On 3 August 2026, Strategy Inc. filed a Form 8-K with the Securities and Exchange Commission that looks, at a glance, like housekeeping: a table of share sales, a table of bitcoin holdings, a table of repurchases. The number that matters sits in the middle table. During the week of 27 July to 2 August, the company sold 1,638 bitcoin for $104.73 million, at an average sale price of $63,957 per coin. Two columns to the right, in the same row, sits the average purchase price of everything it still owns: $75,419.
The largest corporate holder of bitcoin in the world sold, and it sold roughly fifteen percent below its own cost. The footnote explains what for. Of the proceeds, $52.4 million funded dividends on the company’s preferred stock and $52.3 million funded repurchases of one particular class of that preferred stock. There is no statement about bitcoin in any of it, no view on the price, no market call. It was a payment instruction.
The same filing discloses that in the same week Strategy sold 3,011,361 of its own class A shares through its at-the-market programme for net proceeds of $290.6 million. The company was selling its asset and its equity simultaneously. And while that was happening, the S&P 500 closed Friday at a record 7,757.64, up 0.62 percent, the Nasdaq Composite rose 1.3 percent to 26,690.62, and bitcoin traded near $65,000, about half its October 2025 peak of roughly $126,000. That is the real story of the week. Not that a bitcoin holder sold, but why it had to, when it plainly did not want to.
What the filing actually says, line by line
The document rewards a slow read, because nearly every line carries information. Holdings after the sale: 842,138 bitcoin, acquired for an aggregate $63.51 billion. The US dollar reserve the company explicitly maintains to support preferred dividends and interest on its debt stood at $4.0 billion on 2 August. Of the $290.6 million raised by selling shares, $250.0 million went straight into that reserve, $28.9 million into preferred repurchases and $11.7 million into the general cash balance. Not one dollar of it bought bitcoin.
What is missing from the tables is more revealing still. The same at-the-market programme is registered to sell four classes of preferred stock as well as common. Against all four, in the column for shares sold, there is a dash. Zero. The window through which this company raised capital for eighteen months was shut that week, at least for the securities the dividends are attached to. Meanwhile it bought back 912,143 shares of its variable-rate preferred for $81.2 million, under a programme announced on 29 June 2026 that still has $893.8 million of preferred capacity and $1.0 billion of common capacity remaining.
The sequence compresses into one uncomfortable sentence: the company sold bitcoin below cost and equity into a falling share price in order to buy back a security it had issued itself.
The engine only ran forwards while the premium lasted
To see why this is structural rather than accidental, you need the mechanic that made the model work in the first place. It has a shorthand now: mNAV, the ratio of market value to the value of the coins held. Above one is a premium. That premium was not decoration. It was the fuel.
The arithmetic is elegant. Issue a share at twice the value of the bitcoin behind it, put every dollar of the proceeds into bitcoin, and bitcoin per share rises for every existing holder even as their ownership percentage falls. Dilution that makes the diluted shareholder richer is a genuinely clever piece of financial engineering, and while it works it is not a trick at all, it is good capital allocation. The metric Strategy coined for it, BTC Yield, measures precisely this.
What many holders underweighted is that the same equation flips sign when the premium goes. Below an mNAV of one, every newly issued share gives away more bitcoin per share than it brings in. The engine does not slow down; it runs backwards. Which is why Metaplanet in Japan, the largest holder outside the United States with around 43,000 coins at an average cost near $104,106, paused purchases and instead launched a buyback of 150 million of its own shares, 13.13 percent of its issued stock, funded by a bitcoin-collateralised credit facility of up to roughly $500 million. A company created to buy bitcoin is now pledging bitcoin to buy its own shares. Same reversal, different time zone.
The swap that decided everything: from maturity to perpetuity
A vanished premium still does not, by itself, force a sale. A company trading below the value of its assets can simply stop issuing shares and wait. Strategy cannot wait, and the reason is in the 30 June quarterly report.
At the end of 2025 the company carried $8.19 billion of convertible notes and preferred stock with a liquidation preference of $8.03 billion. Six months later the converts had shrunk to $6.71 billion while the preferred had grown to $15.46 billion, with shares outstanding nearly doubling from 78,183 to 153,529. In half a year, dated debt was swapped for permanent preferred capital.
