

PART II: THE OWNER’S VALUATION — POUNDS IN THE GROUND AT TODAY’S PRICE

The Arithmetic a Geologist Understands and a Spreadsheet Misses

The owner’s valuation methodology for a copper mining company is not complicated. What makes it powerful is what it excludes: no forecast of future copper prices, no assumption about future demand growth, no model of what the stock “should” trade at relative to its peers. Just the existing certified reserve, the current cost to extract it, and the current price the market pays.

The four inputs:

The first input is the certified reserve base — proven and probable mineral reserves of copper, measured in pounds or tonnes of contained copper, certified by independent qualified persons under NI 43-101 (Canada), JORC (Australia), or SEC Industry Guide 7 (United States). These are not estimates or guesses. They are statistically confident calculations based on drill hole density, grade interpolation, and mine engineering, published in annual technical reports audited by independent geological firms.

The second input is the C1 cash cost per pound — the direct cost of mining, milling, and smelting copper to produce a saleable concentrate or refined metal, net of any gold or silver byproduct credits. For open-pit porphyry copper operations — the predominant mine type for large-scale producers — industry-disclosed C1 cash costs typically range from $1.50 to $2.00 per pound, based on disclosed All-In Sustaining Cost (AISC) figures from operating company annual reports. Lower-cost operations like Codelco’s El Teniente or Freeport’s Grasberg can operate below $1.50/lb. Higher-cost, lower-grade open-pit operations approach $2.50/lb before sustaining capital.

The third input is the current copper spot price. As of early 2026, copper was trading in the range of $4.20 to $4.50 per pound on the COMEX, with an all-time intraday high of $6.61 per pound reached on January 29, 2026 during a speculative surge. A business owner uses the mid-cycle, steady-state price — not the peak, not the trough, but what the market has consistently cleared at in recent years. Using $4.25/lb as a reference is conservative by 2026 standards.

The fourth input is the net margin per pound: price minus cash cost. At $4.25/lb price and $1.75/lb cost, the net margin is $2.50/lb.

The calculation for Freeport-McMoRan: Freeport disclosed approximately 104 billion pounds of copper in proven and probable reserves in its 2023 10-K filing. Apply $2.50/lb net margin. The total gross operating income embedded in the existing reserve base is approximately $260 billion — calculated entirely at current prices, current costs, from mines already in production, without a single new exploration dollar. Freeport’s market capitalization as of early 2026 was approximately $55 to $65 billion. The arithmetic says: you are paying 25 cents for every dollar of gross operating income embedded in a certified, producing reserve base. The stock’s P/B ratio of approximately 4.18x makes it look expensive on a book value basis. But the book value has no relationship to the reserve value — it is the depreciated cost of building the mine, not the value of the copper in the ground.

This is the gap. Not a forecast gap. Not a DCF gap. A gap between what is physically in the ground at current prices and what the balance sheet acknowledges. It is $260 billion versus a $60 billion market cap. The market is not pricing it literally — other factors, including discount rates, extraction sequencing, and the 25-year extraction timeline, explain much of the divergence. But the fundamental point holds: the accounting records the cost of building the mine, and the owner counts the value of what the mine contains.

For smaller producers, the gap is proportionally more acute and more exploitable. This is the real opportunity.

Subscribe now

PART III: GAAP AND THE MINE THAT AGES BACKWARDS

What the Accountant Sees

Under US GAAP (ASC 360, PP&E with extractive industry guidance) and IFRS (IFRS 6 for exploration, IAS 16 for development assets), mineral reserves are never recognized as assets at fair value. The ore body — the billions of pounds of copper that an independent geological firm has certified through years of drilling and analysis — does not appear on the balance sheet as an asset. What appears instead is the cost of developing access to that ore body: the exploration drilling, the shaft sinking, the open pit earthworks, the processing plant, the infrastructure. These costs are capitalized as PP&E and then depleted against production using the Units of Production method.

This creates an accounting paradox that every mining analyst knows but that financial models fail to handle properly. As production continues, the book value of the mine declines — because accumulated depletion charges reduce the carrying value of the capitalized development costs. But the ore body does not necessarily become less valuable as production progresses. If copper prices rise, the remaining ore becomes worth more per pound, not less. The mine is getting “cheaper” on the balance sheet precisely when the copper it contains may be getting more valuable in the market.

The Grasberg mine in Papua, Indonesia, illustrates this in almost comical terms. Freeport-McMoRan and its predecessor entities began developing Grasberg in the late 1960s and 1970s, at copper prices that were a fraction of current levels. The ore body’s current proven and probable reserves, as of December 31, 2022, included 30.8 billion pounds of copper and 26.3 million ounces of gold, per the Grasberg Wikipedia disclosure sourced from Freeport’s own filings. The current in-ground value of that copper at $4.25/lb — before deducting a single dollar of production cost — is approximately $131 billion. The current in-ground value of the gold at $2,300/oz is approximately $60 billion. Combined: roughly $191 billion in in-ground metal value at a single mine complex.

Freeport’s 48.77% interest in PT Freeport Indonesia (the operating entity) — secured through a decade-long negotiation with the Indonesian government that concluded in 2018, when Indonesia required a 51.23% divestment in exchange for a 20-year permit extension to 2041 — carries a net book value on Freeport’s consolidated balance sheet that is a fraction of the in-ground metal value. It reflects decades of partially depleted development cost at 1970s through 1990s capital expenditure levels. The mine has been physically depreciating on paper for 50 years. The ore body it accesses has been appreciating economically as copper prices rose from $0.60/lb in the 1990s to $4.25/lb today.

That is the GAAP paradox in its most extreme form. The accountant says the mine is worth less than it used to be. The copper market says it is worth more.

Subscribe now

PART IV: THE MARKET’S PREFERRED TOOLS AND WHY THEY FAIL HERE

Why Earnings Multiples Cannot Measure a Mine

When a sell-side analyst publishes a price target on a copper producer, the mechanics are typically straightforward: take the forecast copper price for the next 12 months, build a revenue model, subtract estimated operating costs to get EBITDA, apply a sector multiple to EBITDA, and arrive at a valuation. This is how most industrial companies are priced, and in most industrial companies, it works reasonably well because the company’s primary assets — equipment, inventory, customer relationships — have some relationship to current earnings.

For copper miners, the methodology introduces a structural error that compounds with every step. The copper price forecast is the most uncertain variable in the model, and it is also the variable to which the output is most sensitive. A $0.50/lb error in the copper price assumption — well within the range of error in any 12-month commodity forecast — translates, at Freeport’s production scale of approximately 4 billion pounds per year, into a $2 billion error in annual revenue and roughly a $1.5 billion error in EBITDA. Apply a 10x EBITDA multiple and you have a $15 billion error in the valuation from a single half-dollar assumption on the copper price. The model’s precision — to two decimal places on the price target — vastly overstates its accuracy.

At a cycle peak, when copper is at $4.50/lb, the earnings model produces a low-looking P/E because earnings have surged. Investors believe the stock is cheap and buy. But the earnings at $4.50/lb are not sustainable at the same cost structure — when copper falls back to $3.00/lb, earnings collapse, and investors who bought the “cheap” P/E discover they were buying at a peak. The earnings multiple was a perfect contrary indicator.

At a cycle trough, when copper is at $2.00/lb, the earnings model shows losses or near-zero earnings. P/E is infinite or irrelevant. The consensus is bearish. The stock is sold. But an owner looking at the certified reserve base — 3 billion pounds of copper at a production cost of $1.75/lb, selling at $2.00/lb — sees a thin margin on a large volume, from an asset that will still be in the ground when prices recover. The trough earnings tell you nothing about the asset’s value. They tell you about the current commodity price, which is a temporary state that has historically lasted 18 to 36 months before reverting.

Subscribe now

PART V: STORIES FROM THE GROUND — THREE CASE STUDIES IN COPPER VALUE ARBITRAGE

Case Study 1: The Mountain That Swallowed Freeport — The 2015 Crisis and What It Revealed

The year 2015 was, in many ways, the darkest in Freeport-McMoRan’s corporate history. In 2012 and 2013, at the height of the previous copper bull run, Freeport had made two large acquisitions — Plains Exploration & Production and McMoRan Exploration — for a combined $9 billion, adding oil and gas assets to its copper and gold portfolio. The acquisitions were timed at exactly the wrong moment, just as the commodity cycle was turning. By 2015, copper had fallen from its 2011 peak of $4.65/lb to below $2.20/lb. Oil had crashed simultaneously, compounding the damage to the newly acquired energy business. Freeport’s total debt had ballooned to approximately $20 billion. The stock, which had traded above $55 per share in 2011, fell to approximately $3.50 per share in January 2016. The market capitalization had collapsed from roughly $50 billion at its peak to approximately $5 billion. The company was being priced as if it might not survive.

At the nadir of $3.52 per share in early 2016, Freeport’s total enterprise value — market cap plus net debt — was roughly $25 billion. The company’s proven and probable copper reserves at that time were approximately 100+ billion pounds. Apply the owner’s arithmetic: at $2.00/lb copper and $1.75/lb cash cost, the net margin per pound was $0.25. Total operating income embedded in the reserve base at trough prices: approximately $25 billion. The market was pricing the company at exactly the cash-on-cash value of its entire certified reserve base at the absolute bottom of the copper price cycle — and assigning zero residual value to the oil and gas business, the gold production at Grasberg, the corporate franchise, the management team, and the future exploration inventory.

What did the ore body say in response? Nothing. It was unchanged. The same mountains in Papua, Arizona, New Mexico, and Peru contained the same copper they had always contained. The ore grades certified by independent engineers had not moved. The infrastructure to mine and process the ore was intact and operational. The only thing that had changed was the market price of copper — a temporary condition driven by Chinese growth concerns, a surging US dollar, and commodity fund liquidation, none of which had any bearing on the geology of the Grasberg district or the Morenci open pit in Arizona.

Freeport’s survival required asset sales, equity dilution, and the painful disposal of the oil and gas business at distressed prices. These were real costs imposed by overleveraging at cycle peak — a lesson in capital structure discipline that has nothing to do with asset quality. The copper ore bodies were not impaired. By 2022, when copper reached $4.65/lb again, Freeport’s stock had recovered to over $50 per share — a 14x return from the $3.52 trough. The copper was always there. The market had priced it as if it wasn’t.

The lesson here is not simply that copper prices recover — they do, and they always have. The lesson is more specific: the ore body, the physical asset, was never worth $3.50 per share. Not even close. The balance sheet, battered by debt and impairments, said one thing. The pounds in the ground said another. An owner who ignored the earnings and debt spiral and looked only at the certified reserves, the C1 cash cost, and the physical impossibility of replacing 100 billion pounds of tier-one copper geology in the current market — that owner was buying something real at a price that reflected temporary fear rather than permanent impairment.

