Twenty point nine billion dollars. That is the digital asset gain Strategy — the company the market still calls MicroStrategy — expects to record for the third quarter of 2025. It reverses the second quarter's loss. On the surface it is the largest quarterly profit any corporate bitcoin holder has ever reported, and the press release wrote itself.
Now read the footnote. Over that same quarter, Strategy bought 334 bitcoin. Not 3,340. Three hundred and thirty-four. The two weeks before had seen 1,665 and 950. The purchase was part-funded by $15.7 million of MSTR stock sales and a repurchase of its own STRC preferred shares. The company drew $142.5 million from its USD Reserve account, leaving $4.88 billion, and its USD Cash balance fell from $1.0 billion to $833.4 million.
A $20.9 billion gain, and a 334-coin week. Those numbers do not sit in the same sentence unless you know what Strategy has become. Liquidity vanishes faster than hype. The hype is at a record. The liquidity is quietly thinning.
Strategy did not become a bitcoin company by accident. It became one by design, and the design is one of the most complex capital structures in public markets — more intricate, in its own way, than anything the protocol layer has shipped in the last five years.
The balance sheet first. Strategy holds roughly 848,000 bitcoin at a cumulative cost of $63.97 billion. That is the largest corporate treasury in the asset class, roughly four times the next-largest public holder. The operating business — the original enterprise software line — is now a rounding error against the bitcoin exposure. When software revenue is smaller than the quarterly change in a single mark-to-market line, you are no longer valuing a software company. You are valuing a leveraged bitcoin closed-end fund with a listing.
The funding runs through three engines, and you cannot read a Strategy filing without separating them. The first is the at-the-market equity engine: Strategy sells MSTR common stock into the open market and uses the proceeds to buy bitcoin, and this quarter $15.7 million of the purchase came directly from that channel. The second is the preferred stock ladder — four tiers, STRC, STRK, STRF and STRD, each with its own dividend profile, each designed to pull in yield-seeking capital without diluting common-share voting control. STRC carries a floating rate and a stated 12% dividend. The third is the buyback stabilizer: when a preferred tier trades below its $100 par value, Strategy buys it back with cash to support the price, and this quarter it did exactly that with STRC.
Three engines, one output: more bitcoin per share, financed by the market's willingness to pay a premium for the vehicle. That willingness is the whole game. Everything else — the FASB fair value accounting, the deferred tax machinery, the dividend schedules — exists to service it.
To see how far this has traveled, look at the arc. The first bitcoin purchase landed in August 2020, framed then as a treasury reserve strategy, a defensive move to protect cash from inflation. Five years later the framing is gone. The software business is vestigial; the treasury is the company. Strategy raised billions across equity and preferred channels, survived the 2022 drawdown that killed Terra and took the market's confidence with it, and positioned itself as the flagship institutional bridge between traditional capital and bitcoin. When the spot ETFs were approved in early 2024, I was working on the traditional side of that bridge in Brussels, integrating fund trading algorithms with institutional custody providers under the incoming MiCA framework. The lesson from that work is that institutions do not buy narratives. They buy structures they can diligence. Strategy's structure is now so layered that diligence itself is a full-time job.
The capital stack has four moving parts, and they interact. Common equity is the highest-beta claim — it absorbs all the bitcoin volatility and pays for it with dilution every time the ATM engine runs. The preferred tiers sit above it, paid first, which is why their dividend is a fixed cost against a variable asset. The bitcoin itself is the collateral and the thesis. The cash and reserve accounts are the shock absorbers, and they are the ones shrinking. The genius of the design is that it lets the company keep buying an appreciating asset with other people's capital while keeping voting control concentrated. The flaw is that every layer of that design assumes the market keeps cooperating.
Now the accounting. This is where the $20.9 billion headline dissolves. Under the FASB fair value standard that Strategy adopted, bitcoin is marked to market every quarter. Q3 bitcoin traded higher than Q2, so the mark moved up. The gain is real in the accounting sense and unrealized in the cash sense. It is a paper gain — a change in a valuation input, not a change in the company's ability to pay a bill.