On paper this is an improvement, and it was sold as one. A convertible note matures; a perpetual preferred does not. You cannot miss a maturity you do not have, and a company that cannot miss one cannot default on it. That logic is sound. What it leaves out is what the swap costs on the other side. A bond demands money on a date, and a coupon at a handful of dates in between. A perpetual preferred demands money forever, and in the case of the largest class, twice a month. A maturity problem with a date became a cash-flow problem without one.
The cash flow statement shows the scale with unusual clarity. In the first half of 2026, Strategy paid $629.2 million in preferred dividends. In the first half of 2025, it paid $58.1 million. That is more than a tenfold increase in twelve months. Dividends declared in the half ran to $758.2 million. Market estimates put the current annual bill near $1.26 billion. Standing against it is an asset that produces no income whatsoever.
A segment holding $54 billion of assets and earning zero revenue
That is the crux, and this quarter the accounts make it visible for the first time. Effective in the second quarter of 2026, Strategy reports its bitcoin operations as a separate reportable operating segment. Previously they sat in a non-operating corporate and other category. In the segment table, the revenue line for the Bitcoin segment reads the same in every period presented: nil.
Beside it sits the software business, which used to be the whole company and is now its only source of revenue: $122.4 million in the second quarter against $114.5 million a year earlier, $246.7 million for the half, with half-year gross profit of $164.9 million. A solid, modestly growing enterprise. And beside that, a segment carrying roughly $54 billion of assets, earning nothing, and posting a second-quarter net loss of $8.22 billion and a half-year loss of $20.76 billion, against prior-year profits of $10.02 billion and $5.80 billion respectively. Almost all of the swing is the unrealised mark on the coins, $8.32 billion in the quarter alone. Retained earnings of $6.32 billion at the start of the year had become an accumulated deficit of $15.20 billion by 30 June.
The reclassification itself is presented drily, as a matter of segment reporting, and the company offers no rationale. The timing is striking all the same. In early January 2026, index provider MSCI decided not to exclude companies with large digital asset holdings from its indices for the time being, and announced a broader consultation on how to treat non-operating companies that hold such assets as part of their core operations rather than for investment purposes. The shares jumped six percent on the news; JPMorgan had warned that exclusion could trigger billions of dollars of outflows. The operating versus non-operating distinction on which that consultation turns is exactly what the new segment structure answers in the company’s own presentation. Whether that was a motive is not knowable from outside. That it is the central open question for index membership is not in dispute.
Why the coupon rises as the ability to pay falls
The most uncomfortable feedback loop is written into the variable-rate preferred. Its rate was 11.5 percent in June. On 31 July the company said it would hold the rate at 12.00 percent per annum for all semi-monthly periods beginning on or after 16 August, and attached a condition: management does not intend to recommend a change until the security demonstrates sustained trading at or near its $100 stated amount. At the end of July it traded at $89.46. With roughly 104.6 million shares outstanding, that is about $10.46 billion of notional.
The market therefore sets the coupon. When confidence slips and the price falls below par, the rate has to stay high or go higher to pull it back. The cost of capital rises precisely when the capacity to service it falls. This is not malice, it is a design feature, and it explains why part of the bitcoin proceeds went into buying that same preferred back. Every share retired near $89 permanently removes a twelve percent obligation at an eleven percent discount to par. As a use of cash it is the best available. That it is funded by selling the core asset is the point.
One detail sits at the edge of the filing and says a great deal. The company expects the dividends payable on 31 August and 15 September to be characterised, for US federal tax purposes, as non-taxable returns of capital to the extent of a holder’s basis. A distribution becomes a return of capital when there are insufficient earnings and profits behind it. Holders collecting that twelve percent are, in substantial part, receiving their own money back and reducing their cost basis as they do.
The wrapper problem is not confined to one company
The pattern is now visible across the sector, at every size. On 23 July 2026, the Smarter Web Company in the United Kingdom sold 177.89 bitcoin for $11.68 million, explicitly to repay an $11.7 million convertible facility rather than issue 7.71 million new shares, keeping 2,700 coins. Management framed it as preferring balance-sheet flexibility over dilution, which is honest and also tells you the equity route had become the expensive one. That is the same choice Strategy faced, at a thousandth of the scale, resolved the same way.