Subscribe now

Case Study 2: Taseko’s Gibraltar Mine — The Canadian Porphyry That the Market Forgot

Taseko Mines Limited (NYSE: TGB) is not the kind of company that attracts front-page coverage. It operates the Gibraltar Mine near Williams Lake in British Columbia — a 75%-owned open-pit porphyry copper operation that has been producing since 1972. Gibraltar is not glamorous. It is not in a fashionable jurisdiction like Chile or the Democratic Republic of Congo. It is a British Columbia mine with 50 years of operating history, a reliable workforce, existing road and power infrastructure, and an ore body that, in 2022, was expanded through geological reanalysis to 706 million tonnes at 0.25% copper — yielding 3.0 billion pounds of recoverable copper with a 23-year mine life and average annual production of approximately 129 million pounds.

Pause on those numbers for a moment. Three billion pounds of certified, recoverable copper in a working mine. Twenty-three years of reserve life from infrastructure that already exists. One hundred and twenty-nine million pounds per year of production at operating costs that Taseko has publicly disclosed at approximately $2.00 to $2.30 per pound net of molybdenum byproduct credits.

Apply the owner’s arithmetic at $4.25/lb copper: 129 million pounds per year at approximately $2.00/lb net margin equals approximately $258 million per year in gross operating income from the Gibraltar mine alone. Three billion pounds at $2.00/lb net margin embedded in the certified reserve base equals approximately $6 billion in total gross operating income over the mine’s life, at current prices and current costs. Taseko’s market capitalization as of mid-2025 was approximately $400 to $500 million. The market was pricing the company at roughly 7 to 8 cents on the dollar of the operating income embedded in its certified reserve base, at today’s prices, at today’s costs.

A business owner buying 100% of Taseko’s 75% interest in Gibraltar at the mid-2025 market price of approximately $375 million was paying, on a per-pound basis, approximately $0.125 per pound of recoverable copper. At $0.125/lb for a functioning mine, the market is paying less than the cost of replicating the bare ore body discovery, and is assigning zero value to 50 years of operational infrastructure, the tailings management facility, the community relationships, and the institutional knowledge of a workforce that has been running this specific mine for half a century.

Subscribe now

Case Study 3: Hudbay’s Constancia and the Quiet Power of a Peruvian Porphyry

Hudbay Minerals (NYSE: HBM) tells a slightly different story — one about how an ore body’s disclosed value interacts with a market that does not look past the income statement.

Hudbay operates its Constancia open-pit copper-molybdenum mine in the Cusco region of Peru, along with the Copper Mountain mine in British Columbia (acquired June 2023) and the Snow Lake underground operations in Manitoba. Constancia is the anchor asset. Its total proven and probable mineral reserves as of late 2023 included approximately 1.5 million tonnes of copper across 548 million tonnes of ore at 0.27% copper grade — or roughly 3.3 billion pounds of contained copper at Constancia alone.

The Copper Mountain acquisition of June 2023 added 346 million tonnes at 0.25% copper — approximately 1.9 billion pounds of additional contained copper, with a 21-year mine life. Hudbay acquired Copper Mountain at a time when the previous operator’s cost structure had been elevated; Hudbay’s own disclosures indicated that their operational optimization at the site — replicating the processes they had implemented at Constancia — was expected to reduce Copper Mountain’s cash costs by approximately 50% over the first 10 years. A 50% reduction in cash costs from a certified reserve base means the operating income per pound roughly doubles, without changing a single tonne of the ore body.

The combined Hudbay reserve base — Constancia plus Copper Mountain plus Snow Lake — approached 6 to 7 billion pounds of copper equivalent in proven and probable reserves as of 2023 filings. At $4.25/lb copper and $1.75/lb blended cash cost, the net margin per pound across the portfolio was approximately $2.50/lb. Total gross operating income embedded in the reserve base: approximately $15 to $17.5 billion.

Hudbay’s market capitalization ranged from approximately $2.5 billion to $3.5 billion in 2024, with an enterprise value somewhat higher due to net debt. The market was pricing Hudbay at 15 to 20 cents on the dollar of the operating income embedded in its certified reserves, at current prices and current costs. The balance sheet explanation for this discount is familiar: the reserves are not assets under GAAP. What appears on Hudbay’s balance sheet is the net book value of mines at historical development cost less accumulated depletion — a number that bears no relationship to the value of 6 billion pounds of copper in already-producing mines.

The story at Copper Mountain also illustrates something important about how the market prices operational challenges. The mine had been struggling with production variability and elevated costs under its previous operator, generating negative headlines and disappointing quarterly results. The market, anchored to trailing earnings and recent production metrics, priced the asset accordingly. Hudbay’s acquisition thesis was simpler: this is a porphyry copper deposit with 21 years of certified reserve life in a tier-one Canadian jurisdiction, and its operating problems are fixable through standard operational discipline. The ore body does not care about its operating history. It cares about the C1 cost achievable from its geology, which Hudbay’s own Constancia experience gave them reason to believe could be substantially lower than the previous operator had achieved.

Subscribe now

PART VI: CONFIRMED CATALYSTS — EVENTS ALREADY IN MOTION

The 17-Year Lock

Every confirmed catalyst for copper ultimately derives from one physical fact: the discovery-to-production timeline of 17 years creates a rigid upper bound on how much new copper supply can enter the market between now and the early 2040s. You can confirm this number independently: Sprott Asset Management, S&P Global, the IEA, and the UNCTAD have all published versions of the same figure, ranging from 15 to 17 years as the average discovery-to-production period. A copper deposit found today cannot produce commercial copper until approximately 2041 or 2042. Every copper mine that is currently in production is therefore structurally irreplaceable for the next 15 to 17 years, regardless of what happens to copper prices in the interim.

This is the lock. It makes currently producing, permitted, infrastructure-equipped copper mines categorically more valuable than they would be if new supply could respond quickly to price signals — as agricultural commodities can, or as oil shale can through relatively rapid well drilling. Copper cannot respond quickly. The geology and the permitting timeline will not allow it.

Catalyst 1: The EV Copper Demand is Already in the Supply Chain — Not a Forecast

Global EV production is not a forecast. It is current factory output. In 2023, the global EV fleet grew by approximately 14 million vehicles, bringing the total to roughly 40 million EVs on the road globally. Each new EV contains approximately 83 kilograms of copper. Fourteen million new EVs consumed approximately 1.16 million tonnes of additional copper demand in a single year — purely from the EV increment. At current global primary mine production of approximately 22 to 23 million tonnes per year, this single demand category added approximately 5 percent of additional consumption above the baseline, from a standing start. The IEA projects global refined copper demand to grow from 27 million tonnes in 2024 to 33 million tonnes by 2035 — an increase of 6 million tonnes, or 22 percent, in a decade. Against primary mine production that the IEA projects will peak in the late 2020s and potentially decline thereafter, the arithmetic of deficit is not speculative. It is the intersection of confirmed demand growth and documented supply constraints.

Catalyst 2: Grid Modernisation — The Invisible Copper Demand Nobody Models

The electricity grid is copper-intensive in a way that most demand models undercount. The US electrical grid alone — often described as one of the most overdue infrastructure upgrades in any developed economy — requires an estimated 100,000 miles of new transmission lines over the next decade, plus the transformation of distribution infrastructure to accommodate bidirectional power flow from distributed solar generation. Transmission lines are copper. Distribution transformers are copper. The switchgear, substation equipment, and industrial cables connecting new renewable generation to existing grids are copper.

Wood Mackenzie has estimated that US grid modernisation alone could require an additional 500,000 to 600,000 tonnes of copper per year above current baseline demand, sustained for at least a decade. Europe’s equivalent grid investment programme, mandated under the EU’s Green Deal and REPowerEU frameworks, represents a comparable additional demand. China’s continuing grid expansion — the country adds more transmission capacity in a single year than the entire existing US transmission system — sustains the largest single increment of copper demand in the global market.

None of this is a projection about consumer behaviour or technology adoption. It is infrastructure investment mandated by governments, contracted by utilities, and funded by regulatory frameworks that are already in place. The copper for this grid modernisation will be consumed whether or not EV adoption exceeds expectations, whether or not battery technology evolves, and whether or not AI data centre demand grows as fast as currently projected. Grid copper is the baseload demand that makes every other copper demand story more secure.

Catalyst 3: The Ore Grade Decline — A Structural Cost Floor That Supports Price

Average copper ore grades globally have declined approximately 40 percent since 1991, per IEA data. In practical terms, this means that mining the same pound of copper today requires processing roughly 40 percent more rock than it did 35 years ago, consuming more diesel, more water, more grinding media, and more personnel. Capital costs for new mine development in Latin America have increased approximately 65 percent since 2020, approaching levels previously associated only with greenfield projects, according to IEA analysis. The consequence is that the minimum copper price required to incentivise new mine construction — the “incentive price” — has risen structurally, not cyclically.

When copper periodically falls below $3.00/lb, the market tends to interpret this as a return to normal. But $3.00/lb does not incentivise new mine construction at current capital and operating cost levels. The incentive price for a new greenfield copper mine in Chile or Peru today is estimated at $3.50 to $4.50/lb, accounting for current construction costs, energy costs, and financing requirements. Any copper price below this range does not generate the returns required to justify the 10 to 15 year investment horizon of a new mine. This means that when copper falls below $3.50/lb, new mine investment slows or stops — reducing future supply — which, after the 10 to 17 year development lag, pushes prices back above the incentive level. The incentive price is a structural floor, not a ceiling. Existing mines with production costs of $1.50 to $2.00/lb earn extraordinary returns at any price above the incentive level, because they were built when costs were lower and their capital is already sunk.

Subscribe now

Catalyst 4: The Supply Disruptions Are Structural, Not One-Off Events

The copper market experienced three significant supply disruptions in 2023 to 2025 that are worth examining not as individual events but as a pattern. First Quantum Minerals was forced to shut its Cobre Panama mine in November 2023 following widespread Panamanian public protests over environmental concerns, removing approximately 350,000 to 400,000 tonnes of annual copper production from the market — roughly 1.5 to 2 percent of global supply. Then, in 2025, accidents at Ivanhoe Mines’ Kamoa-Kakula complex in the DRC and at Freeport’s Grasberg mine in Indonesia disrupted production at two of the world’s largest copper operations simultaneously. Copper prices, which had already been supported by structural supply concerns, responded by reaching an all-time high of $6.61/lb on January 29, 2026.

Subscribe now

Catalyst 5: The US Critical Minerals Policy Shift — Already Active

The US government has formally designated copper as a critical mineral, making domestic and allied-country copper producers eligible for Inflation Reduction Act incentives, Department of Energy loan guarantees, and Export-Import Bank financing. The CHIPS and Science Act creates further semiconductor manufacturing investment that will drive domestic copper demand for advanced electronics. The bipartisan support for critical minerals policy — with both parties agreeing that Western supply chain security requires domestic and allied-country mineral production — creates a policy environment in which US-permitted copper projects are more likely to advance through the permitting process more quickly than the historical 17-year average.

