Four Out of Five: MSCI's Ratio Test and the Slow Unwinding of Strategy's Capital Flywheel

Four out of five. That is the entire rule. Buried in the technical annex of an MSCI consultation document, the threshold is almost insultingly simple: a company that trips four of five financial ratio tests, and trips them across two consecutive annual review periods, loses its seat in the Global Investable Market Indexes. The rule is not proposed. It is approved. It takes effect in November 2026, with the first formal test scheduled for May 2026. The first named casualty is Strategy — the Delaware software company formerly known as MicroStrategy, which converted its balance sheet into a leveraged bitcoin position and, in the process, converted itself into the most-watched equity proxy in crypto. The second is Metaplanet. The third is Yellow Cake. That third name is the one I keep returning to. Yellow Cake holds uranium. It has never touched a satoshi. It was flagged in the same pass, by the same rule, as a digital asset treasury company. Which means the rule is not about bitcoin at all. It is about a corporate form — the listed holding vehicle — and corporate-form rules have a way of outliving the assets that inspired them. MSCI is not a regulator. It is something quieter and in some respects more consequential: a taxonomy. Its Global Investable Market Indexes are the reference set that a large share of the world's passive capital is contractually obligated to track. A fund that benchmarks to an MSCI index does not get to have an opinion about Strategy. It holds the stock because the index contains the stock. When the index stops containing the stock, the fund sells. Not because it wants to. Because its mandate is written in a way that leaves no discretion. That is the mechanism people keep underestimating. Passive flows are not sentiment. They are a settlement instruction. They execute on a rebalance calendar regardless of what the price is doing, regardless of what the fundamentals say, regardless of whether anyone at the fund has ever read a Michael Saylor tweet. The bid they provide is real, and it is uniquely indifferent to valuation. Strategy's own history explains why that indifference mattered so much. The company spent two decades as a mid-cap enterprise software vendor with unremarkable margins and a legacy licensing business. Then, beginning in 2020, it began converting corporate treasury into bitcoin and financing the purchases through the capital markets — at-the-market equity issuance when the stock traded at a premium to its holdings, convertible notes when volatility made the embedded option cheap, and later a layered preferred stack. The software business never disappeared. It simply became a rounding error next to the balance sheet. Index inclusion gave that structure something it could not manufacture on its own: a permanent, price-insensitive marginal buyer. That is what the MSCI rule proposes to remove. Not the bitcoin. The buyer. I have spent enough time inside smart contracts to know that the interesting failures are never in the arithmetic. They are in the places where the specification goes quiet. In 2018 I traced roughly 450 lines of MakerDAO's original Solidity to verify the collateralization logic, and the two liquidation edge cases I flagged were not division errors — they were scenarios the spec had simply not contemplated. The math was correct. The silence was not. MSCI's rule has the same shape. It runs in two rounds. The first is a qualitative screen: does the company hold substantive operating assets? The second is quantitative: five financial ratio tests covering expense ratios, cash flow generation, changes in asset value, and reliance on external financing. Trip four of the five, and the quantitative round is failed. But the first gate is not quantitative. "Substantive operating assets" is a judgment call wearing the costume of a criterion. In contract terms, it is an external call whose implementation you cannot read — a function that returns a boolean and whose source is held by the counterparty. The five ratios are auditable. The first gate is not. And in my experience, the unauditable gate is always the one that decides the case. There is a second gap, and it is larger. The specific parameters and thresholds of the five ratios have not been published. Anyone who tells you they know exactly where Strategy lands on test three is guessing. I am not going to guess. What I can say is that an unpublished threshold combined with a qualitative screen leaves the index provider with substantial interpretive room — and that interpretive room is precisely the ground Strategy's objection stands on. The company's argument is not that it would pass the tests. It is that MSCI is importing policy judgment into index construction, which would compromise the neutrality that makes an index an index rather than a portfolio manager with a house view. That is a serious argument, and it is being made with the correct instruments: references to GAAP and IFRS operating-segment treatment, disclosure of discussions with SEC staff, and an appeal to procedural consistency. The company's position is that its bitcoin activity qualifies as an operating segment under the accounting standards that govern how it reports. MSCI's position is that a treasury vehicle belongs in the same excluded bucket as investment funds and passive instruments. Two rulebooks. One asset. No reconciliation clause. That is the actual dispute. Here is where I think the market has made a straightforward error, and it is a timeline error rather than a directional one. The effective date is November 2026. The first test is May 2026. The buffer requires failure across two consecutive annual review periods. Run that forward. A company flagged in May 2026 must be flagged again in the following cycle before removal is triggered. The earliest realistic removal is therefore 2027, plausibly 2028. November 2026 is not D-day. November 2026 is when the clock starts. This matters because the consensus trade is built around a single date. Rebalancing does not work that way. Index deletions are executed across scheduled reconstitution windows, spread over quarters, and the forced selling is metered out in tranches rather than dumped in one print. The market is pricing a cliff. The mechanism describes a slope. Now the part that JPMorgan's number does not capture. The widely cited estimate is roughly $2.8 billion of passive outflow if Strategy is removed. I have no independent verification of that figure and I would