The document Strategy filed with the SEC on Monday, July 6 was routine in format: a Form 8-K, the same weekly update the company has used for years to report what it bought. This one reported what it sold. Over the seven days ending July 5, the company disposed of 3,588 bitcoin at roughly $60,000 a coin, against the $75,476 it had paid for them on average, and used the money to cover the dividends on its preferred stock.
Strategy, the largest corporate holder of bitcoin, has started selling it at a loss to pay dividends. This article covers three things: what the company sold and why, why selling principal to meet a fixed coupon (the fixed interest-style payment the preferreds are owed) is different in kind from taking profit, and which three public numbers will show, week by week, how the next quarter goes.
One disclosure before the argument. Two pieces on this website, A Coupon Is Not a Cash Flow and Three Reflexive Machines, published in June, described this event before it happened. Part of this article checks those claims against what occurred, which means the author is grading his own homework. The mitigation is that both pieces carry publication dates and exact sentences, quoted below, so the reader can verify the order of events independently. Neither prediction was entered on the scorecard at the time; that gap is addressed at the end, the honest way: with a new forward-looking entry, not a backdated trophy.
What Happened
The facts first, all from the company's own filings. Between June 29 and 30, Strategy sold 1,363 BTC at an average of $59,256. Between July 1 and 5, it sold another 2,225 BTC at an average of $60,773. Total: 3,588 BTC for about $216 million, the largest bitcoin disposal in the company's history. For scale, its previous sale, in late May, was 32 coins.
The company's remaining 843,775 BTC cost it $63.69 billion to accumulate, an average of $75,476 per coin. Every coin in this sale went out the door roughly 20 percent below what the company paid for it on average. The proceeds went to the quarterly dividends on four series of preferred stock (STRF, STRK, STRD, and the euro-denominated STRE), the monthly dividend on a fifth (STRC), and to refilling the cash reserve that fronts those payments, which stood at $2.55 billion on July 5.
Three terms carry the whole story, so here they are in plain language. A preferred stock is a promise: holders get a fixed cash dividend before common shareholders get anything. A cost basis is what you paid; selling below it means realizing a loss rather than a gain. An ATM program ("at the market") lets a company sell new shares of its own stock continuously, which is how Strategy funded almost everything for five years: sell shares at a premium, buy coins, repeat.
Now the sequence of the two weeks, because the order is the story. From June 22 to 28, the ATM sold 12.7 million MSTR shares for $1.15 billion. On June 29, the company announced a "Digital Credit Capital Framework": a formal cash-reserve policy, two $1 billion buyback authorizations, and a board-approved program allowing up to $1.25 billion of bitcoin sales, plus further sales to fund dividends whenever management judges that "more advantageous than issuing class A common stock." The same day, it raised the STRC dividend rate to 12 percent. Then, from June 29 to July 5, the ATM sold nothing, the buybacks bought nothing, and the company sold coins instead.
The backdrop: the same filing disclosed an expected $8.32 billion loss on digital assets for the second quarter, almost all of it unrealized, with the stack's carrying value at $49.67 billion on June 30 against roughly $63.9 billion of cost. The quarter's loss was paper. The sale is the first part of it that is not.
Two Different Sells
There are two kinds of selling, and the market's reaction machinery mostly cannot tell them apart. The first is discretionary: a holder trims a position, usually at a profit, at a moment of its choosing. The second is obligated: a structure owes fixed cash on fixed dates, the cash must come from somewhere, and when the other sources dry up, the calendar chooses the moment. The first kind is a portfolio decision. The second is a funding mechanism running in reverse. Only the second kind matters here, and this was the second kind.
The June piece on treasury-company yield made a specific claim about where this leads. Describing STRC's funding mechanics, it said: the moment issuance freezes, "the structure leans on its reserves and then on selling the underlying." That sentence was published roughly three weeks before the 8-K that matched it. Reserves first: the $2.55 billion USD Reserve fronted the payments. Then it sells the underlying: 3,588 coins, replenishing what the reserve paid out.
