The number that matters is not 2,305.
Two corporate treasuries — Strategy and Strive — added 2,305 bitcoins between them this week, according to the disclosure that crossed my desk. The headline wrote itself, because the headline always writes itself: institutions keep buying. That is precisely the problem. A quantity of coins acquired is the one figure a company can publish while disclosing nothing that would let you decide whether the purchase created value or merely displaced it.
The number that matters is 846,000.
That is Strategy's disclosed bitcoin position. Every publicly listed company on earth holds 1,273,000 coins. Run the division. One balance sheet controls roughly 66 percent of every bitcoin held in corporate treasuries anywhere. Add the second-largest holder and you are at 68.5 percent. A sector that markets itself as diversified institutional adoption is, at the top, a duopoly with a rounding error attached.
There is no comparable structure in any other asset class. No single ETF issuer holds two-thirds of all ETF assets. No one bank holds two-thirds of all bank deposits. The corporate bitcoin sector is not an index. It is one company, plus a tail of imitators, plus a story.
To understand why that matters, you have to understand what these companies actually are.
A bitcoin treasury company is not a crypto company. It is a capital structure with a bitcoin exposure bolted to it. The operating business, where one exists, is vestigial. The product is the balance sheet. The mechanism runs like this: the company raises money in public markets — through convertible notes, at-the-market equity issuance, or successive series of perpetual preferred stock — and converts that money into bitcoin. If the market values the company at a premium to the bitcoin it holds, the company can issue shares, buy more bitcoin, and increase the amount of bitcoin backing each remaining share. Do that repeatedly and the flywheel spins.
The metric that governs the machine is the premium itself — market capitalization divided by the market value of the bitcoin held. Analysts call it mNAV. Above one, the structure accretes. Below one, it destroys. Everything else — the founder's public profile, the conference keynote, the treasury-strategy manifesto — is decoration sitting on top of that single ratio.
I have watched this mechanism from close range before. During the summer of 2020, when Compound and Aave were handing leverage to retail like samples at a trade show, I modeled the liquidation cascades and published the math. I was told I did not understand DeFi. Six weeks later the cascades arrived on schedule. The lesson I took was not that I had been right. It was that the quantity a market publicizes — total value locked, coins accumulated, wallets onboarded — is almost never the quantity that determines whether a structure survives. What determines survival is the funding.
The sector's own figures deserve a flag before we go further. My own records do not reconcile cleanly with the 846,000 number or the 1,273,000 total. Treasury-position reporting is a lagging exercise, filings arrive in feet, and news copy routinely rounds, double-counts, or inherits a stale snapshot from a database nobody re-verifies. Treat every figure below as an opening position requiring a filing-level check, not a settled fact. I do not publish a number I have not re-derived.
Strive's entry into the top five is the more interesting datapoint. A treasury at 26,355 coins does not reach the top five by buying on the open market at a weekly cadence. Physics does not allow it. The plausible mechanism is consolidation — acquiring another listed vehicle's treasury, or merging the shell and inheriting the coins. That is a different activity from accumulation, and it belongs in a different column of the table.
The concentration is not diversification. It is a single point of failure wearing a sector's clothes.
Start with the arithmetic of the tail. If the sector total is 1,273,000 and Strategy holds 846,000, then everything else — every mining company with a treasury arm, every Metaplanet, every freshly announced treasury strategy from a regional exchange listing — sums to roughly 400,000 coins. That is the competitive field. It is about 31 percent of a category whose entire investment thesis is institutional exposure to bitcoin.
Consider what an allocator is actually buying when they buy the sector. They believe they are getting a basket. They are getting a levered bet on one company's capital-markets access, repackaged fifty ways. If Strategy's premium compresses, the sector's narrative compresses with it, and the tail re-rates downward, because the tail's entire valuation argument is "we are the next Strategy." A tail priced on the leader's multiple has no independent floor.
I have run this pattern before. In 2021 I traced a minting operation called Quantum Cat, followed the 12 ether that left the minting contract within hours of launch, and published the wallet flow. What killed that project was not the theft. It was the discovery that every "partner" listed on its site was downstream of the same two dev wallets. Concentration looks like strength until you map it. Then it looks like a single point of failure with good branding.
The missing denominator is the funding method. Coins acquired tells you nothing. Bitcoin per share tells you everything.
Company A issues 10 million new shares at a premium, buys 1,000 bitcoin, and lifts its bitcoin-per-share from 0.0100 to 0.0106. Accretive. Company B issues 10 million new shares at a discount, buys 1,000 bitcoin, and drops its bitcoin-per-share from 0.0100 to 0.0094. Dilutive.
