The original bitcoin treasury
In December 2013, Michael Saylor tweeted: "#Bitcoin days are numbered. It seems like just a matter of time before it suffers the same fate as online gambling." The tweet is still up, which we respect.
Seven years later, on 11 August 2020, MicroStrategy announced it had bought 21,454 bitcoin for an aggregate $250m and adopted bitcoin as its primary treasury reserve asset — the first publicly traded company to do so. Everything that followed, including the several dozen imitators now collectively described as , descends from that filing.
This was not the company's first experiment in aggressive corporate risk. In March 2000 MicroStrategy restated its 1997, 1998 and 1999 results, erasing roughly $66m of revenue and converting reported profits into losses. The stock fell about 62% in a single day on 20 March 2000, from around $260 to $86, and Saylor's paper wealth fell by more than $6bn in that session. The SEC alleged improper revenue recognition, and in December 2000 the case settled without admission or denial: $10m of disgorgement and $350,000 penalties each from three executives, about $11m in total, with Saylor personally paying $8.28m.
So the pattern is not new. What changed in 2020 was the asset.
The volatility harvesting machine
Strategy bought most of its bitcoin with money it borrowed at a 0% coupon, which sounds like a magic trick and is instead an options trade wearing a bond costume.
A convertible note bundles a plain cash loan with an embedded call option on the issuer's stock. Investors accepted no interest because the option was worth more to them than the coupon they gave up, particularly when MSTR's implied volatility was extraordinary. Conversion prices were set well above the share price at issuance, so the company got cheap cash immediately and only handed over equity if the stock rallied hard.
The crucial part is who bought them. Not credit investors making a judgement on Strategy's ability to repay, but convertible arbitrage hedge funds — long the bond, short the stock against it, delta-hedged and rebalanced as MSTR moved. Their trade was to buy the volatility embedded in the convert cheaply and harvest it by trading the shares. That means the demand for Strategy's debt was never really demand for Strategy's credit. It was demand for MSTR's volatility, which is a different thing and disappears under different conditions. This is a textbook piece of : the risk sits in hedge fund books rather than on a bank balance sheet, and nobody publishes the aggregate.
As of the most recent 2026 filings, convertible debt stands at $6.71bn across 0% notes maturing in 2027, 2028, 2029, 2030, 2031 and 2032, after the company repurchased $1.50bn of the 2029s for about $1.38bn of cash, cutting the convert stack from $8.21bn. Preferred stock outstanding totals $3.65bn: 6.00% Series A Perpetual Strike, 10.00% Series A Perpetual Stride, and 8.50% Series A Perpetual Stretch, alongside the STRC preferred against which the company launched a $1.0bn repurchase programme.
The preferred is the part that bites. The converts pay nothing. The preferred pays cash, forever, in dollars.
The premium collapses
The whole structure rested on one number: the multiple of net asset value, or mNAV — what the equity is worth divided by what the bitcoin behind it is worth. At its peak in late 2024 that multiple was about 3.4x. Investors were paying three dollars and forty cents for a dollar of coin, on the theory that management would use the premium to buy more coin and raise bitcoin-per-share, which would justify the premium, which would let them issue again.
In late June 2026 it broke below parity for the first time ever — one account puts the crossing at 27 June, another has it touching about 0.99x in late June. The measures now disagree depending on definition: reporting around 10 August 2026 put Strategy's market capitalisation at $38.37bn and the multiple at roughly 1.06x, with other recent coverage citing 1.04x–1.06x on an enterprise-style basis and 0.68x–0.70x on a basic market-cap measure. Which number you believe depends on the definition used. Nobody has settled it.
There was a dress rehearsal. Between the November 2021 bitcoin high near $68,945 and 21 November 2022, the coin fell about 77% to $15,790 while MSTR fell about 81%, from $81.63 to $15.72 split-adjusted. For FY 2022 the company reported $1.29bn of bitcoin impairment charges and a $1.47bn net loss, ending the year with 132,500 coins. The wrapper fell further than the asset then, too.
