On the most recent session, Strategy printed as one of the twenty most active single-name equities on the US tape, ahead of Robinhood Markets. The headline is easy to repeat. It is also a lagging indicator — a receipt for attention already spent, not a signal about what comes next. So I went to the filing instead of the headline, and I read the share count line before I read the revenue line.
The quarterly disclosures show three things that the volume ranking does not. Share count continues to expand through the at-the-market program. The aggregate convertible note and perpetual preferred stack sits in the double-digit billions of dollars. And the metric the company circulates as its core performance indicator — "BTC Yield" — is a non-GAAP ratio, self-computed, self-reported, and not attested by the auditor who signs the financial statements. Meanwhile the software segment that provides the entity with its legal identity generates cash flow immaterial against the claims now ranking senior to the common equity.
Volume is a receipt. The liability stack is the account. Code does not lie; intent does. In capital markets, the function that source code performs on-chain is performed by the disclosure document — and only when someone reads it against the arithmetic rather than against the press release.
The instrument, stated precisely
Strategy, formerly MicroStrategy, is no longer a software company that holds bitcoin. It is a bitcoin treasury vehicle that retains a small software business for legal, tax, and operational continuity. The distinction is not semantic. The operating business contributes a rounding error to enterprise value while carrying the entirety of the legal identity and the reported earnings base.
The model is straightforward to state. Raise capital through three channels — at-the-market common equity issuance, convertible notes, and perpetual preferred stock. Convert the proceeds into spot bitcoin. Report the resulting change in bitcoin per common share. The company calls that change BTC Yield, and it presents the figure as a growth rate.
Two definitions are required before any of the rest of this holds. Modified net asset value, or mNAV, is the ratio of the company's market capitalization to the market value of its coin holdings. BTC Yield, in the company's own framing, is the percentage change in bitcoin per share over a period, adjusted for issuance. One is a valuation ratio. The other is a dilution-adjusted accumulation rate. They are mechanically linked, and the link is the entire story.
I have read a lot of these structures, and the methodology does not change with the asset class. In late 2017, I spent three months inside the 0x Protocol v2 order-matching engine, line by line, during the ICO cycle. The lesson I carried out of that audit was not about integer overflow specifically. It was that the gap between a project's deck and its code is where the risk lives. The deck describes intent. The diff describes behavior.
The same discipline applies to an equity story. A press release is a deck. A 10-Q is a diff. Most coverage of this company quotes the deck.
Market context sharpens the point. Crypto is in a sideways regime — no trending impulse, no capitulation, positioning rather than conviction. In that regime, attention concentrates into a small number of instruments, and volume rankings become sentiment artifacts rather than flow measurements. That is precisely the environment in which an accretion narrative is most persuasive and least tested.
The metric that breaks at the worst possible price
Start with the arithmetic of accretion, because the arithmetic itself is not disputed by anyone.
Let B be coin holdings and N be common shares. Issue ΔN new shares at market price P. That raises P·ΔN in cash, which acquires ΔB = P·ΔN / p coins at spot price p. Bitcoin per share rises if and only if (B + ΔB) / (N + ΔN) > B / N. Expand the inequality and it collapses to a single condition: P > (B / N) · p. The share price must exceed the intrinsic bitcoin backing per share. That is the definition of mNAV greater than one.
The entire accretion engine therefore has one gate. When mNAV is above one, issuance transfers value from incoming buyers to existing holders in coin terms. When it is below one, the identical mechanics transfer value in the opposite direction. There is no third state, and there is no management discretion that changes the sign.
This matters because of where the discontinuity sits. BTC Yield approaches zero precisely as mNAV approaches one, and turns negative below it. The company's headline KPI fails at the exact price level where its funding engine stalls. A metric that collapses at the moment of maximum stress is not a risk measure. It is a procyclical amplifier, and it is presented quarterly as evidence of strength.
It also measures a transfer, not a creation. The coin count per share rises because an incoming buyer paid a premium for exposure. That premium is the source of the accretion. It is not generated by the software business, not by trading, not by yield strategies. It is donated by the marginal purchaser of newly issued shares. The existing holder's gain and the new holder's entry price are the same number viewed from opposite sides.
I have traced this pattern before in a different form. In May 2022 I was asked to examine Anchor Protocol's sustainability model after the TerraUSD break. The 19.5% headline rate was not funded by trading fees. It was funded by minting. I cross-referenced on-chain flows against the tokenomic whitepaper and the discrepancy was structural rather than incidental — a reward distribution that could not be satisfied by any fee base the protocol could plausibly generate. Fifty pages of transaction logs made the arithmetic visible to anyone willing to read it. Ponzi schemes leave trails in the data.