Both companies report "we acquired 1,000 bitcoin this week." Both press releases read identically. One created shareholder value. One transferred it away from existing holders to whoever bought the new shares. The 2,305-coin figure cannot tell the two apart. It does not disclose the issuance price, the share count, the instrument, or the coupon. A weekly accumulation report without a share count is a revenue figure without a cost line. I do not report on revenue without costs.
My first serious piece of work was a signature-malleability flaw in an exchange protocol's v1 contracts — a nonce-handling defect that permitted double-spending through the relayer. It was patched, eventually, after the developers stopped questioning whether a twenty-something was technically competent. The lesson never changed: the flaw always lives in the part of the system nobody markets. In a treasury company, nobody markets the share count.
The instrument choice carries its own cost, and each option prices differently. Convertible notes issued near a zero coupon are cheap money that converts into equity if the stock rises — the cost is deferred dilution, and it lands precisely when the company is doing well enough for that to matter. At-the-market equity issuance is instant dilution at whatever the tape gives you. Perpetual preferred stock is the most honest of the three, because it states its price out loud.
And here the sector's own yields are the tell. When the yield is too high, the exit is rigged. Preferred series carrying high single-digit or low double-digit dividend rates are not a signal of confidence. They are the price an issuer pays to borrow in an instrument that ranks above the common equity in liquidation, and therefore absorbs risk the common holders are not absorbing. That dividend is a cash obligation. It must be paid in dollars, by a company whose only product is an asset that pays nothing. Follow the money, not the announcement. I trace the wallet, not the whisper — and the wallet that matters here is the share count.
The flywheel is reflexive, and reflexive structures fail from the inside.
The mechanism is self-referential in a specific and dangerous direction. The premium exists because the market believes the company can keep buying. The company can keep buying because the premium lets it issue equity on favourable terms. The favourable terms require the premium. Remove the premium and the advantage inverts: every new share issued below the bitcoin-backing level transfers value away from the people who bought yesterday.
This is the same topology as the UST–LUNA seigniorage loop I dissected in 2021. Different asset, different mechanism, identical shape — a variable that must keep rising in order to sustain the mechanism that makes it rise. I wrote then that the loop was unsustainable. I was early, but I was right about the shape, and the shape determines the terminal state.
The failure mode is not dramatic. It requires no fraud, no hack, no regulatory raid. It requires only a change in mood. The premium narrows, issuance slows, per-share bitcoin stops growing, the narrative weakens, the premium narrows further. Sixty billion dollars of Terra value did not vanish because the code broke. It vanished because the reflexivity reversed and there was no bid underneath.
There is a maturity wall inside this machine as well. Convertible notes carry dates. Preferred series carry redemption terms and dividend schedules. If a cycle turns at the wrong point on that ladder, the company faces a choice between raising capital at a bad price, paying obligations in cash it does not generate, or selling the only asset it holds. None of the weekly accumulation reports mention the ladder. Ask why.
And there is a historical precedent that these companies' marketing departments have never once cited. Closed-end funds — vehicles whose entire purpose is to hold a portfolio and trade as equity — almost universally trade at discounts to net asset value after listing. The average discount runs into the several percent range, and it persists for decades. The initial-premium puzzle is such a well-documented anomaly in academic finance that it has its own literature. A treasury company trading at a large multiple of the value of its coins is, structurally, a closed-end fund running the anomaly in reverse, at a magnitude the literature does not contain. That is not a bull case. That is a question.
The accounting rules have already changed, and the earnings volatility is scheduled.
From fiscal 2025, US GAAP requires crypto assets to be measured at fair value with changes running through net income. In practice, that makes these companies' reported quarterly earnings a levered function of the bitcoin price. In a quarter where bitcoin falls thirty percent, a treasury holding 846,000 coins books a mark-to-market loss in the tens of billions — not because anything happened, but because the calendar turned.
That is bookkeeping, not cash flow. It does not directly threaten solvency. It does matter for three reasons. It makes reported earnings useless as a signal about the business. It feeds screens and indices that consume reported earnings as an input. And it manufactures a quantity of headline volatility the company cannot control and will consistently be blamed for.
The volatility cuts both ways. In a bull market, treasury companies will report earnings that look like those of a leveraged fund. Retail investors who do not read footnotes will treat those numbers as operating performance. They are not. They are the asset price, restated. Hype is the only asset in a vacuum mint — and a fair-value gain with no cash behind it is the purest form of that asset. There is a further wrinkle that almost no retail holder has priced: US tax law now applies a minimum tax to large corporations on book income, which is precisely the number that fair-value accounting inflates. A gain that exists only on paper can still generate a real liability. Nobody puts that in the tweet.
The risk that is priced at zero — index classification.