This is not a guarantee of specific projects or specific timelines. It is a confirmed shift in the regulatory posture of the largest economy in the world toward favouring the production of copper from Western, allied, and domestic sources. For companies with permitted copper deposits in the United States, Canada, Australia, and allied Latin American jurisdictions, the policy environment has materially improved, reducing the permitting risk that has historically been a primary discount factor for development-stage assets.

Subscribe now

CONCLUSION: THE OWNER’S ADVANTAGE IN A MARKET THAT COUNTS QUARTERS

The copper investment case, stripped to its essentials, runs as follows.

An industry that makes its living finding and digging up copper has essentially stopped finding new deposits. Fourteen major discoveries in the last decade. Two hundred and twenty-five in the preceding 23 years. The deposits being mined today — the ore bodies that Freeport, Hudbay, Capstone, Sandfire, Ero, Taseko, and MMG have certified in their annual technical reports — are largely not being replaced. They are being depleted. And depletions, unlike earnings, are permanent.

At the same time, the demand for copper is growing from multiple independent directions simultaneously. Electric vehicles. Grid modernisation. Renewable energy. AI data centres. None of these demand sources requires the others to succeed; they each independently add copper demand, and they all depend on the same scarce supply.

A new copper mine that is discovered today takes 17 years to produce its first commercial pound. The incentive price to build it is $3.50 to $4.50/lb — well above what the market averaged for much of the previous decade. The ore grades in any hypothetical new deposit will likely be lower than the existing producing mines, making it more expensive to operate. The political and environmental permitting environment in every major mining jurisdiction is more challenging than it was 20 years ago.

Against this backdrop, the certified reserve base of an existing, producing copper mine has a value that GAAP does not record and that earnings models structurally understate. The accounting says: “This property is worth what we spent building it, minus what we’ve already depleted.” The owner says: “This property contains X billion pounds of copper, which generates Y net margin per pound at today’s prices and today’s costs. Multiply X by Y and that is the operating income embedded in this asset. No new mine needed. No discovery required. No 17-year wait.” That is the arbitrage.

The companies that exploit it are not making a forecast. They are reading a geological report, checking the current copper price, doing arithmetic on a production cost statement, and comparing the result to a share price. The gap between those two numbers — the owner’s value and the market’s price — is widest at exactly the moments when the earnings models are most negative and the consensus is most bearish. It is widest when copper is near its trough, when producers are reporting losses or minimal earnings, when the financial press is writing stories about China’s slowdown or the end of the commodity cycle.

It is, in other words, widest precisely when it is most uncomfortable to act on.

The ore body does not read the financial press. It does not respond to quarterly earnings calls. It sits in the ground, unchanged, waiting — exactly as it has been doing since before the mine was built, and exactly as it will continue to do until the mining industry finally, slowly, catches up to the demand that is already visible in the world being built around it.

This white paper is for educational and informational purposes only. All reserve figures, production cost data, market statistics, and company financial data are drawn from publicly available sources including company annual reports (10-Ks, 40-Fs, 20-Fs), NI 43-101 and JORC technical reports, IEA publications, S&P Global Market Intelligence data, and documented market transactions as cited throughout. Nothing herein constitutes investment advice.

If you like this read, check out my other writings 👇

Most great investments don’t announce themselves. They don’t come wrapped in a press release from Goldman Sachs. They don’t appear on the “most active” tab of your brokerage. They sit there, quietly, like a pile of wood at the edge of your property that you keep walking past — until one morning the temperature drops, and suddenly you realize what you’ve been ignoring.

This is a story about one of those piles of wood.

The company is Taseko Mines Limited, listed on the NYSE American under the ticker TGB. As of May 5, 2026, the stock trades at $6.85 a share with a market capitalization of approximately $2.5 billion. Book value per share is $1.56. Its ROE has been inconsistent — 10.8% in 2021, negative in 2022, a strong 21% in 2023, and back to slightly negative in 2024 and 2025. On the surface, this looks like a mediocre miner in a cyclical industry with a leverage problem and no earnings to speak of. The P/E ratio is undefined because there are no meaningful earnings — only the suggestion of them.

And yet.

When you open the annual report and read past the headline numbers — when you stop looking at what the accountants say and start looking at what a business owner would see — something strange happens. The picture inverts. Completely.

This is the story of how that inversion works, why the market is missing it, and — crucially — why I am not yet buying it.

Subscribe now

The Problem With Accountants

Here is a truth that took me a long time to fully internalize: accounting was not designed to tell you what a business is worth. It was designed to tell you what happened, historically, in a conservative way, to protect creditors. GAAP is a retrospective system built for debt contracts, not for owners.

Charlie Munger put it simply: the most dangerous thing in investing is a man with a set of financial statements and no understanding of what they represent. The numbers themselves are not the reality — they are a shadow of the reality, cast by a very particular light source pointing in a very particular direction.

In the mining industry, this distortion is almost comically severe. The reason is geological.

When a mining company identifies a copper reserve — after drilling hundreds of wells, assaying samples, running geological models, applying metallurgical recovery factors, and filing a formal technical report under NI 43-101 or SEC guidelines — it is essentially describing a physical asset that exists in the ground right now. The copper is there. It has been verified by qualified persons. It has a grade, a mass, and a location. You can calculate how many pounds of copper sit inside that rock with the same confidence that you can calculate how many gallons of milk are in a dairy farm’s tank.

But here is where it gets strange. Under GAAP, that reserve is not carried at the value of the copper it contains. It is carried at cost — the historical cost of exploration and acquisition. The copper grade climbs, commodity prices rise, and inflation makes the replacement cost of the mine exponentially more expensive. None of that shows up on the balance sheet. The asset just sits there, at its original price tag, aging quietly into irrelevance as a number, while appreciating furiously as a physical reality.

This is what we are dealing with at Taseko. The balance sheet says one thing. The earth says something very different.

Subscribe now

The Inventory: What Is Actually in the Ground

Let me build this up the way a business owner would — not a hedge fund manager running a DCF, not an accountant filing a 10-K, but someone who just bought the company outright and is walking through the warehouse with a clipboard.

The Gibraltar Mine, British Columbia. This is Taseko’s operating core and has been for decades. It is one of the largest open-pit copper mines in Canada. The ore grade is about 0.25% copper — genuinely low-grade by global standards, which is relevant for costs and we will come back to it. Within the Gibraltar system there are four sub-pits: Pollyanna, Connector, Gibraltar, and Extension. The proven copper reserves, doing the math directly from the technical reserve tables:

Pollyanna carries 91 million tonnes at 0.24% copper = 436.8 million pounds. Connector carries 118 million tonnes at 0.25% = 590 million pounds. The Gibraltar pit itself holds 121 million tonnes at 0.24% = 580.8 million pounds. Extension carries 84 million tonnes at a richer 0.31% = 520.8 million pounds. The Gibraltar mine complex in total: 2,128.4 million pounds of proven copper.

That number alone — 2.1 billion pounds — is staggering when you hold it against the $2.5 billion market cap. But we are not finished.

The Florence Copper Project, Arizona. This is not just another mining project. This is something genuinely new, and it changes the whole character of Taseko’s cost structure and the timeline to cash. More on this shortly. The proven reserves at Florence: 258 million tonnes at 0.35% copper, which works out to 1,812 million pounds (the technical report rounds it to 1,806 million).

The Yellowhead Copper Project, British Columbia. This is a development-stage project — Taseko does not operate it yet, and it would require enormous capital to bring into production. The proven reserves stand at 458 million tonnes at 0.29% copper = 2,928 million pounds.

Combined total across all three asset bases: 6,862.5 million pounds of proven copper.

Subscribe now

That is six point eight billion pounds. At a low base average LME copper price of approximately $4.50 per pound, the gross in-ground value of just the copper — before any costs, before any discounting — is roughly $30.9 billion.

The market cap is $2.5 billion.

Now, the immediate and entirely correct objection is: “You can’t just multiply reserves by the price it’ll sell for. You have to account for operating costs, capital requirements, time value of money, and execution risk.” Absolutely right. Let us do exactly that.

What a Business Owner Actually Pays Attention To: The Cost Structure

This is where the story gets interesting, because the cost side of this equation is not uniform across Taseko’s assets. It bifurcates sharply. And that bifurcation is the entire thesis.

The Gibraltar mine runs at what the industry calls C1 cash costs — the full operating cost to produce a pound of copper, net of by-product credits. Historical data from Taseko’s own filings is clear: Full year 2023 averaged $2.37 per pound, among the most efficient years. Full year 2024 averaged $2.66 per pound, pressured by labor strikes and lower grades. Full year 2025 came in at $2.66 per pound as well. Q4 2025, after operational improvements and a return to higher grades, dropped to $2.47 per pound. Q4 2023 at peak efficiency hit $1.91 per pound.

At today’s LME copper price of roughly $4.50 to $5.00 per pound, the margin on Gibraltar copper looks like this: at $2.66 all-in, Gibraltar earns between $1.84 and $2.34 per pound of copper produced. In 2025, the company produced roughly 93 million pounds of copper, translating to operating economics somewhere between $171 million and $218 million in gross cash margin from Gibraltar alone, before corporate overhead and finance costs.

But Florence Copper is a different animal entirely.

Subscribe now

The Technology That Changes Everything: ISCR and Why It Matters

Here is where I need to slow down and explain something that most people, even in the mining industry, don’t fully appreciate.

Conventional copper mining — what Gibraltar does — is a grinding, energy-intensive, expensive business. You blast open rock, haul it with 400-ton trucks burning diesel, crush it in a primary jaw crusher, grind it in rotating SAG mills filled with steel balls (the most electricity-intensive step in all of mining), then run the powdered slurry through a flotation circuit using chemical reagents — almost all of which are petroleum-derived — to produce a concentrate that is 20-30% copper. Then you ship that concentrate, pay treatment and refining charges (TC/RCs) to a smelter in China or Japan, and receive back cash net of those deductions. The payable percentage (by clients) is roughly 96% of contained copper. The TC/RC deductions add $0.20 to $0.35 per pound to the effective cost.

The whole process is thermally and energetically violent. It is a war against physics, run at industrial scale, every day of the year.

Florence Copper uses none of this. The process requires no blasting, no loading, no hauling, no crushing, no conveying of mineralized material, resulting in 75% fewer carbon emissions, 65% less energy, and 78% less water consumed per pound of copper produced than conventional open-pit copper mines in Arizona.

Instead, Florence injects a weakly acidic solution — roughly the same pH as household vinegar — into a naturally fractured copper orebody via a series of injection wells in the copper oxide zone, 400 to 1,200 feet below surface, causing copper minerals to dissolve into solution prior to being pumped to surface through recovery wells, and processing it into LME grade A copper cathode sheets using an electrochemical solvent extraction and electrowinning process.

No trucks. No mill. No smelter. No TC/RC deductions. No flotation reagents. No tailings pond.

Florence Copper is the first greenfield site globally to employ ISCR at commercial scale, establishing operational precedent for this technology.