The larger distortion sits one layer down, in the tax line. In Q2, when bitcoin fell below Strategy's cost basis, the company booked a deferred tax asset of $4.12 billion and applied a full valuation allowance against it — the accounting way of saying "we do not expect to be able to use this." In Q3, bitcoin climbed back above the cost line, so the allowance was released. The effect: income tax expense dropped from $6.00 billion to $1.88 billion. A $4.12 billion swing in the profit and loss statement came from a management judgment about the future, not from a single satoshi changing hands.
Here is why the tax reversal deserves more than a footnote. A deferred tax asset is only worth something if there is future taxable income to shield. Strategy's future taxable income is, for practical purposes, its future bitcoin gains. So the Q3 decision to release the $4.12 billion valuation allowance is really a management assertion that bitcoin will be higher than cost in the periods ahead. That is a directional bet embedded inside an accounting entry — and it is a bet the auditor can challenge. When a single judgment moves the tax line by $4.12 billion, the earnings number is not measuring operations. It is measuring confidence.
Strip the tax reversal and the fair value noise and you are left with the only line that matters operationally: what did the company do with real cash? It bought 334 coins. It bought back STRC. It drained its liquidity buffers.
I have run this play before. In 2017, before the 0x token sale, I led a rapid due diligence sprint on the liquidity aggregation contracts. The whitepaper promised the world; the contracts buckled under high-frequency conditions. I pitched the fund into ZRX anyway, because the technical foundation was sound beneath the marketing — and we exited at 400% six months later. The lesson I carried into every subsequent mandate is simple: never trust the yield; audit the source. The yield here is a 12% preferred dividend and a $20.9 billion gain. The source is a financing machine that depends on the market continuing to pay a premium for the wrapper.
So audit it. The reflexivity loop runs like this. Sell MSTR and preferred shares, raise cash, buy bitcoin, bitcoin rises, book value rises, the premium to net asset value — mNAV — widens, which lets the company sell more shares, which funds more bitcoin. It is elegant. It is also self-referential: the loop is powered not by the underlying asset but by the market's belief in the loop. I watched the same structure at a smaller scale during DeFi Summer 2020, when incentive emissions manufactured APYs that had no source in real yield. I rotated $2 million of capital into stablecoin pairs before the emissions collapsed and preserved 90% of principal while the leveraged cohort was liquidated. The pattern is identical here, only the instrument is a common share instead of a farming token. When the yield is manufactured by the structure rather than earned by the asset, the structure — not the asset — is the risk.
Is this a Ponzi? No, and I will not let the lazy comparison stand. A Ponzi pays old participants with new participants' money and holds nothing real. Strategy holds 848,000 bitcoin, an asset that is liquid, verifiable and saleable. If the machine stops, there is a physical floor. That is the difference between a structural risk and a fraud. But structural risk is not a small thing. The premium — mNAV above 1 — is the oxygen. Every dollar of the flywheel's advantage comes from selling equity at a premium to the assets it buys. If the premium compresses toward 1, "sell shares, buy bitcoin" stops creating value and quietly becomes "sell bitcoin to service preferred dividends." The mechanism does not slow down gracefully. It reverses. Reflexivity cuts both ways, and it cuts faster on the way down.
Consider what the 12% STRC dividend actually costs. A preferred yield that high is not a gift; it is a price. It tells you the market demanded a steep coupon to hold subordinated paper in a vehicle whose solvency is tied to a volatile asset. Every dollar of that coupon is a dollar that does not compound into bitcoin per common share. If the preferred stack grows while bitcoin's growth slows, the dividend burden becomes a slow leak in the very metric — bitcoin per share — that the whole model exists to maximize. The company is funding its upside by selling a claim on its downside, and the price of that claim is rising.
Here is the part the headline buried. The purchase rhythm is decelerating: 950, then 1,665, then 334. Three consecutive weeks, the last one the smallest. You can read that three ways — the capital markets window narrowed, management deliberately slowed, or cash simply got tight. Every one of those readings is cautious. When the engine that is supposed to compound the asset starts sputtering, the number to watch is not the paper gain. It is the pace.
And the buffers are visibly drawing down. USD Cash fell to $833.4 million. The USD Reserve was tapped for $142.5 million to cover dividends and interest, leaving $4.88 billion. The company is paying out to preferred holders and buying back STRC at the same time it is funding a token purchase. That is a liquidity profile being consumed faster than it is being replenished — at least in the quarter we can see.