The companion piece argued that this is precisely what separates Strategy from Terra-Luna: a discretionary brake. Terra's peg defense was automatic, unbounded, and enforced by code; nobody could stop it, and it destroyed the structure in 72 hours. Strategy's board can choose to sell bitcoin, cut a dividend, or buy back securities, and each lever is bounded and voluntary. That brake has now been pulled, at scale, at a loss. The prediction that the brake existed and would be used is confirmed. So is the uncomfortable half of the same argument: a brake is not a free option. Using it costs principal, and the principal is the entire investment thesis.
Here is the detail that deserves more attention than the sale itself. Under the new program, bitcoin sales fund dividends when management determines it is more advantageous than issuing stock. During the sale week the ATM was available and unused. Read plainly: the people with the most information about MSTR decided that selling bitcoin 20 percent under water was better than selling MSTR shares at current prices. That is a valuation opinion about the stock, delivered by its own management, in the form of a trade.
And the surprise, the one figure in the filings that looks backwards until the mechanics click: on the same day the company set up a program to sell coins to pay its coupons, it raised the biggest coupon. STRC's rate went to 12 percent, effective July 1, to hold the instrument near its $100 anchor. The bill got bigger the same week the funding switched from selling shares to selling the asset. The framework even loosened the anchor's language: the company now says it will not necessarily raise the rate just because STRC trades below its stated amount. A softer promise, at a higher price.
The structural problem is arithmetic, not opinion. A fixed dollar bill divided by a falling price equals more coins. At the cost basis of $75,476, this quarter's $216 million costs about 2,860 coins. At $60,000 it cost 3,588. At half the cost basis it would cost about 5,700. The mechanism sells more of the asset the cheaper the asset gets, which is the definition of pro-cyclical, and it was adopted by choice: the board authorized it, presumably because the alternatives (missed dividends, or heavy dilution below net asset value: selling new shares for less than the bitcoin behind each existing share is worth, which drains value from the shareholders already in) were judged worse. Both things are true. It is the least bad option, and it points the wrong way.
The Bill and the Dials
Scale check, before the burn math: 3,588 coins is 0.4 percent of the stack. One week of selling is a data point, not a regime. The company could resume ATM issuance tomorrow and this article's forward claim would start dying immediately. That is what would make it wrong, and it is stated here so the reader can watch for it.
The bill: one quarter of preferred distributions, plus a month of STRC, consumed roughly $216 million of bitcoin (the filing bundles the dividend payments and the reserve refill together, so treat that as an order of magnitude, not a line item). The reserve policy requires the company to hold at least 12 months of expected dividend and interest obligations in cash, and the reserve stands at $2.55 billion; by the company's own compliance arithmetic, a full year of those obligations fits inside that number. The dedicated sale authorization is $1.25 billion, untouched by this sale, which ran under the separate dividend-funding clause.
The dials, in the order the board is likely to reach for them: the STRC rate, reviewed monthly, which can go down as well as up now that the anchor language is softer; the two $1 billion buyback authorizations, which cannot draw on the cash reserve and would themselves be funded by selling more coins; and the $1.25 billion monetization capacity. Every one of these is discretionary, which is why this remains a corporate-finance squeeze and not a death spiral. Terra had no dials. Strategy has three, and has now proved it will use them. The two June theses resolve together, not against each other: "not Luna" was right, and so was "the risk did not disappear, it moved," because it has now surfaced where the coupon piece said it would: in the principal.
The scorecard entry, stated so it can miss: unless bitcoin recovers above Strategy's cost basis (about $75,500) or the company restores share issuance as its primary funding source, its Q3 2026 preferred distributions will again be funded at least in part by bitcoin sales, disclosed in its 8-K filings by the end of October. Tracked on the predictions scorecard, where a miss will stay as visible as a hit.
What the reader takes from this piece is three numbers, all published weekly in the company's own filings, none requiring a paid terminal. The coin count: 843,775, and whether it keeps falling. The USD Reserve: $2.55 billion, against a policy floor of 12 months of obligations. The STRC rate: 12 percent, and whether the next move is up (more burn) or down (a weaker anchor). When the reserve drains while the coupon rises, the machine is consuming principal faster. When the ATM resumes at size, the squeeze is over. The filings will say which, before any commentary does. "Then it sells the underlying" stopped being a prediction on July 6. What remains is arithmetic, and it updates every week.