The economics of this process are almost offensive in their simplicity. Taseko’s technical team has guided to a C1 cash cost for Florence of approximately $1.10 per pound of cathode. At a $4.50-5.00 copper price, that is a margin of $3.40 to $3.90 per pound.

To put that in perspective with the “$2.5 billion market cap” line at the top: Florence alone, at full capacity, could theoretically pay for the entire current market cap of Taseko in less than nine years of operations, before accounting for Gibraltar’s own cash flows.

Once Florence Copper achieves its nameplate capacity of 85 million pounds of LME Grade A copper metal per year, Taseko will become the third largest copper cathode producer in America.

Subscribe now

The Hidden Value: What the Balance Sheet Refuses to Show You

Now let us apply the framework from the first document on hidden asset values — specifically the “Depreciated PP&E” concept — to Taseko’s situation, because it maps almost perfectly.

The Gibraltar mine has been running since the early 1970s. Much of the physical infrastructure — the mill buildings, the conveying systems, the tailings dam, the power connections, the road networks — has been depreciated for decades. On the balance sheet, Property, Plant and Equipment stands at roughly $2 billion Canadian. But the replacement cost of a comparable operation — a fully permitted, operating copper mine in Canada with over 2 billion pounds of proven reserves, an active mill processing 130,000+ tonnes of ore per day, with all environmental permits in place — is not $2 billion. It is multiples of that.

This is the insight that is almost never discussed about mining companies: the permit itself is worth more than the building. Environmental permits for large open-pit copper mines in British Columbia are effectively impossible to obtain today. The regulatory timeline, the First Nations consultation requirements, the environmental assessment process — all of these have lengthened dramatically since Gibraltar was first permitted. The mine that Gibraltar represents could not be permitted from scratch in under 20 years today, and perhaps not at all. What Taseko owns, therefore, is not just a mine. It is a grandfathered regulatory position — a Certificate of Need, if you want to borrow the healthcare analogy — that has no market value on the balance sheet but commands a significant premium in any rational acquisition scenario.

This is also true, even more dramatically, for Florence. The ISCR process in Arizona required permits from both the Arizona Department of Environmental Quality and the EPA. The combination of the unique geological conditions required for ISCR to work, the permitting process, and the multi-year Production Test Facility phase means that the demonstration of technical viability and operational economics may encourage mining companies to reassess suitable resources using in-situ recovery methodologies — but nobody else can just go build a second Florence. The site-specific hydrogeology, the prior community relationships, the standing permits, and the already-amortized PTF costs are all embedded in Taseko’s asset at zero marginal book value.

Subscribe now

The By-Products No One Is Talking About

Every good mine has a secret second story. At Gibraltar, the second story is molybdenum.

Molybdenum is a silvery metal used primarily as an alloying agent in high-strength steel — infrastructure, aerospace, defense, oil and gas pipelines. It trades in a relatively thin market, and prices can be volatile. But the point is that Gibraltar produces it as a by-product of copper mining. The molybdenum comes up in the same ore, goes through the same mill, and requires only modest incremental processing to separate. The marginal cost of the molybdenum Taseko produces is essentially zero — or rather, it is already embedded in the C1 cost calculation as a negative number, because the molybdenum revenue offsets copper costs.

The sulphide ore at Gibraltar carries approximately 21,000 metric tonnes of proven molybdenum. At current market prices of roughly $63,900 to $64,486 per metric tonne (based on March-May 2026 LME trends), the total intrinsic value of the molybdenum in the ground is approximately $1.34 to $1.35 billion.

That molybdenum sits on the balance sheet at close to zero. It will be produced over the remaining life of the mine. It will offset a portion of copper operating costs year after year. And it represents roughly 54% of the entire current market cap of Taseko, in by-product minerals alone.

Then there is the Yellowhead project’s by-products. The 458 million proved tonnes carry gold at 0.031 gpt and silver at 1.3 gpt. At spot prices of approximately $4,600 per troy ounce for gold and $75 per ounce for silver:

14.2 tonnes of gold = 456,540 troy ounces × $4,600 = $2.10 billion. 595.4 tonnes of silver = 19,142,527 troy ounces × $75 = $1.44 billion.

Total precious metal value within the proved resource base: $3.54 billion. Against a market cap of $2.5 billion.

Subscribe now

This is the base case. The “recession scenario.” The floor where copper is at lows, oil is expensive, and the world has forgotten why it wanted copper in the first place. Even in that scenario, the gold and silver in the ground — not the copper, not the molybdenum, just the precious metals — exceed the current enterprise price you pay for the entire company.

In aggregate — $3.54 billion in gold and silver, $1.35 billion in molybdenum, and $3.8 to $21.2 billion in copper (using the $0.56 to $3.10 net-back spread across 6,862.5 million pounds) — the total intrinsic resource base ranges from $3.8 billion at the most conservative estimate to approximately $26 billion at the high end, against a market cap of $2.5 billion.

A 52% undervaluation at the floor. A 937% gap at the ceiling.

Subscribe now

The Copper Market: The Wind at Its Back

The question is not whether copper is important. The question is how important and how soon.

According to S&P Global’s January 2026 study “Copper in the Age of AI: The Challenges of Electrification,” the accelerating pace of electrification is projected to swell copper demand to 42 million metric tons by 2040, a 50% increase from current levels, yet existing supply is currently poised to decrease as the mining sector faces challenges across the copper value chain.

The supply gap threatens to constrain technological advancement and economic growth as copper becomes increasingly essential for AI data centers, electric vehicles, renewable energy infrastructure, and defense systems. Without significant changes to supply, global copper production is projected to peak at 33 million metric tons in 2030 before declining.

BloombergNEF’s Transition Metals Outlook warns copper demand for the energy transition could triple by 2045 and that the metal may enter structural deficit as early as 2026.

An electric vehicle contains roughly 80-100 kilograms of copper, three to four times as much as a conventional internal combustion car. A single offshore wind turbine uses approximately 3 to 15 tonnes of copper depending on size. Data center construction — driven by the AI buildout — consumes copper in the wiring, cooling infrastructure, and power distribution systems. Defence spending is increasing globally. The arithmetic of demand is not a prediction. It is a function of decisions that have already been made, announced, and begun.

On the supply side, the structural problem is even cleaner. In developed regions, opening new mines takes 20 to 30 years, with the U.S. averaging close to 29 years. Six countries are responsible for roughly two-thirds of mining production, and China commands roughly 40% of total smelting capacity and 66% of copper concentrate imports, making the global supply chain vulnerable to policy shocks and trade barriers.

This last point is specifically relevant for Florence. Taseko’s Arizona cathode bypasses China entirely. It goes from the wellfield to the SX/EW plant to an American manufacturer — no concentrate, no Chinese smelter, no geopolitical exposure. In a world where supply chain sovereignty has become a first-order policy concern, the value of domestically produced, LME-grade copper cathode with zero refining intermediaries is not adequately captured by the LME spot price alone. There is a premium — qualitative, and perhaps increasingly quantitative — embedded in Florence’s product that does not show up in Taseko’s revenue line yet.

Subscribe now

The Contrarian’s Pause: Why I Am Not Buying Today

I have spent several pages building a case for why Taseko’s assets are extraordinary. Now let me do something more useful and spend equal time tearing it apart.

The first principle of thinking about any investment is simple: invert. Don’t ask “why will this work?” Ask “how could this destroy me?” The bear case on Taseko is not vague or philosophical. It is concrete and it is serious.

The CAPEX abyss. Of the 6,862.5 million pounds of proven copper, only a fraction is immediately monetizable without massive capital investment. Florence is producing, and its reserves of 1,812 million pounds are progressively extractable without large incremental capital — the $230 million has been spent and the wells are flowing. Gibraltar is operating (most of it at least), and its 2,128 million pounds will be produced over the remaining mine life at current throughput rates. But Yellowhead — the 2,928 million pounds that makes up 43% of the total reserve base and the engine of the “937% upside” calculation — needs approximately $2 billion in initial capital expenditure and has no current infrastructure of any kind. It is, in the parlance of the field, a development-stage project. You cannot sell it today at intrinsic copper value.

You own an option, not the copper. And this option is not cheap (at PS of approx 4, + a small cash balance relative to the CAPEX that is required to unlock the company’s full intrinsic value)

And even Gibraltar and Florence require ongoing maintenance capital. The balance sheet as of December 31, 2025 shows cash of approximately $188 million Canadian against long-term debt of approximately $711 million. The company is not in a position to self-fund major expansion. That means (very likely) equity dilution, given the fact that debt is quite high — at about 95% of equity.

The copper-oil double bind. Here is an insight that rarely appears in the bullish mining literature: copper and oil have a complex, unhealthy relationship. Copper demand is highly correlated with economic growth. When growth slows — especially industrial growth, which is what eventually always follows a period of over-investment in a capital markets bubble — copper demand falls first and hardest. It is the most economically sensitive of all base metals, precisely because it is used in everything productive.

At the same time, copper production is deeply dependent on oil. Gibraltar’s cost structure is materially affected by diesel — 400-ton haul trucks, the primary energy input for blasting (ANFO explosive), off-grid power generation for remote operations, and the thermal input to reagent manufacturing. Taseko’s own Q1 2026 guidance acknowledged this explicitly, noting that diesel prices at current levels would add $0.10 to $0.15 per pound to Gibraltar’s annual operating costs. When oil rises because the economy is strong, copper demand is high, margins hold. But if oil rises in a stagflationary environment — rising input costs with declining product prices — the margin gets crushed from both ends simultaneously.

This is not an abstract concern. The froth flotation circuit that turns raw ore into concentrate is essentially impossible without petroleum-based flotation reagents. Every pound of concentrate that comes out of Gibraltar has hydrocarbon molecules embedded in its production cost. This dual pressure — to oil prices and from economic contraction — means Taseko’s earnings profile is riskier than the copper price alone would suggest.

The Wilshire 5000 signal. I want to be careful here not to be too clever, but the macro backdrop matters for the timing of an investment even when the fundamental value is clear. As of May 2026, the Wilshire 5000 to GDP ratio — the Buffett Indicator — stands at approximately 226%. This is not a trivial data point. The historical average for this ratio, going back decades, is somewhere between 80% and 100%. At 226%, it reflects either a permanently higher plateau of corporate earnings power relative to the economy, or a significant mispricing of risk across asset classes. If the latter, the unwinding (economic contraction) — whenever it happens — will depress copper prices through demand destruction before it resolves. Taseko’s copper reserves do not disappear in a recession. But the price at which they are monetized over the next five to ten years does compress, and the cash flows that allow the company to develop Yellowhead shrink with them.

Copper is cyclical. Intrinsic resource value is permanent. The distance between those two truths is exactly the gap between a good idea and a good investment.