Index providers classify companies. Classification determines which indices a company enters. Index entry determines which passive funds are forced to buy it. Forced buying determines part of its valuation. A company whose primary activity is holding an appreciating asset and financing itself against it looks, structurally, less like an operating company and more like a closed-end fund.
If classification committees reach that conclusion — and no law prevents them — the result is mechanical: removal from broad-market indices as holding vehicles rather than operating businesses, forced selling by index funds, and a permanent de-rating that has nothing to do with the bitcoin price. The risk is asymmetric in the way that matters. In a bull market nobody discusses it, because flows are inbound and the narrative is all confirmation. In a drawdown, when the committee has cover and the politics are easy, it becomes a live option. Watch the committee, not the chart.
I have seen how quickly an unexamined structural risk converts into an exit. Last year I traced an AI-agent network running fifteen fake influencer accounts, roughly five million dollars of stolen funds, and a shell company registered in Seoul. Every one of those accounts had a convincing identity and a coherent posting history. The failure was never the front end. It was the corporate layer nobody looked at until the money was gone. A profile picture is not a shield against fraud — and a ticker with a famous founder attached is not a balance sheet.
Custody is the largest uninsured exposure in the asset class, and it is a footnote.
Eight hundred forty-six thousand bitcoins. At any plausible price in the current cycle, that is somewhere north of eighty billion dollars, held in a bearer asset. Whoever holds those keys holds them without deposit insurance, without a central-bank backstop, and without the adviser-custody rules that govern trillions in ordinary client assets.
The weekly disclosure does not say who holds them. It does not say whether the keys sit in a qualified custodian's vault or a hardware wallet in an office. It does not say what insurance exists, whether custody is multi-signature, whether the signing quorum is geographically distributed, or how the estate plan reads if the founder is gone tomorrow morning.
Compare that disclosure standard to what a money-market fund must publish about its counterparties, or what a bank must publish about operational risk. The gap is not a technicality. It is the entire risk surface of the position, left unexamined. I learned the lesson about unexamined custody in 2021 — money leaves in hours, and it leaves through the door nobody bothered to lock. The difference here is the size of the door.
Gross accumulation is not net accumulation, and the sellers never make the headline.
Two thousand three hundred and five coins went in. The release implies that is the story. It is not. A market is a flow, not a purchase. If three other corporate treasuries, or an ETF, or a sovereign entity, or a bankruptcy estate sold four thousand coins into the same week, the net flow was negative and the headline was written backwards. The release does not disclose the average purchase price, whether the coins were bought at the top of a local range or into a drawdown, the venues, or the counterparties. It does not say whether any counterparty is itself distressed.
These are not exotic requests. They are the minimum standard for reporting any institutional transaction. Crypto's press cycle grants an exemption that would be unthinkable in equity research. I do not accept the exemption.
Now the part that people who agree with the criticism above will not like.
The bulls are not wrong that the treasury-company model is a genuine innovation. It solves a real problem. Before these vehicles, an institution that wanted bitcoin exposure inside its existing custody, audit, tax, and compliance rails had three bad options: buy spot and manage keys itself, buy a futures-based fund with roll costs, or buy a mining stock and inherit an operating business it did not want. The treasury company gives them a fourth: equity in a public vehicle whose balance sheet is essentially the asset.
That is real. The capital structure is sophisticated. Zero-coupon convertibles, at-the-market programmes, layered preferred — this is the toolkit that closed-end funds and REITs have used for decades, applied to an asset with no cash flow. Doing it without a cash-flow stream is harder than it looks, and the teams who built these structures deserve credit for the engineering, not only for the marketing.
The bulls are also right that concentration cuts both ways. A single holder with a public, repeated commitment never to sell is, for as long as it holds, a supply lock. Roughly four percent of circulating supply has been removed from the float and parked inside a structure whose stated intent is permanence. In a market where issuance is fixed by protocol, that is a durable bid under the price.
And I will admit the part my own record makes uncomfortable. I called the DeFi leverage cycle correctly on mechanism and badly on timing. Being early is indistinguishable from being wrong until the moment it is not. If bitcoin doubles from here and the premium holds, everything above will be filed under bearish blogger, bearish again. That is a real possibility. Structure does not schedule itself. But structure also does not forgive.
Watch four columns that no accumulation report prints. Bitcoin per share, quarter over quarter. The premium, against the value of the coins. The convertible maturity ladder, against cash on hand. And the custody attestation, if one is ever published.
If per-share bitcoin keeps rising while the premium holds, the machine is working and the skeptics are wrong. If per-share bitcoin starts falling while the company keeps announcing record purchases, the machine has inverted, and the only people who will notice before the tape does are the ones reading filings instead of headlines.
Two-thirds of an asset class is not a position. It is a hostage situation with a press office.
Question the number, not the narrative. The number survives the audit.