The difference here is the collateral, and it is not cosmetic. Anchor's liability was denominated in a token whose supply expanded to meet it. Strategy's liability is denominated in dollars, and its asset is bitcoin, which trades in a deep, externally priced, continuously cleared market. I will return to that distinction because it invalidates the laziest version of the bear case. But the metric hygiene problem is identical. The reported number describes the inflow. It says nothing about the durability of the inflow.
What BTC Yield does not count
Here is the omission that gets skipped in every bullish thread.
BTC Yield is computed against common shares. The perpetual preferred series issued to fund purchases — the instruments the company has marketed under several tickers — carries cumulative cash dividends and ranks senior to the common. Every dollar of preferred dividend is a claim on the same coin stack that the common's per-share figure divides.
The consequence is a denominator borrowed against a senior claim. The numerator counts coins. The denominator counts one class of claim on those coins while ignoring a class that ranks ahead of it. A cleaner metric would compute coin exposure per unit of fully diluted, fully ranked capital. Nobody publishes that figure, because it moves in the wrong direction.
I ran into the same failure mode auditing an AI-agent yield protocol in early 2024. The contracts consumed an off-chain data feed with no cryptographic verification of provenance, and the resulting yield numbers were arithmetically valid and economically meaningless. A self-attested metric is an oracle with no attestation layer. Verify the hash, trust no one — including the issuer of the ratio.
The second omission is the cost side of the conversion. Convertible notes issued at low or zero coupons are not cheap capital. They are a sold volatility position. The company received cash today and wrote the bondholder a call on its own equity struck at a premium. If the stock clears the strike, conversion dilutes the common. If it does not, the principal is due in cash or must be refinanced into whatever market exists at that time.
The maturity profile runs into the early 2030s. That is genuinely the structure's strongest feature and I will credit it. But "long-dated" is a statement about when the question is asked, not about whether it has an answer. A refinancing wall is still a wall; the calendar just gives you more time to look at it.
The third omission is servicing capacity. Preferred dividends are cash obligations. The software segment's cash flow is small relative to the dividend burden now attached to the capital structure. Servicing capacity for the senior layer therefore depends on continued capital markets access or continued asset appreciation. Neither is an operating cash flow, and neither is under management control.
What mark-to-market does to earnings quality
Accounting rules now require bitcoin to be measured at fair value with changes running through net income. The consequence is that reported earnings become a derivative of spot price. In a quarter where bitcoin rallies, the company reports a large gain that is unrealized and generates no cash. In a quarter where it falls, the loss is equally unrealized and equally non-cash.
Meanwhile the obligations that do consume cash — preferred dividends, coupon payments, operating expenses — do not mark to market. They are fixed. Earnings quality therefore degrades in a specific, describable way: the income statement becomes increasingly dominated by a non-cash item directionally correlated with sentiment, while the outflows are contractual and insensitive to that sentiment. Analysts who model the equity off reported EPS are modeling a variable that has no relationship to liquidity.
During the FTX review I worked on in late 2022, I traced roughly eight billion dollars of customer funds through wallet clusters to a proprietary trading desk. The mechanism there was commingling, and the internal control environment was, in the plainest reading of the record, absent. The relevant contrast here is not moral. It is informational. At FTX, the ledger was concealed. Here, the ledger is published quarterly and the leverage is disclosed in advance. Silence is the only honest ledger — but loudness is not the same as solvency, and disclosure is not the same as safety.
The quiet funding channel
There is a second-order mechanism that rarely appears in retail coverage.
Index inclusion converts market capitalization into mechanical demand. Passive vehicles must hold shares in proportion to free float. The at-the-market program issues new shares, which increases float, which increases the passive allocation, which absorbs the issuance. That is a feedback loop between the supply of shares and the demand for them, mediated by an index rule book rather than by any investor's judgment.
It is not a conspiracy. It is a mechanical consequence of a rule set. But it means the marginal buyer for newly issued paper is partly price-insensitive during expansion, which flatters the premium, and equally mechanical on the way out when weights reset or a rebalance removes the bid. Audit the edges, not just the center.
Disclosure cadence compounds the problem. The company knows its issuance and holdings continuously. The market sees them quarterly, punctuated by voluntary announcements. Between filings, roughly ninety days pass during which the only verified window into the accretion engine is a press release. A premium that depends on continuous verification is being priced on intermittent data.
The volume number itself is contaminated
A trading-volume rank is not a clean measure of conviction. It aggregates accumulation and distribution without distinguishing between them. It includes dealers hedging options exposure, which means single-name options activity can generate share volume that has no directional content at all. It includes short-term momentum capital that will rotate on a single session's price action.