The Florence ramp-up question. Florence is producing. Q1 2026 saw 1.5 million pounds of copper cathode produced, with solutions flowing in the wellfield since late 2025 and initial copper leaching production in line with expectations based on internal modeling. But 1.5 million pounds in a quarter is a long way from 85 million pounds per year. Full annualized production in Q1 works out to 6 million pounds — about 7% of nameplate. The ramp from 7% to 100% involves wellfield expansion, increasing acid injection rates, managing solution chemistry, and resolving the operational learning curves inherent in the world’s first commercial-scale ISCR project. Total Florence production for 2026 is expected to be in the range of 30 to 35 million pounds of copper cathode. At 35 million pounds, Florence is 20-30% of nameplate. Full capacity is not a 2026 story. It is a 2027-2028 story at the earliest.

This matters because the $3.75 billion lifetime gross profit I calculated earlier is a 22-year number. In the near term, Florence is contributing partial cash flows against a fully loaded interest expense. The company’s ability to grow into its intrinsic value is a function of time, operational execution, and the copper price environment during the ramp period — none of which is in the investor’s control.

Subscribe now

The PDP Insight: What Florence Is Versus What Gibraltar and Yellowhead Are

There is a concept in oil and gas valuation called PDP — Proved Developed Producing reserves. These are reserves in wells that are already drilled, already completed, already flowing. You do not need to spend another dollar to produce them. The next molecule of oil that comes up that wellbore is essentially free capital, subject only to operating costs. PDP is the highest quality reserve category in the entire oil and gas taxonomy, and it is priced accordingly — often at the highest multiples per BOE because the execution risk has already been paid for.

Florence’s in-situ reserves are the copper equivalent of PDP. The wellfield exists. The acid is flowing. The solution is being pumped. The SX/EW plant is running. Copper cathodes are being harvested. No new drilling. No new permitting. No new capital. Every pound that comes up is a pound you capture at $1.10 of cost against $4.50 of price.

Gibraltar is more like PUD — Proved Undeveloped, or in this case, Proved Developed but requiring ongoing maintenance capital. You know the ore is there, you know the equipment works, but the mine will eventually exhaust the current pit and need to expand, and the expansion costs real money.

Yellowhead is a wild card — resource certainty without production reality. It exists as an asset only in a financial model, not in the physical world.

The implication is that the valuation spread on Taseko’s asset base is wider than it looks from the outside. The high-quality, immediately-value-generating assets (Florence plus the running portions of Gibraltar) are genuinely compelling. The development-stage assets are contingent value that requires capital, time, and benign market conditions to realize.

A sophisticated buyer of Taseko equity today is not buying the whole $26 billion intrinsic number. They are buying the high-quality near-term assets at a reasonable price, and getting the long-dated optionality on Yellowhead and expanded Gibraltar essentially for free. That is a very different and more defensible proposition.

Subscribe now

The Molybdenum Wildcard: The By-Product That Could Be a Business

The molybdenum story at Gibraltar deserves its own brief chapter because it illustrates exactly the kind of hidden value that GAAP obscures.

Molybdenum prices have been extraordinarily volatile over the past decade, as the Taseko analysis document shows. The 10-year chart is arresting: prices ranged from roughly 100 CNY/kg through 2018-2019, then exploded to nearly 900-1,000 CNY/kg in the 2022-2023 period before settling back to around 244 CNY/kg at the time of writing. The 21,000 metric tonnes of molybdenum in the Gibraltar sulphide ore is, at 244 CNY/kg (approximately $33,600/tonne), worth approximately $706 million at today’s price. At the 2022-2023 peak pricing near 400 CNY/kg, the same reserves were worth $1.2 billion.

This by-product is not a projection. It is contained in ore that is already being mined. It comes out of the same plant, in the same process. Taseko does not have to build anything to capture it. It just comes up with the copper, is separated in the molybdenum circuit, and is sold.

The by-product credit — roughly $0.20 to $0.35 per pound of copper, depending on molybdenum prices — is the primary reason Gibraltar can maintain C1 costs in the $2.40-2.66 range despite being a low-grade mine. Without the molybdenum credit, C1 costs would be approximately $0.30 per pound higher.

Subscribe now

Capital Structure and Dilution: The Last Fortress

One of the quiet ways good investments get ruined is through share issuance. You buy a company at a discount to intrinsic value, and then management issues so many new shares at cheap prices that your per-share value erodes even as the company grows. This is the dilution trap, and mining companies are particularly prone to it because they are perpetually capital-hungry.

Taseko’s dilution profile, at this moment, is actually quite clean. As of February 18, 2026, there are 364,557,150 common shares outstanding and 8,091,464 stock options. All 8 million options carry strike prices well below 50% of the current market price — meaning they are deeply in-the-money and likely to be exercised, adding approximately 8 million shares or a 2.2% dilution in the worst case. In a company of this size, with this asset base, a 2% dilution from a fixed option pool is essentially noise.

There are no outstanding warrants, no convertible debentures at punitive conversion rates, no history of toxic equity financing — the kind of structure that kills junior miners.

Subscribe now

Why This Is On My Watchlist, Not In My Portfolio

The conclusion of this analysis is not “buy immediately.” It is “watch very closely.” The distinction matters.

The intrinsic value case is clear and compelling. The assets are real, verified, and priced by the market at a substantial discount to their in-ground economic value. Florence is producing. Gibraltar is performing. The macro tailwinds for copper are as strong as any I have seen for a commodity in recent memory. The dilution risk is low. The management has executed the Florence construction on time and on budget — a genuinely difficult achievement for a first-of-kind greenfield mining project.

But timing is the variable that determines whether a right idea becomes a profitable investment.

The Wilshire 5000 to GDP ratio at 226% is not just a chart curiosity. It represents trillions of dollars of capital sitting in equity positions at valuations that presuppose a level of economic growth and corporate earnings expansion that, historically, has never been delivered. When that valuation reversion begins — and it will, because it always does — the first assets to sell are the cyclical ones. Copper is cyclical. Mining companies are cyclical. The market will sell Taseko before it sells Microsoft, because Taseko’s earnings are volatile and its narrative is difficult to tell at a cocktail party. That selling will create the entry point. And although we have significant proven reserves of Gold and Silver, they would still have to be unlocked by a capital catalyst.

The time to buy Taseko is when copper is under $3.00 per pound and the headlines say “copper glut” and the mining stocks have been abandoned. That is when the $1.10 C1 cost at Florence becomes the most powerful fact in the entire investment universe — because Florence is cash-flow positive even at $2.00 copper, while every conventional mine in the world is bleeding.

Alternatively, the time to buy is when Yellowhead reaches a final investment decision with committed financing, because at that point the market can begin to price it as a producing asset rather than a development option.

Until then, the thesis sits here on the watchlist, fully formed, waiting for the price to come to the idea rather than the idea chasing the price.

There is something almost Buffett-like in that discipline. He once described investing as waiting at the plate until the pitcher throws exactly the ball you want, with no called strikes. The pitcher for Taseko is still warming up. The ball I want — a copper-bearish macro selloff, a mining sector exodus, a temporary decline in TGB toward the $4-5 range, and a capital catalyst — has not yet been thrown.

When it is, I intend to swing hard.

Subscribe now

The Full Picture, Summarized Without Mercy

What we have in Taseko Mines is a company that owns:

6.8 billion pounds of proven copper reserves, of which roughly 4 billion (Gibraltar plus Florence) are immediately or near-term actionable without large new CAPEX. It owns the first commercial-scale ISCR copper operation in the world, producing at $1.10 per pound against a $4.50-5.00 copper market. It owns 21,000 metric tonnes of molybdenum by-product worth over $700 million at current prices. It owns 14.2 tonnes of gold and 595 tonnes of silver worth a combined $3.54 billion.

The market cap is $2.5 billion.

The refined copper market is projected to shift from a surplus in 2025 to a deficit of 150,000 tonnes or more by 2026, against a backdrop where new mines take 20-29 years to permit in developed countries.

The bears will say: leverage, cyclicality, unproven ISCR ramp-up, $13-18 billion in eventual CAPEX needed to monetize all reserves, and a macro backdrop that has historically preceded commodity demand destruction.

All of those are fair. None of them are fatal to the thesis over a five to ten year horizon.

The mountain of copper in British Columbia and Arizona is not going anywhere. The acid in the Florence wellfield is dissolving it right now, bringing it to the surface at a cost structure that would have seemed like science fiction to a conventional miner fifteen years ago. And the world — with its EVs and its data centers and its wind turbines and its defense infrastructure — needs every pound of it.

The only question is the price at which you choose to acquire your share of what the earth has already made.

I know my price. I am waiting.

Disclosure: This is a personal investment analysis document, written for educational and watchlist purposes. It does not constitute financial advice. All figures sourced from Taseko Mines public filings, Taseko Q1 2026 operational update, S&P Global, BloombergNEF, Wood Mackenzie, Florence Copper project documentation, and International Copper Study Group public reports. The author does not currently hold a position in TGB.

If you like this read, check out my other writings 👇

A Note on Method Before We Begin

Every analyst who has ever covered Cameco, NexGen, or Kazatomprom has opened a financial model and begun discounting future cash flows. They have built tables of uranium price assumptions, escalating production schedules, operating cost projections, EBITDA multiples, and present value calculations using discount rates they cannot justify with anything more rigorous than gut feeling and precedent.

None of that appears in this paper.

Instead, this paper asks a simpler question — the question a mine owner asks, not a financial modeler. That question is: how many pounds of uranium are in the ground, what does it cost to get them out, and what can I sell them for right now?

The answer to that question — a business owner’s answer, derived from certified reserve reports and current spot prices, not from spreadsheets built on assumptions — is systematically and repeatedly higher than what public markets price these companies at. Understanding why that gap exists, how large it is, why it has persisted for decades, and why specific, already-confirmed catalysts are now closing it — that is what this paper is about.

Subscribe now

Part One: The Asset

What a Uranium Company Actually Owns

Start with a physical fact that no financial model can alter: uranium is a finite, non-renewable substance embedded in specific rock formations at specific grades in specific places on earth. It cannot be manufactured, substituted for fuel purposes, or meaningfully extracted from seawater at any economic cost. The 439 operating nuclear reactors in the world as of 2025 need approximately 180 million pounds of uranium oxide (U₃O₈) per year to generate the electricity they were built to produce. Those reactors do not need uranium that might be in the ground somewhere. They need uranium that has already been found, certified by qualified geologists under internationally audited standards, and measured to within a narrow margin of error.

That certified, measured, verified quantity of uranium in the ground is called a Proven and Probable Mineral Reserve. Under National Instrument 43-101 (Canada) and JORC Code (Australia) — the two primary reserve certification standards for uranium companies — a reserve can only be declared when it has been drilled at sufficient density to confirm the grade and continuity of the ore body, and when the economics have been demonstrated viable at a stated commodity price. This is not a guess. It is an engineering statement about what is physically extractable from a specific hole in the ground.

The business owner’s arithmetic begins here. Take the reserve in pounds. Subtract the cash operating cost per pound to get it out of the ground and into a drum of yellowcake (uranium concentrate). What remains is the net operating income embedded in the ground — before a single future assumption is required.

