On 10 August, Strategy filed its weekly update with the SEC, and it had sold bitcoin again. The count was 1,690 coins, at an average of $64,262, raising $108.6m. That much was expected. What changed is where the money went.
This article covers three things. What the last two filings did to the mix of stock and coins the company funds itself with. Which coins it is selling, which the filings never say outright, though the arithmetic does. And why the open call on this site's scorecard is harder to grade today than it looked in July.
The reason any of it matters is that call, made on 7 July. It says Strategy will keep paying the dividends on its preferred stock out of bitcoin sales. Two things would let it off: a price recovery, or a return to funding itself by selling shares. The call runs to the end of October. Both escape routes are now in play at once, and the two filings do not settle which.
Two warnings before the numbers. The call is the author's own, so this is a progress report on his own homework. And the figures below come from news coverage of the filings, not from the filings themselves. Neither the SEC's own archive nor Strategy's press page would serve the documents to the tools used for this site. Reports agree with each other, the numbers are internally consistent, and every calculation here is shown so a reader can redo it. But the primary documents have not been read. Until they are, treat every company figure below as reported rather than verified.
What the Filings Say
Start with the week of 27 July to 2 August. Strategy sold 1,638 coins at an average of $63,957, raising $104.73m. It split that money almost exactly in half: $52.4m to preferred dividends, $52.3m to buying back its own preferred shares. In the same week it sold 3,011,361 of its ordinary shares for $290.6m net. Stock raised 2.8 times what coins did.
Now the week of 3 to 9 August. Coins sold: 1,690, for $108.6m. Preferred shares bought back: $108.6m. Those two figures are the same to the dollar, which leaves nothing over. Every cent the coins raised went into the buyback, and the dividends were paid from somewhere else. In the same week the company sold 6,585,682 ordinary shares for $653.1m net, of which $650m went straight into its cash reserve. Stock raised six times what coins did.
Across the two weeks together the ratio is 4.4 to one: $943.7m from stock against $213.3m from coins. In the first week of July it ran the other way. The company sold no stock at all that week, and sold its first large batch of coins instead. The mix has inverted. The coins now pay a different bill.
One point of confusion in the coverage is worth a line, because it changes the story. The write-ups present these arrangements as a new policy. They are not. Strategy set the rules out on 29 June, six weeks before either filing and a week before the call on this site was made. That document caps bitcoin sales for the cash reserve at $1.25bn. It permits sales to cover dividends and interest "when deemed more advantageous than equity issuance". It requires the reserve to hold at least twelve months of those payments. It sets aside $1bn to buy back preferred stock and another $1bn for ordinary stock. It also put the monthly dividend rate on the STRC preferred at 12.00% from 1 July. Nothing since has changed any of it. What changed is which lever the company reached for.
Two other movements are worth recording. The cash reserve is filling fast: $3.75bn a fortnight ago, $4.0bn a week later, $4.65bn on 9 August. And the preferred shares are coming back cheap. The first week's buyback took 912,143 shares for $81.2m in total, counting money from stock and coins together, which works out at $89.02 a share. The second week's took 1,152,020 shares at $94.27. Both sit below the $100 the shares are meant to be worth. A company paying a 12% coupon is buying that coupon back at a discount, and coverage puts $785.2m of the allowance still unspent. The discount, though, is shrinking.
Which Coins Went Out
Each week the filing reports two figures that say more side by side than either says alone: how many coins are left, and what they cost on average.
On 2 August: 842,138 coins, average cost $75,419. On 9 August: 840,447 coins, average cost $75,385. Multiply each pair out and the total the company has spent on its pile fell from $63,513m to $63,357m. That is $156m of cost removed, spread across 1,691 coins, which puts the average cost of the coins that left at about $92,300.
That figure should stop a reader. The average coin in the pile cost $75,419. The coins sold cost roughly a fifth more than that. Had the company sold average coins, the average would not have moved at all; it fell, so dearer coins went out the door. At the $64,262 those coins fetched, the loss booked was near $28,055 each, against the $11,157 an average coin would have shown.
Three cautions, because this is an inference and not a disclosure. Rounding does not rescue it: the two averages are given to the dollar, so the implied cost only moves between about $91,800 and $92,800. It does rest on one assumption, that the average purchase price the company discloses follows the particular coins removed. If Strategy computes that figure another way, the calculation falls apart. And no reason is offered here for the choice. Picking dear lots has an obvious tax use, but no filing says so, and guessing at motive is not reporting.
A small mismatch, left in rather than tidied away: the holdings fell by 1,691 coins while the filing reports 1,690 sold. Almost certainly a fraction of a coin lost to rounding. The sum above uses 1,691, the figure the two holdings imply; using 1,690 instead moves the answer by $55.
The Call Is Harder to Grade Than It Looks
Here is the scorecard entry, in the words it was filed under on 7 July. Unless bitcoin recovers above Strategy's cost basis, about $75,500, or the company restores share issuance as its primary funding source, its third-quarter preferred distributions will again be funded at least in part by bitcoin sales.
Two halves of that sentence now point in opposite directions. Share issuance is plainly primary again, at 4.4 to one over the fortnight and six to one in the latest week, which is the escape clause firing. Yet coins are still being sold in size, every week. The resolution rule says the entry misses if the dividends are covered without bitcoin sales while the price stays low. In the week to 9 August, the dividends were indeed covered without bitcoin sales.
So there are two honest readings. On the first, the entry is failing: stock has taken over the funding, and last week's dividends came from somewhere else. On the second, it is holding: bitcoin is still being sold every week to service the preferred stack, and retiring a 12% instrument is servicing it. The wording did not anticipate the money splitting this way, and this article does not choose between the readings.
There is a sharper version of the problem. The company reserved the right to switch, in writing: coins fund dividends "when deemed more advantageous than equity issuance". The entry is therefore predicting the output of a switch that management may flip in any given week. Three questions follow from that, and none of them belongs to this article. Whether the wording is too loose. Whether the voiding condition has fired. Whether the resolution rule needs tightening. Those are calls for the author of the entry, on a public ledger under his own name. It resolves on 31 October.
The other half of the entry, the price half, is far easier to settle, and this is the one place the site's own data does the work. The entry voids if bitcoin trades back above roughly $75,500 for a sustained period. It has not come close. Across the 40 days of the third quarter so far, bitcoin closed below $75,500 on every single one. The nearest approach was $8,979 short, on 22 July. The median gap was $11,554. The last close at or above that line anywhere in sixteen years of daily prices was 27 May. On 9 August, the last day the file covers, bitcoin closed at $64,777, some 14% below. The recovery branch has not been live for a single day of the quarter.
What to Watch
Three numbers, all in the weekly filing, none of them behind a paid terminal. The first is the split of the bitcoin money between dividends and buybacks: half and half two weeks ago, nothing and everything last week. The second is the ratio of stock raised to coins sold, 2.8 and then 6.0, which is the switch itself. The third is the average cost of the pile, $75,385 and drifting down, because every dollar it falls says more dear coins went out the door.
Watch those three and the switch is visible in the filings before any commentary describes it. The coins now pay a different bill, and the same filings will say when they go back to paying the old one.
Where the figures come from: the company numbers are as reported by news coverage of Strategy's weekly filings, not read from the filings themselves, as the warning above says. Every price here is a daily close from this site's own price file, which ends on 9 August 2026.