What a volume rank does establish is concentration. Attention concentrating into one instrument is the same phenomenon as risk concentrating into one instrument, because both are functions of crowding. When the marginal participant base is dominated by price-sensitive, short-horizon capital, realized volatility rises, not because anything changed in the underlying, but because the holder base changed.
The realized record already shows this. In the 2022 drawdown, the equity fell materially harder than the asset it holds. That asymmetry is not an anomaly. It is the structural signature of an instrument that is a levered proxy, and it will repeat in the next drawdown for the same mechanical reason.
One signature, no veto
The capital structure carries a dual-class voting arrangement. The founder holds super-voting shares. The practical effect is that the issuance cadence, the decision to buy or not buy, and the choice of instrument all require one signature and cannot be vetoed by the economic owners of the company.
In late 2023 I ran a post-Merge stability assessment for an institutional client preparing a large allocation. I monitored roughly two thousand validators for three months and found that more than seventy percent were running the same execution client. That is a single point of failure that no amount of capital diversification addresses. Single-asset treasuries and single-person strategies share that property. The concentration is the risk, and it does not appear on any balance sheet line, which is exactly why it gets omitted from models.
The competitive floor is rising
Spot bitcoin ETFs now provide the same underlying exposure at a fee in the low basis points, with no dilution, no key-person dependency, no dividend obligation, and daily creation and redemption. Leveraged single-stock products replicate the high-beta property for anyone who wants it explicitly rather than implicitly.
So what remains exclusive to the equity wrapper? Index membership. Deep options liquidity. Access from retirement accounts that cannot custody spot. An active accretion policy that no fund can legally replicate. That is not nothing. It is also not a moat. It is a bundle of features that competitors can replicate and regulators can reprice. The premium is a rental agreement, and rentals renew at the landlord's discretion.
What the bulls have right
The lazy version of this critique calls the structure a Ponzi. It is not, and the distinction is technical rather than rhetorical.
A Ponzi requires a promised return, funded by new deposits, with no underlying asset. Anchor had a nominal yield. This has no promised yield. A Ponzi requires an accounting fiction. This reports under SEC rules, with audited financial statements and quarterly disclosure of every material financing event. A Ponzi's terminal event is a run, triggered by the discovery of the shortfall. This structure has no redemption promise and no peg to defend, so there is no run to have. What it actually is, is a reflexive leverage vehicle. The distinction matters because the failure modes differ. A Ponzi fails abruptly and criminally. A reflexive vehicle can unwind over years through premium compression — and it can also stabilize if the underlying appreciates, which the Ponzi cannot.
The bulls are also right about the liability design, and this deserves more credit than it receives. The convertible notes are largely unsecured, long-dated, and low-coupon. There is no margin call, no collateral maintenance requirement, and no lender that can force liquidation into a drawdown. In 2022, that was precisely the mechanism that destroyed most leveraged crypto balance sheets. This one did not have it. That was a deliberate structural choice, made early, and it has been validated twice.
They are right about the opportunistic issuance, too. Issuing equity when the premium is wide and abstaining when it is narrow is rational capital allocation by the standard of any corporate treasurer. The criticism is not that the policy is irrational. The criticism is that the reported metric used to evaluate the policy omits the senior claims and breaks at the level where it matters most.
And they are right about the historical function. Before spot ETFs existed, a listed equity was the only compliant route for a very large population of accounts — retirement vehicles, restricted institutional mandates, custody-constrained allocators — to hold bitcoin exposure. The company built the first scaled template for corporate treasury allocation, and that template has since been copied by a long list of issuers across multiple jurisdictions. Whatever happens to this premium, that template is now a permanent feature of the market's architecture.
The honest bear case is narrower than fraud. It is that the central performance metric cannot be independently verified from outside, that the dilution cost is masked by appreciation during up cycles, and that the whole edifice rests on a premium nobody has contractually agreed to pay. That is a weaker claim than fraud and a stronger one than a valuation opinion.
The four numbers to track
Four figures matter from here. The premium, expressed as market capitalization over coin-holdings value, and specifically whether it compresses toward one, because that is where the accretion engine changes sign. The quarterly change in share count, which is the honest measure of transfer. The coverage of preferred dividends by the operating business, which is the only internal source of servicing capacity. And the index weight, because the passive bid is the quietest and most mechanical component of the demand stack.
One falsification criterion is worth stating plainly. If the premium holds above a wide margin through a sustained bitcoin drawdown, then the rental is more durable than the mechanics suggest, and the model has a moat I have not identified. If it compresses through a drawdown while share count keeps expanding, then the engine was never an engine.
The blockchain remembers what humans forget, and the equity ledger does too. It simply remembers quarterly.
If the premium is the product, what exactly is owned when the market stops paying for the product?