For Cameco Corporation, the world’s second-largest uranium producer, the arithmetic is direct. The company’s annual reports disclose that it has access to approximately 433 million pounds of proven and probable mineral reserves (its attributable share across McArthur River, Cigar Lake, and its other operations). As of mid-2025, the uranium spot price was approximately $71.75 per pound. Cameco’s reported cash operating costs for its underground Canadian operations run approximately $20–$30 per pound, with the midpoint at $25. Net operating margin per pound: $71.75 minus $25 = $46.75. Applied to 433 million pounds of certified reserves: approximately $20.2 billion in gross operating income embedded in existing, certified, drilled-out reserves.

Cameco’s market capitalization in early 2026 was approximately $22–$25 billion. That means the market is paying roughly 1.1 to 1.2 times the gross operating income of the company’s proven reserves — and ascribing essentially zero value to Cameco’s Fuel Services segment (uranium conversion and refining), its 49% ownership stake in Westinghouse Electric (the world’s leading nuclear reactor technology company, which alone was valued at approximately $7.9 billion in its 2023 acquisition by Brookfield and Cameco), and its long-term contract book that locks in multi-year sales at prices above the current spot market.

This is the gap. Not a speculative gap, not a forecasted gap, but a gap between a number you can calculate today — pounds times net margin per pound — and the price the market is charging for the whole enterprise.

Subscribe now

Why the Balance Sheet Is Useless for This Asset

Under both US GAAP (ASC 932) and IFRS (IFRS 6), mineral reserves are not recognized as assets at fair value on the balance sheet. Full stop. The accounting treatment is to capitalize the cost of exploration and development drilling as Property, Plant, and Equipment, and then deplete those capitalized costs against production using the Units of Production method.

What this produces is absurd from an economic perspective. Cameco has been drilling and developing McArthur River since the 1980s. The capitalized development costs that remain on the balance sheet after decades of accumulated depletion represent what was spent to find and develop the reserves, not what those reserves are worth. A deposit that cost $400 million to develop and has been 60% depleted shows a net book value of $160 million — completely independent of whether the remaining pounds are worth $1 billion or $5 billion at current prices.

The balance sheet is therefore not merely unhelpful for uranium analysis. It is actively misleading. A naive analyst who screened for uranium companies trading at low Price-to-Book ratios would find nothing, because all major uranium producers trade at significant premiums to their accounting book value. The book value bears no relationship to the economic value of the pounds in the ground. A sophisticated analyst ignores the balance sheet entirely and does the reserve math directly.

Wall Street does neither of these things cleanly. Instead, it builds discounted cash flow models that require uranium price assumptions three, five, and ten years into the future — assumptions that carry enormous uncertainty — and discount everything back at arbitrary rates. The result is a valuation that looks precise but is built entirely on forecasts that could be wrong by 50% in either direction. The business owner’s method — pounds in the ground times net margin per pound at today’s price — requires no forecast at all. It is an arithmetic calculation about what already exists.

Subscribe now

Part Two: The Story of How This Market Works — And Why the Gap Persists

To understand why uranium reserves are chronically mispriced, you need to understand the specific structure of the uranium market. It is unlike any other commodity market, and those structural features create the mispricing that this paper is documenting.

The Long-Term Contract System: Why Spot Prices Understate Realized Revenue

Uranium does not trade on an open exchange like copper or oil. There is no London Metals Exchange for uranium, no futures market that sets a continuous public price, no real-time bid-ask spread visible to anyone with a Bloomberg terminal. Instead, nuclear utilities buy uranium primarily under long-term supply contracts — privately negotiated agreements that run for 3 to 15 years between the utility and the mining company. The contracted price is typically higher than spot, because utilities pay a premium for supply security.

The World Nuclear Association has documented this structure consistently: as of the most recent available data, approximately 75% of global uranium supply is sold under long-term contracts, with only 25% traded on the spot market. This means the spot price — the number reporters and analysts quote — represents only the marginal clearing price for 25% of the market. The other 75% is contracted at prices that can be meaningfully higher. Cameco’s average realized price in 2023 was approximately $57.78 per pound, against a spot market that averaged approximately $55–$60 per pound during the year — but in 2024, as contracts signed in 2021–2023 at escalating term prices began to deliver, Cameco was realizing significantly above spot.

This means the business owner’s arithmetic using today’s spot price actually understates the net margin per pound for a company like Cameco that has locked in long-term contracts above spot. The embedded value of the reserve is higher than the spot-based calculation suggests.

Subscribe now

The Inventory Liquidation that Suppressed Prices for a Decade — and Why It Is Over

The single most important structural fact about uranium pricing from 2011 to 2021 is one that most financial analysts never examine. For the decade following the March 2011 Fukushima Daiichi nuclear accident in Japan, uranium prices collapsed from approximately $70/lb to as low as $18/lb by 2016 and spent years in the $20–$30 range. Every financial model built during this period — models that Wall Street analysts used to “value” uranium companies — was implicitly or explicitly forecasting that these depressed prices would persist or recover only slowly.

The actual cause of the price collapse was the liquidation of a specific, finite, and non-recurring inventory: Soviet-era weapons-grade uranium. From 1993 to 2013, a program known as “Megatons to Megawatts” converted 500 metric tonnes of Russian highly-enriched uranium (originally built for nuclear warheads) into low-enriched uranium suitable for reactor fuel. This created approximately 500 million pounds of equivalent reactor-grade uranium over 20 years — or about 25 million pounds per year of secondary supply that required zero mining to produce. Combined with Japanese reactor shutdowns (Japan had operated 54 reactors; all were eventually idled after Fukushima), the market was simultaneously losing 10–15% of global demand while gaining a persistent secondary supply source. The significant increase in supply relative to demand resulted in the massive drop in prices.

The critical point: both of these suppression factors had a defined end date. The Megatons to Megawatts program concluded in December 2013. Japanese reactor restarts were a regulatory question, not a physics question. The secondary supply overhang would eventually be consumed, and the reactors would eventually restart or be permanently decommissioned. In either case, the structural suppression was temporary.

Uranium producers who understood this stayed in the business, reduced output to preserve reserves, and waited. Uranium producers who valued their companies using the discounted cash flow models of 2015 — which mechanically extrapolated a $28/lb uranium price into the future — concluded their assets were nearly worthless. The ones who calculated pounds times net margin at a normalized price, and found that number to be many times the market cap, understood what they owned.

Subscribe now

Part Three: Three Case Studies in the Gap Between Book and Reserve

Case Study 1 — Cameco at the Trough (2016)

By the end of 2015, Cameco Corporation — the premier uranium mining company in the Western world, operator of the McArthur River mine (the world’s highest-grade uranium mine, at approximately 17% U₃O₈ average grade versus a global average of less than 0.1%), and owner of certified reserves totaling hundreds of millions of pounds — was trading on the Toronto Stock Exchange at approximately CAD $15 per share. Analysts were uniformly bearish. Price targets clustered around CAD $12–$18. The consensus DCF model, fed with $30/lb long-term uranium price assumptions, produced valuations that in several cases were below the cash on Cameco’s balance sheet plus the depreciated book value of its Canadian mining assets.

What the analysts were doing, specifically, was this: they took the prevailing spot price of approximately $28–$35/lb, escalated it at a modest rate in their models, applied it to future production schedules, discounted at 8–10%, and arrived at a net present value. From that NPV they subtracted net debt, divided by shares outstanding, and called it a price target. Every step of this process was theoretically defensible. Every step also systematically undervalued what was in the ground.

In July 2018, after years of losses from operating McArthur River and Cigar Lake at cash costs above the prevailing spot price, Cameco did something that financial models did not predict: it shut both mines down entirely. McArthur River, the world’s highest-grade uranium operation, was placed on care and maintenance. Cigar Lake continued at reduced capacity. Cameco’s management stated publicly that it preferred to buy uranium in the spot market to fulfill existing contracts rather than mine it at a loss. This was not a statement of despair. It was the statement of a business owner who understood the value of what was in the ground and refused to liquidate it at prices below its worth.

The financial community interpreted the mine closures as a negative signal. Production guidance cuts drove further analyst downgrades. The stock declined further, reaching approximately CAD $12–$14 per share in late 2018. By this point, the business owner’s arithmetic was stark: Cameco held approximately 455 million pounds of proven and probable reserves. At $28/lb spot with $25/lb operating costs, the net margin was $3/lb — barely positive and not sufficient to justify operating the mines. But the reserves had not disappeared. They were still in the ground. A buyer who purchased the entire company at the 2018 market cap of approximately CAD $5–$6 billion ($3.8–$4.5 billion USD) was purchasing the right to produce hundreds of millions of pounds of uranium from the world’s highest-grade mines — at the option to produce them when prices made it economic. At any uranium price above $40/lb (well within the historical range), the net margin per pound was $10–$15+, and across 455 million pounds, the embedded operating income was $4.5–$6.8 billion. The market, guided by DCF models assuming $30/lb uranium forever, was selling this for $3.8 billion.

Uranium spot prices began their recovery in late 2021 and accelerated dramatically through 2022, 2023, and early 2024. In January 2024, the uranium spot price broke above $100/lb for the first time in 17 years. Cameco restarted McArthur River in November 2022 and resumed full operations at Cigar Lake. By early 2026, Cameco’s market capitalization had recovered to approximately $22–$25 billion USD. Investors who purchased at the 2018 trough near $12–$14 CAD per share had earned returns of approximately 4–5x their investment over six years — not from a business transformation or a competitive reinvention, but from the market finally pricing what had been in the ground all along.

The business owner who looked at Cameco in 2018 and calculated “455 million pounds times $15/lb net margin at a $40/lb price scenario = $6.8 billion embedded operating income, versus a $4 billion market cap” was not making a forecast. He was reading a reserve report and multiplying by a number. The analyst who built a DCF at $30/lb was making seventeen specific forecasts about uranium prices, production ramp schedules, operating cost inflation, exchange rates, discount rates, and capital spending — every single one of which could be wrong. The business owner’s method was more certain, more grounded, and in this case, more correct.

Subscribe now

Case Study 2 — NexGen Energy and the Arrow Deposit: A Reserve So Large It Bent the Market

NexGen Energy is a Canadian uranium developer with a single asset that has become one of the most debated topics in the uranium investment world. Its Rook I project, centered on the Arrow deposit in Saskatchewan’s Athabasca Basin, hosts proven and probable mineral reserves of 240 million pounds of U₃O₈ at an average grade of 2.37% — with a high-grade core zone (the A2 sub-zone) averaging approximately 15–17% U₃O₈, grades that are extraordinary by any global standard. For context, the global average uranium mine grade is approximately 0.05–0.10%. Arrow’s high-grade core is 150–300 times richer than the average uranium mine in the world.