Then there is the STRC signal, and it is loud. Strategy committed to a 12% dividend "until STRC trades healthily around $100." You do not promise price support for a security that is trading at or above par. That language is a confession: STRC has been trading below its $100 face, and the company is using cash and dividend policy to hold it up. A buyback plus a yield commitment is a stabilizer, and stabilizers exist because something is unstable. The company also proposed paying daily dividends across the four preferred tiers, with a common-shareholder vote set for October 28. Read that carefully. Moving from quarterly to daily payouts increases the frequency of cash leaving the building. Why would a treasury do that? To make the preferred shares more attractive to yield capital — to keep the funding channel open. It is a demand-side tell dressed as a shareholder benefit.
The consensus reading of this quarter is that Strategy is winning — that a $20.9 billion gain proves the model works. The contrarian reading is that this was never a bitcoin story. It is a capital markets story wearing a bitcoin costume.
Start with the decoupling. Bitcoin's price in Q3 did the work. Strategy's management did not. The gain is a passive mark on an asset the company happened to hold. A spot bitcoin ETF would have captured the identical move with no premium, no leverage, no preferred ladder and no tax judgment. That is the uncomfortable comparison Strategy cannot answer: if the goal is bitcoin exposure, the ETF is the cleaner instrument. The only thing Strategy sells that an ETF does not is leverage to the upside and a premium that can compound — and both of those are exactly the features that reverse violently when sentiment turns.
Which brings the competition into focus. Spot ETFs like IBIT now offer regulated, low-fee, no-premium access. Metaplanet in Japan runs the same treasury playbook at smaller scale with an Asia-premium arbitrage. Semler Scientific and a dozen smaller US names copy the template with thinner liquidity. Strategy is the template author, but templates get commoditized. The premium is a trust premium on Strategy's financing ability — and trust premiums are the first thing to erode when a cheaper substitute appears.
There is a quieter threat that almost nobody prices: index inclusion. Strategy's weight in major indices is partly a function of how index providers classify a company whose balance sheet is dominated by a digital asset. If MSCI or S&P revise the rules on digital-asset-heavy constituents, passive flows could be forced out regardless of bitcoin's price. That is an upstream supply cut to the demand side of the flywheel — and it is not something management can hedge.
And then the narrative itself has started to turn. The most telling artifact of this quarter is not the filing. It is the media framing around it: "Strategy Sells $216M in Bitcoin. Is Saylor a Buyer or a Seller Now?" The question did not exist a year ago. The market's mental model is migrating from accumulation to distribution. Narratives in this asset class do not need to be correct to move price — they only need to be repeated. Once "is he selling?" enters the vocabulary, a bad week of purchase data can make it self-fulfilling.
I will add the governance wrinkle, because it belongs in the contrarian case. Saylor posted the $21 billion figure on X before the formal filing. That is a selective-disclosure question under Regulation FD — material information released to a social media audience ahead of the structured channel. It also fits a governance profile where control is concentrated: a dual-class structure gives Saylor voting power far beyond his economic stake. This is not a company with checks and balances around its core strategy. It is a company with a single conviction at the center and a shareholder base that has bought the conviction. When the conviction and the financing window diverge, there is no committee to slow the machine down.
None of this means Strategy collapses next quarter. The funding channel is still open — the fact that stock sales helped finance even a 334-coin purchase proves capital is still willing to enter. That is the single strongest piece of evidence that the reflexivity engine has not yet inverted.
But the early-warning board is lit. Watch four things, and watch them in this order. First, mNAV: the moment the premium compresses toward 1, the flywheel loses its mechanical advantage. Second, the weekly purchase cadence: 334 is a data point, three more small weeks is a trend. Third, the October 28 vote on daily dividends: a yes is a tell that the preferred channel needs propping up. Fourth, the KPMG audit: every number in this article is a preliminary estimate the auditor has not yet reviewed, and a large revision would put a shareholder-litigation spotlight on a quarter built on accounting judgments.
Liquidity vanishes faster than hype. The $20.9 billion is the hype. The 334 coins, the drained reserve and the price support for STRC are the liquidity. In a sideways market, the winners are not the loudest balance sheets. They are the ones whose funding does not depend on the story staying true. Strategy's story is still true today. The question worth holding into next quarter is a simple one: what happens to a reflexivity machine on the day the market stops paying for the reflection?