treat any single-number estimate of index-driven flow with caution — the underlying float attribution requires aggregating 13F filings across dozens of managers, and the number moves with every quarterly filing cycle. But grant the figure for the sake of argument. What does $2.8 billion actually do? Against a name with Strategy's turnover, it is days of volume. Adrian Morris's argument that the liquidity can absorb it is, on magnitude, probably correct. I would not bet against him on the size of the wave. But magnitude is the wrong axis. The right axis is composition. A passive fund is the least price-sensitive node in Strategy's entire capital loop. It buys the stock without asking what the premium to net asset value is. It buys at mNAV 1.2, it buys at mNAV 2.0, it buys at mNAV 0.9. Its indifference is the point. Remove that node and you have not removed $2.8 billion of demand — you have removed the only demand that never asked a question about price. Which brings us to the number that actually matters: mNAV, the multiple of Strategy's market capitalization to the net asset value of its bitcoin holdings. Everything in the capital structure is a function of it. Above 1.0, issuing equity is accretive to bitcoin-per-share, and the at-the-money program is a machine that converts premium into coins. Below 1.0, the same program is dilutive, and a rational management team stops using it. Convertible issuance depends on equity volatility rich enough that the embedded option subsidizes the coupon. Preferred dividends depend on either operating cash flow — small relative to the stack — or fresh issuance. The whole structure is a loop, and the loop is gated on a single multiple. So the correct question is not how many dollars leave. It is how much premium compression those dollars cause. And on that question, the honest answer is that nobody outside the company's cap table has the data. I do not have a verified float breakdown in front of me. Neither, I suspect, do most of the people quoting the $2.8 billion figure. Forensics is just history written in hexadecimal — and a 10-K is history written in GAAP. Both are admissible. Neither is complete. What I can say with confidence is structural. A premium sustained partly by a price-insensitive buyer is a premium with a floor underneath it. Remove the floor and you discover what the price-sensitive bid actually is. I have watched this exact experiment run before, not in equities but in DeFi. Curve pools in 2022 looked bottomless until the incentive programs ended, and then the depth that remained was the depth that had always been real. Liquidity that only exists because someone is contractually obliged to provide it is not liquidity. It is a subsidy with a schedule. The Yellow Cake detail is the one I would put at the top of any risk memo. Uranium, not bitcoin. Physical commodity custody, offtake agreements, logistics. Flagged by the same rule in the same pass. That confirms the test is form-based rather than asset-based, and form-based tests generalise. Every listed holding wrapper — commodity trusts, royalty vehicles, any future structure that parks a hard asset on a public balance sheet and calls it a strategy — is now inside the blast radius of a definition that has not yet been finalised. And a definition that is not finalised is a definition that can move. The rulebook is where the risk lives; the headline is where the risk is sold. Now the contrarian read, because the consensus chain here has four arrows and only two of them are solid. The chain runs: removal causes outflow, outflow causes price pressure, price pressure causes premium compression, compression stalls the flywheel. Arrow one is mechanical and reliable. Arrow two is contested and probably smaller than feared. Arrow three is close to tautological — if the stock falls and the bitcoin does not, the multiple compresses, full stop. Arrow four is where the reasoning breaks. The flywheel does not require mNAV above 1.0 to survive. It requires mNAV above 1.0 to grow at its current rate. There is a meaningful difference between a machine that stops accelerating and a machine that reverses, and the market is currently pricing the second while describing the first. A company that stops issuing equity and simply holds its bitcoin is a different company, not a dead one. Slower. Less interesting. Still solvent unless the debt stack says otherwise, and the debt stack is the part nobody is stress-testing. What I think is genuinely mispriced is not the outflow. It is the precedent. Index providers converge — not because they coordinate, but because they read each other's rulebooks and because their clients ask why one provider's eligibility criteria differ from another's. If even two of the remaining major providers publish a comparable standard within the next few quarters, $2.8 billion becomes a rounding error and the entire "digital asset treasury company" category gets repriced as a structurally discounted wrapper. That is the scenario worth insuring against, and it does not require anyone to be wrong about the size of the first wave. The second mispricing is temporal. The market is treating November 2026 as the event. It is not. May 2026 is the event — the first formal test, the first published flag list, the first moment the qualitative gate returns an actual answer. Everything after that is arithmetic and patience. Three signals to watch, in order of information density. First, the May 2026 test result and whatever MSCI publishes alongside it. Not the outcome — the language. If the qualitative screen's criteria get defined more tightly, the rule is softening. If they stay vague, the interpretive room was the point. Second, mNAV relative to 1.0. Not the stock price, not the bitcoin price. The ratio between them. A sustained break below parity is the only unambiguous signal that the loop has changed direction, and it will show up there before it shows up in any headline. Third, whether S&P, FTSE, or Nasdaq publishes anything structurally similar within two quarters of the MSCI effective date. Silence from all three is the bullish case. One of three is a warning. Two of three is a regime. The ledger never lies, it only waits to be read. And the entry worth reading here is not the outflow line. It is the premium — the gap between what the market says the coins are worth and what the market pays for someone else to hold them. That gap has a rulebook written under it now. The only open question is how many annual review periods it takes before the rulebook finishes the sentence.