The economics disclosed in NexGen’s 2021 NI 43-101 feasibility study are specific and remarkable: average cash operating costs of approximately US $4.36–$6.70 per pound over the mine life. At a $71.75/lb spot price, the net operating margin per pound ranges from $65 to $67 on the high-grade core material. Applied to 240 million pounds of probable reserves: embedded gross operating income of approximately $15.4–$16.1 billion from a single mine. The company’s entire market capitalization as of early 2026 was approximately $4–$5 billion USD — less than a third of the operating income embedded in its certified reserves.

The financial community’s persistent difficulty with NexGen is that it has no revenue. The mine is not yet in production. A DCF analyst looks at NexGen and sees years of negative cash flow ahead (construction costs, development capital) before any production begins. They discount those future cash flows at 8%, 10%, or 12% depending on their risk adjustment, layer in uranium price assumptions for 2028 and beyond, and arrive at a valuation. Because the mine is not producing, the numerator of any income-based calculation is zero or negative for several years. Most financial models for NexGen have therefore either (a) severely undervalued the company during periods of uranium market pessimism, or (b) struggled to justify the company’s market capitalization against any conventional metric when uranium prices are low.

The business owner’s approach cuts through this entirely. Arrow’s 240 million pounds of probable reserves are in the ground right now. They have been drilled, certified, and published in a publicly available NI 43-101 report. The operating cost — derived from a feasibility study conducted by independent engineering firms — is approximately $6.70/lb at the life-of-mine average. At $71.75/lb spot price, net operating margin is $65.05/lb. The entire reserve base is worth $15.6 billion in embedded operating income at today’s price, today’s cost. The question is not “will this be worth something in the future” — it already is worth $15.6 billion. The question is only whether the market is paying full price for it now.

As of early 2026, with a market cap of approximately $4–$5 billion, the answer is clearly no. The market is paying roughly 30 cents on the dollar for the embedded reserve value. The construction risk is real — Arrow has not yet been built — but the ore body is certified. The economic framework is established. The regulatory approval process advanced materially in March 2026 when NexGen received construction approval from the provincial government of Saskatchewan. The environmental assessment is complete. The mine will be built.

NexGen received its construction go-ahead in early 2026. The company is targeting a production start in the late 2020s. By the time Arrow reaches full production — approximately 30 million pounds of U₃O₈ per year in the first half of its mine life — it will represent approximately 16% of current global uranium mine supply from a single deposit. At any uranium price above $30/lb, the mine is extremely profitable. At the $71.75/lb spot price in mid-2025, it would be generating approximately $1.9 billion per year in gross operating income from a single operation. The market cap of $4–$5 billion, measured against these numbers, reflects a discount that can be explained only by the systematic failure of conventional valuation to account for what is in the ground.

The Arrow deposit has contained the same 240 million pounds since it was drilled out between 2014 and 2019. The grade has been confirmed repeatedly by independent engineers. The operating cost estimate has been refined by a rigorous feasibility study. None of these facts require a forecast. A business owner looking at Arrow says: “240 million pounds at $65/lb net margin = $15.6 billion in embedded operating income. I am being asked to pay $4.5 billion for the entire company.” An analyst running a DCF says: “The mine is not yet in production, the uranium price in 2030 is uncertain, and my discount rate adjusts for political risk in Saskatchewan.” The analyst’s model produces a number that looks precise but is built on a scaffold of assumptions. The business owner’s calculation is based on what already exists.

Subscribe now

Case Study 3 — The 2021 Sprott Intervention: When a Physical Trust Exposed the Market Structure

By mid-2021, the uranium spot market was a ghost town. The price had languished near $30/lb for three years. Kazatomprom had cut production. Cameco had shuttered its flagship mines. The Megatons to Megawatts program had been over for eight years, and its secondary supply contribution was gone. Yet prices remained stubbornly depressed, and uranium mining companies traded at fractions of their in-ground reserve values, because financial analysts continued to model forward prices at $35–$45/lb and discount everything back at rates that made even Cameco look only modestly undervalued.

The spot market’s available inventory — the pounds of physical uranium oxide available for purchase in the spot market at any given time — was estimated at approximately 23 million pounds in August 2021. This is a remarkably small number. The global reactor fleet consumes 180 million pounds per year, meaning the spot market held roughly six weeks of global consumption. The market was structurally tight but nobody was forcing the price discovery.

In July 2021, Sprott Asset Management acquired the Uranium Participation Corporation (a pre-existing Canadian physical uranium vehicle) and reorganized it into the Sprott Physical Uranium Trust (SPUT), listed on the Toronto Stock Exchange. SPUT’s mechanism was simple: issue trust units to investors, use the proceeds to purchase physical uranium in the spot market, and store it indefinitely. Unlike a futures contract or a mining stock, SPUT actually took physical delivery of the uranium — removing it from available spot inventory and holding it in licensed storage facilities.

The effect was immediate and substantial. Within weeks of its launch, SPUT had begun purchasing uranium at a rate that shocked the small spot market. By August 2022, SPUT had purchased approximately 57 million pounds of U₃O₈ — more than twice the spot market’s entire available inventory at the trust’s launch, purchased over roughly twelve months. Total capital deployed: approximately $2.77 billion, representing approximately 250 individual purchases from 28 different counterparties. By mid-2025, SPUT’s total holdings had grown to more than 60 million pounds — without selling or lending a single pound.

The spot price response was direct: from approximately $28–$30/lb at the trust’s July 2021 launch, uranium prices had risen to $50/lb by late 2021, $63/lb by mid-2022, and eventually broke above $100/lb in January 2024 for the first time in 17 years, briefly touching $106/lb in February 2024 — a level last seen in 2007 when the Cigar Lake mine flood triggered the previous bull market.

Uranium mining stocks moved with prices. Cameco shares went from approximately CAD $20 in mid-2021 to approximately CAD $65 by late 2024 — a 225% return. Uranium Energy Corp, an ISR-focused US producer, went from approximately $3 per share to peaks above $8 during the same period. NexGen Energy went from approximately $4 to peaks above $10. The entire sector rerated, not because earnings estimates were revised upward by financial models (the companies were not yet producing at full capacity), but because the physical market demonstrated, with actual purchases at documented prices, that the uranium in the ground was worth more than the models said.

The Sprott intervention reveals something fundamental about the uranium market that every DCF-based valuation misses: the spot market is structurally thin, and it only takes one large actor buying physical material to move prices dramatically. The reserve value embedded in Cameco’s 433 million pounds, or NexGen’s 240 million pounds, or Kazatomprom’s billions of pounds, existed before Sprott existed. Sprott did not create value — it revealed it. The business owner calculating pounds times net margin per pound in 2019 or 2020 was sitting on the same answer that the market eventually reached in 2023 and 2024. The gap between the two was patience, not analysis.

Subscribe now

Part Four: The Structural Supply Deficit — Why the Business Owner’s Math Gets Better Over Time

The critical quantitative fact about the uranium market as of 2025 is this: the world’s 439 operating nuclear reactors require approximately 180 million pounds of U₃O₈ per year to operate. Global primary mine production — uranium actually coming out of the ground — was approximately 130–140 million pounds in 2023, the most recent year with complete data. The deficit between demand and primary supply is therefore approximately 40–50 million pounds per year, or roughly 25–28% of demand.

This deficit is currently bridged by secondary sources: utility inventories accumulated during the post-Fukushima low-price era, enrichment tails (reprocessed from old enrichment waste), and Russian state inventories — none of which are infinite and all of which are declining in availability. The US Energy Information Administration’s 2024 report stated explicitly that American nuclear utilities face a growing uranium supply gap over the next decade, with the shortfall expected to reach 184 million pounds — more than three years of US reactor consumption — without new domestic supply.

No new tier-one uranium mine has entered production since Cigar Lake came online in 2014. Lead times for new mine development, including exploration, feasibility studies, permitting, and construction, are typically 10–15 years for greenfield projects in even favorable jurisdictions. This means the supply that will be available to the world’s reactors in 2030 is largely already determined: it will come from mines that are either already operating, already in advanced development, or — for a small fraction — being fast-tracked from advanced exploration. There is no scenario in which the supply deficit resolves quickly.

For the business owner who owns certified reserves in a producing or near-producing mine, this structural supply deficit is not a forecast — it is a documented arithmetic fact. The reserves are more valuable precisely because supply cannot easily expand to meet demand.

Subscribe now

Part Five: Confirmed Catalysts — What Is Already Happening That the Market Has Not Fully Priced

Catalyst 1: The Russian Uranium Import Ban — Legislation Enacted, Not Proposed

On May 13, 2024, President Biden signed into law the Prohibiting Russian Uranium Imports Act. This legislation bans the import of Russian uranium into the United States with limited exception waivers for utilities that have no alternative supply, phased out completely by 2028. This is not a proposal. It is not a policy recommendation. It is enacted law.

The significance for reserve valuation is direct and quantifiable. The United States currently operates approximately 93 nuclear reactors, requiring approximately 40–45 million pounds of uranium annually. Russia and Russian-controlled entities (Rosatom and its subsidiaries) supplied approximately 12 million pounds of uranium equivalent annually to US utilities as recently as 2022, plus additional conversion and enrichment services. That supply must now be replaced from non-Russian sources.

The Western producers with proven reserves in geopolitically stable jurisdictions — Cameco in Canada, Energy Fuels and Uranium Energy Corp in the US, Paladin Energy and Boss Energy in Australia — are the direct beneficiaries. Each pound of demand redirected away from Rosatom is a pound that must be contracted from Western producers. The term price — the multi-year contract price above spot — has responded: as of late 2024, long-term uranium contract prices were approximately $70–$80/lb, meaningfully above spot. Cameco’s disclosed contracted position in 2024 included deliveries at prices “in the $70s” per pound for a substantial portion of its output.

This catalyst does not require uranium prices to go higher to increase reserve value. It requires only that existing Western reserves become the exclusive or primary source of supply for a market that was previously served by Russian supply. The Western reserve base is finite. The demand redirected from Russia is 12 million pounds per year minimum — permanent, legislated, and growing as the 2028 sunset on any waivers approaches.

Subscribe now

Catalyst 2: Small Modular Reactors — The First US Design Certification and the Analog to Shale Technology

The US Nuclear Regulatory Commission issued its first design certification for a Small Modular Reactor in January 2023, approving the NuScale VOYGR design. This is a regulatory milestone that took more than a decade to reach. A design certification means the NRC has formally determined that the reactor design meets all US safety standards — the most important regulatory gate in the SMR development process.

The technology parallel to hydraulic fracturing in the Permian Basin is precise. Before commercial shale technology was deployed around 2008–2012, the Permian Basin contained the same oil it contains today. The hydrocarbons were there. They were just economically inaccessible. Horizontal drilling and multi-stage hydraulic fracturing reduced breakeven production costs from $80+/bbl to $35–$45/bbl, fundamentally changing the economics of accessing the existing resource. Pioneer Natural Resources’ Midland Basin acreage, acquired at $2,000–$5,000 per acre in the 2010–2015 period, was valued at $30,000–$50,000 per acre following the technology’s commercial deployment. The acreage existed the whole time. The technology made it worth 10–25x more.