Why it matters to this crash
Here is the arithmetic that makes Strategy a mechanism rather than a curiosity. Annual preferred dividends plus interest expense run at roughly $1.76bn. The software business — the actual company, the one that sells business intelligence tools — produced $122.4m of revenue in Q2 2026, which annualises to about $490m. The obligations are around 3.6 times the revenue of the operating business, a shortfall of roughly $1.27bn a year. Bitcoin, whatever else it does, generates no cash.
That leaves three ways to pay: sell stock, sell coins, or refinance. Strategy is currently doing the first two, in public, one Monday filing at a time.
We have tracked the reversal weekly. The company has bought no bitcoin since mid-June. Over five weeks to mid-August 2026 it sold about $2.1bn of common stock, used the proceeds to repurchase roughly $347m of STRC preferred and build a cash reserve to $4.8bn, and sold $213.3m of bitcoin (Bloomberg). In one instance it sold 1,690 bitcoin at an average $64,262 and used the $108.6m to retire 1,152,020 STRC shares. We wrote that up as the flywheel spinning backwards and again as a preferred-stock retirement vehicle, because that is now the honest description of the business: common equity is being issued to retire the fixed claim, which means holder is funding the senior holder's exit.
The company's own disclosure frames the runway in months. Its expected payments were covered by the USD reserve for roughly 16 months as of 30 June 2026, and 26 months as of 24 July 2026. That is management telling you, in a filing, how long it can do this before something else has to happen.
For six years Strategy was the largest marginal buyer of bitcoin. Caroline Mauron of Orbit Markets put the new state of affairs to Bloomberg exactly: "Strategy's trading is pro-cyclical. They are able to buy more when it goes up, and they are forced to sell when it goes down." Bitcoin's 30-day realised volatility has since fallen to a three-year low, which is what a market missing its biggest price-insensitive bidder looks like. The holdings are still enormous: about $58–59bn of coin, 840,447 bitcoin as of the 8-K covering 10–16 August 2026, at an average cost of $75,385 apiece, against a coin price that spent August in the $64,000s before rallying to $72,630 by 20 August.
So far this is orderly. $213m of sales against $59bn of holdings is a rounding error, and we said so at the time in Strategy is running the flywheel backwards. It is the good version of a leveraged holder deleveraging: cash raised, claims retired, no forced seller. It is also the clearest live demonstration that the model dozens of companies copied only runs in one direction.
What would make this dangerous
The strangest thing in the file is that the equity does not appear to have accepted any of it. Strategy rose 12.64% on 18 August 2026 to $104.21, its highest of the period, and was up 15.7% over the five days to 20 August 2026, while the company was a net seller of the asset shareholders are notionally buying it for. Over a year the stock is down about 73% against bitcoin's 45%, so the long-run repricing is happening; the short-run bid is not. We flagged the disagreement and lean to the boring explanation, that MSTR is now pure beta and a 7% day in the coin mechanically produces a 12% day in the wrapper regardless of what the treasury is doing.
The specific things that would turn this from a controlled unwind into something else:
A gap-down repricing of the equity to the low mNAV measure. If the market resolves the 1.06x-versus-0.70x argument in favour of the lower number, the stock has a long way to fall in a short time, and the delta hedges behind $6.71bn of converts have to be adjusted into that . Convertible arb is a hedged trade until the hedge has to be rebalanced by everyone at once.
Volatility staying low. The converts were bought by people buying MSTR volatility. Bitcoin's 30-day realised vol is at a three-year low. If that persists and feeds through to MSTR, the cheapest financing channel the company ever had gets expensive or closes, and the refinancing option in the list of three shrinks to two.
Coin sales scaling. $213.3m is nothing. A quarter where the sales run at the $1.27bn annual shortfall, into a market that no longer has its largest structural bidder, is a different picture. Watch the weekly 8-Ks for the size, not the fact.
The cash reserve running down. The company said 26 months of coverage as of 24 July 2026. That clock is public and it only runs one way unless equity issuance or coin sales top it up.
The 2027 convert. The first of the 0% notes matures in 2027. A maturity is a date on which the option either converts or has to be paid in cash, and the conversion prices were set well above the share price at issuance.
What would make it systemic rather than merely painful is the imitators. Strategy is the template, and the ether treasury companies — SharpLink and BitMine among them — were at their period highs on the same August days Strategy was. If the template gets repriced, it gets repriced for all of them at once.