For an ordinary investor the practical consequence is a question of unintended exposure. Because these vehicles sit inside broad equity indices, a portfolio built entirely from index funds can hold a leveraged, preferred-financed bitcoin position without its owner ever choosing one. That is not a scandal, it is how capitalisation weighting works, and the sums involved are small relative to any diversified index. But it is worth knowing that the MSCI consultation is unresolved, and that the outcome will move these shares independently of bitcoin.
The cleaner comparison is with the instruments that do the same job without the capital structure. A spot bitcoin exchange-traded product holds coins and charges a fee. It has no preferred stock, no perpetual dividend, no premium to defend and no incentive to sell into weakness. It also offers no upside from clever issuance, which was the entire appeal of the treasury model in the years when the premium was wide. Investors who bought the wrapper for the leverage got the leverage. It simply runs both ways.
The counterarguments, and they are stronger than the headline allows
Reading this as a distressed liquidation overstates it considerably. The sales are tiny. In the second quarter, Strategy bought 85,296 bitcoin and sold roughly 1,395 of them. Year to date, disposals total around 3,863 coins, about half a percent of holdings. Anyone calling that a capitulation has not done the arithmetic. Michael Saylor’s own response to the criticism was a piece of careful semantics, that when he said never sell your bitcoin he was speaking as one saver to another and that Strategy is a public company, not his wallet, and he maintains the firm expects to remain a net buyer over the long run. On the second half of that, the numbers currently agree with him.
The structural defence has weight too. Precisely because the preferred has no maturity, there is no date on which anyone can force the company to do anything. Preferred dividends can be suspended; that is painful and expensive, but it is not a default. The $4.0 billion reserve covers a $1.26 billion annual dividend bill for roughly three years on simple arithmetic. The balance sheet still shows $52.56 billion of assets against $7.24 billion of liabilities. This is not an insolvent company. It is a company with a liquidity plan.
Even the drawdown can be put in perspective. A fall from roughly $126,000 in October 2025 to about $65,000 is close to fifty percent, which measured against this asset’s own history is the shallowest cycle decline on record; previous bear markets took 77 to more than 90 percent. And the index question that looked like the single largest risk was deferred by MSCI in January, not decided. A case that a sound balance sheet is being steered through a bad patch can be made with real numbers.
The objection simply misses the point. The question is not whether Strategy can pay. It is where the payments come from. In the first half the company raised $8.25 billion selling common stock and $7.53 billion selling preferred, $15.78 billion together. Weeks later it needed the proceeds of 1,638 coins to cover a $52.4 million dividend. Not because the money was gone, but because the channel it had come through for two years was closed that week. That is the information.
What investors should take from this
None of this is an argument against bitcoin. Bitcoin behaved with complete neutrality throughout: it fell, as it has in every cycle. This is an argument about wrappers, and the lesson generalises to any vehicle that packages an asset inside a share, from commodity funds to listed private equity to property holding companies.
Four questions are enough. First: does the asset inside generate a cash flow? Bitcoin does not; a let building does, and that single difference governs everything downstream. Second: is there an obligation on the liability side that demands cash continuously, and does it have an end date? Perpetual preferred paying twice a month is the most demanding possible answer. Third: does the financing depend on a stock market premium that can vanish? If so, the business model is tied to a mood rather than to an asset. Fourth: who stands ahead of me in the queue? At Strategy, $15.46 billion of liquidation preference ranks ahead of the common, against a common market capitalisation lately around $34.7 billion, with the shares near $95 and roughly 52 percent below this year’s high.
Put those four questions to a vehicle and get an uncomfortable answer to all four, and what you are holding is not leveraged exposure to the underlying asset. It is a credit transaction with that asset posted as collateral. In good years the difference is invisible. It becomes visible at the exact moment a line in a regulatory filing explains that the proceeds from selling the asset went to pay a dividend. At BMInsider we think that is the most useful thing to take from this week, whatever you happen to think of bitcoin itself.
Try TradingView Free for 30 Days
Plus get a $15 discount on your first subscription through this link.