SMRs create a directly analogous situation for specific categories of uranium reserves that are currently sub-economic. In-situ recovery (ISR) uranium deposits in the United States — concentrated in Wyoming and South Dakota — are amenable to production at cash costs of approximately $15–$25/lb. At current spot prices of $71.75/lb, these operations are already economic. But the impact of SMRs is felt on the demand side: the IEA has projected that 100–200 GW of new nuclear capacity could be brought online by 2040 through SMR deployment, which would increase annual uranium demand by 30–50 million pounds — a 17–28% increase on a market already running at a 25% supply deficit.

For companies like Uranium Energy Corp, which holds licensed ISR wellfields and processing facilities in Wyoming that can restart production within 6–12 months of a sustained price signal, the SMR catalyst unlocks reserve value without any incremental capital investment. The wellfields are drilled. The processing plant exists. The permits are valid. The uranium is in the ground. A $90–$120/lb uranium environment — which the IEA’s demand projections would likely produce — would make these assets worth 2–3x their current balance sheet value, which is carried at historical development cost far below economic value.

Subscribe now

Catalyst 3: Data Center Power Demand and the Re-Rating of Nuclear as Baseload Technology

The emergence of large-scale AI data centers has created a power demand problem that the existing electric grid cannot solve cleanly. A single hyperscale data center requires 100–500 MW of continuous, reliable power — not intermittent power from wind or solar, but baseload power that runs 24 hours a day regardless of weather. Nuclear power is the only zero-carbon technology that meets this requirement at scale.

Three Mile Island Unit 1, shut down in September 2019 after becoming uneconomic during the years of cheap natural gas, was contracted to restart specifically to power Microsoft’s data centers in the region, with a 20-year power purchase agreement announced in September 2023. The Palisades nuclear plant in Michigan — shut down in May 2022 — received a $1.5 billion federal Department of Energy loan in 2024 to facilitate the first reactor restart in US history. The Department of Energy under multiple administrations has explicitly included nuclear power in its clean energy strategy.

These are not exploratory discussions. They are operational decisions backed by committed capital. Each additional operating reactor hour consumes uranium. Each reactor restart or new reactor construction triggers a procurement cycle in which the utility must contract for uranium supply — typically 3–10 years in advance of fuel loading. The procurement cycle for nuclear fuel means that demand from reactors entering service in 2028–2032 must be contracted from roughly 2025–2027. That contracting wave is already underway. Cameco disclosed in its 2022 Sprott investor presentation that it had contracted 75 million pounds of uranium from the beginning of 2021 through mid-2022 alone — more new long-term contracting in 18 months than the prior decade combined.

Subscribe now

Part Six: What Wall Street Gets Wrong and Why It Keeps Getting It Wrong

To understand the persistent mispricing of uranium reserves, it helps to understand specifically how financial analysts approach the sector and where their methodology systematically fails.

The standard framework is a discounted cash flow model. The analyst forecasts uranium prices for 5, 10, and 20 years. They apply those prices to a production schedule. They subtract operating costs, capital expenditures, royalties, and taxes. They discount the residual cash flows at a weighted average cost of capital. The result is a net present value per share that becomes the price target.

Every step in this process introduces error in a direction that systematically undervalues the asset:

On price forecasting: Uranium analysts who produced price decks in 2014, 2015, 2016, and 2017 universally forecast long-term prices in the $40–$55/lb range. The actual price in January 2024 was $106/lb. No model built in 2015 with a $45/lb long-term price assumption would have produced a Cameco price target higher than CAD $20. The actual price in early 2025 was CAD $60+. The forecast was wrong not because the analysts were incompetent but because uranium price prediction requires knowing when utilities will enter a contracting cycle, when governments will build reactors, when secondary supply will be exhausted, and when producers will reduce output — all of which are unknowable with precision.

On discount rates: An analyst who applies a 10% discount rate to a uranium reserve in the Athabasca Basin is making an implicit claim that the uranium will be worth 10% less in present value terms for every year it remains in the ground rather than being sold. This is an arbitrary assumption. The uranium does not disappear. Its value does not naturally erode. The grade does not decline. The discount rate is a device for handling uncertainty, but when the uncertainty itself has a directional bias (structural supply deficit, legislated demand from non-Russian sources, SMR deployment), applying a symmetric discount rate as if the future is equally likely to be better or worse than today is itself an error.

On earnings-based metrics: P/E ratios are essentially useless for uranium miners. During a price trough, earnings are negative or minimal — not because the company’s assets have diminished in value, but because extracting uranium at $25/lb cost and selling it at $28/lb is unprofitable. Applying a P/E screen to uranium companies during a price trough produces the perverse result that the companies with the best assets (Cameco, which shut mines rather than sell at a loss) look the worst on earnings-based metrics.

On the DCF’s fundamental assumption: A DCF model assumes the correct valuation of a resource company is the present value of future cash flows from extracting the resource over its mine life. But the business owner knows something the model does not: if you have pounds in the ground that cannot be replicated at any cost by any competitor, waiting for a higher price is free. The reserves do not expire. The uranium does not evaporate. A company that holds 433 million pounds of the world’s highest-grade uranium can afford to wait for $80/lb, $90/lb, or $100/lb — because the resource will still be there when those prices arrive. The DCF penalizes waiting. The business owner rewards it.

The 2018 Cameco decision to shutter McArthur River and buy uranium on the spot market rather than mine at a loss is the purest possible demonstration of this principle. Management was explicitly declining to sell at the prevailing price because they judged the reserves to be worth more than the market offered. They were correct. The financial community punished the stock for the production guidance cut. The business owner recognized it as a rational decision to preserve the value of an irreplaceable asset.

Subscribe now

Part Seven: A Note on Thorium

Thorium deserves mention in the context of uranium reserve valuation — not as an imminent substitute, but as a context for understanding the supply scarcity argument in uranium.

Thorium (Th-232) is approximately three to four times more abundant in the earth’s crust than uranium and is distributed more widely across a broader range of geological formations. India has very large thorium reserves (approximately 300,000 tonnes, among the world’s largest) and has pursued thorium reactor development as a deliberate energy security strategy since the 1970s. Advanced reactor designs — including the Molten Salt Reactor and the Liquid Fluoride Thorium Reactor — can use thorium as a fuel source through a conversion cycle that produces uranium-233 as an intermediate fission fuel.

The relevance to uranium reserve valuation is this: thorium cannot currently be used directly in any of the world’s 439 operating nuclear reactors. Every light water reactor (the dominant global reactor design) uses enriched uranium fuel. There is no existing commercial infrastructure for thorium fuel fabrication, processing, or reprocessing. A thorium-capable reactor fleet would require massive new investment in a fuel cycle that does not yet commercially exist.

In the next 10–15 years, thorium poses no threat to uranium demand. The existing reactor fleet will consume uranium. The reactors currently under construction (approximately 60 globally) are uranium-fueled light water reactors. The SMR designs currently in advanced regulatory review are uranium-fueled. Any world in which thorium meaningfully displaces uranium demand is a world in which entirely new reactor designs have been commercialized, built, and deployed at sufficient scale to matter — a process requiring at minimum two to three decades under the most optimistic assumptions.

For the purpose of valuing uranium reserves today, thorium’s abundance is theoretically relevant but practically irrelevant. It is like observing that there is oil in Venezuela to undermine the reserve value of Permian Basin production — theoretically true and practically immaterial to what the existing, accessible, certifiably-producing reserve is worth right now.

Subscribe now

Part Eight: The Business Owner’s Checklist

After this analysis, a business owner evaluating uranium reserves would look for four things — none of which require a financial model:

One: Pounds in the ground, certified to NI 43-101 or JORC standard. Not resources — reserves. Resources are geological estimates; reserves have been demonstrated economically viable at a stated price. Probable reserves (the lower of the two reserve categories) are still subject to a feasibility-level economic analysis. A company with 200 million pounds of measured and indicated resources may have 100 million pounds of probable reserves. Use the reserves.

Two: Cash operating cost per pound, as disclosed in the most recent feasibility study or annual report. For operating mines, this is the C1 cash cost or equivalent, disclosed regularly in financial statements and supplementary data. For development projects, this is the life-of-mine operating cost from the feasibility study. Compare this to the current spot price and the current long-term contract price. The margin between the two is the embedded operating income per pound.

Three: Multiply pounds by the net margin. Reserves times net margin per pound equals the embedded operating income in the reserve base. Compare this to the Enterprise Value of the company (market cap plus net debt). Any company where the Enterprise Value is below the embedded operating income of its proved reserve base is selling below reserve value — a state that has historically resolved through price recovery.

Four: Check the jurisdictional and political risk. Saskatchewan (Cameco, NexGen) is perhaps the world’s most stable mining jurisdiction. Kazakhstan (Kazatomprom) carries some geopolitical risk. Australia (Paladin, Boss Energy) has strong rule of law and established mining codes. The United States (Energy Fuels, Uranium Energy Corp) has domestic security legislation actively incentivizing domestic uranium production. Jurisdictional risk adjusts the value — but it does not change the pounds.

Subscribe now

Conclusion: The Ground Doesn’t Lie

There is a sentence in Warren Buffett’s 2009 shareholder letter that captures the essence of what this paper is arguing. He wrote: “Price is what you pay, value is what you get.” In uranium, what you get — when you buy a company — is a specific, certified, engineered quantity of a material that the world’s nuclear reactors will consume indefinitely, produced at a specific, documented cost, in a specific, geologically unique deposit that cannot be moved, replicated, or substituted.

Financial models can discount that. Wall Street can ignore it during bear markets. Analysts can apply arbitrary price assumptions and arrive at price targets that suggest these companies are fairly valued or even expensive relative to current earnings. None of that changes how many pounds are in the ground.

The 2016–2021 trough in uranium produced companies trading below what a straightforward multiplication of pounds times net margin per pound justified. Cameco traded below the embedded value of its McArthur River reserves alone. The gap was closed — slowly, then all at once — as the structural supply deficit became undeniable, as Kazatomprom and Cameco cut production, as SPUT removed physical inventory from the market, as the Russian import ban legislated a demand shift toward Western producers, as data centers created new baseload power demand that only nuclear could reliably meet, and as the first SMR design certification signaled the technology pathway to meaningfully higher uranium demand in the 2030s.

The lessons from this cycle compound: the right answer in 2016 was not a discounted cash flow model. It was a reserve report, a cost disclosure, and a multiplication. The pounds were in the ground. The costs were documented. The price would recover — because it always has, because the reactors have no alternative fuel, and because supply cannot expand quickly enough to meet even existing demand, let alone the demand being created by energy transition, data centers, and military procurement priorities.

The ground doesn’t lie. The spreadsheet does.

All data referenced reflects publicly available information from company annual reports, NI 43-101 feasibility studies, SEC filings, World Nuclear Association publications, Sprott Asset Management disclosures, and documented market transactions as cited throughout. No investment advice is implied or intended.

If you like this read, check out my other writings 👇