The 696 Problem: Metaplanet's Warrant Ratchet and the Limits of a 41% Rollback

The 696 Problem: Metaplanet's Warrant Ratchet and the Limits of a 41% Rollback

Hook

There is a number inside Metaplanet's latest disclosure that deserves more scrutiny than the headline it generated. It is 696.

That was the number of shares a single Series 10 stock acquisition right delivered on exercise before the board acted. After the board acted, it became 410. The 41% reduction in the company's potential share count — from roughly 319.5 million to 188.2 million — is that same fact restated in aggregate. Nothing else moved. No bitcoin was sold. No capital was returned to anyone. A conversion ratio on an insider compensation instrument was rewritten, and the rewrite is being presented as a concession to shareholders who spent weeks demanding one.

I have spent enough time inside settlement logic to recognize the shape of this. A variable that is supposed to be fixed gets an adjustment clause bolted to it. The clause fires rarely at first, then routinely, then continuously. Nobody notices until the aggregate becomes large enough to lead a filing. The ledger remembers what the code forgot. Here the ledger is a Japanese securities disclosure, and what it remembered is that a compensation pool swelled from roughly 46 million shares to about 319.5 million while the company was telling the market that every capital raise was disciplined.

Context: What Metaplanet Actually Is

Metaplanet is not a bitcoin company that happens to be listed. It is a listed company that converted itself into a bitcoin balance sheet, and the conversion is the entire thesis. Formerly Red Planet Japan, a hotel operator with a distressed legacy business, it repositioned in 2024 around a single objective: accumulate bitcoin, issue equity to fund the accumulation, and let the resulting per-share coin count perform the function that an operating business would normally perform.

The model is familiar because Strategy industrialized it. The mechanics fit in one sentence. If a company's equity trades at a premium to the market value of the coins it holds, then issuing new shares and buying coins with the proceeds raises the coin count per share. If the equity trades at a discount, the same operation lowers it. Premium is accretive. Discount is dilutive. There is no third state, and no amount of operating performance can substitute for the missing premium.

That binary is why the model is unusually sensitive to instruments that look like accounting footnotes. An at-the-market offering program, a convertible note, a moving-strike warrant, an employee stock acquisition right — each is a claim on the same balance sheet, and each competes for the same per-share metric the equity story depends on. The metric is not revenue. It is bitcoin per fully diluted share. Anything that changes the denominator changes the number investors are actually buying.

Japanese market structure amplifies this. Metaplanet's shareholder register is retail-heavy in a way that most Western treasury vehicles are not, and the Japanese retail base has spent two decades learning to read dilution instruments the hard way. The moving-strike warrant — the "MS warrant," in local parlance — is the clearest example. Unlike a conventional warrant, the exercise price ratchets downward as the share price falls, which means the holder is structurally guaranteed a spread while existing shareholders absorb the supply. Metaplanet used that instrument in 2024, and the criticism it drew still shapes how its retail holders read every subsequent disclosure.

That history is the context for the Series 10 controversy. The shareholders who spent weeks in open revolt were not responding to a single line item. They were responding to a pattern they had already seen once, in a different instrument, under a different name, with identical arithmetic.

The scale matters too. Metaplanet's accumulation program required an extraordinary volume of equity issuance — international offerings, at-the-market programs, warrant exercises — each one feeding the balance sheet and each one feeding the conversion formula. By the time the board acted, the company had moved from a hotel operator to one of the largest corporate bitcoin holders in Asia. It had also, without ever disclosing it as a risk factor in plain terms, converted a compensation plan into a call option on its own fundraising cadence.

Series 10 stock acquisition rights sit squarely in that lineage. In Japanese corporate practice, a stock acquisition right — shinkabuyakuken — is a standard instrument with four definable parameters: the number of rights outstanding, the exercise price, the number of deliverable shares per right, and the exercise window. Three of the four are usually stable. The third is not, if the instrument carries an anti-dilution adjustment clause. That is where the Metaplanet story actually lives.

The instrument is stark from inception. The exercise price is 10 yen. At any share price the company has traded at over the past two years, a 10-yen strike is not a meaningful cost to the holder. It is a nominal formality layered on top of what is functionally a free grant. The consequence is that the "value" of a Series 10 right is almost entirely intrinsic value of the underlying shares, which scales linearly with the conversion ratio. Raise the ratio, raise the value. Lower the ratio, lower the value. There is no optionality premium to speak of, no meaningful time value, no stochastic component. It is a leveraged claim on the share price dressed in the language of compensation.

And the ratio had been raised, repeatedly, by exactly the capital-raising activity shareholders were told was in their interest. That is the crux.

Core: The Arithmetic Is the Argument

Start with the arithmetic, because the arithmetic is the argument.

The company disclosed that the Series 10 pool will fall from approximately 319.5 million potential shares to approximately 188.2 million. Working backward from the two conversion ratios gives the number of rights outstanding: 319.5 million divided by 696 is roughly 459,000. 188.2 million divided by 410 is also roughly 459,000. The right count did not change. Only the multiplier did.

That reconciliation tells you what the board actually did. This was not a cancellation of grants. It was a reset of a formula. The instrument stayed in place. The holders stayed in place. The conversion ratio was rolled back to 410, a number the company itself identifies as the level immediately before its international share offering in September 2025.

The company's own language around that date is the most revealing sentence in the entire disclosure. Metaplanet characterized the September 2025 offering as the point where capital raises stopped being strongly accretive. Read that carefully. It is a dated admission that the accretion engine — the mechanism the entire corporate strategy rests on — crossed a boundary at a specific moment, and that the warrant ratchet had been running on the wrong side of that boundary ever since.

Now reverse-engineer the growth of the pool. The aggregate went from roughly 46 million shares to about 319.5 million. That is a multiple of 6.95. The aggregate is nothing more than the right count multiplied by the conversion ratio, and the right count was stable. Therefore the ratio moved by the same factor: from approximately 100 shares per right to 696, a multiple of 6.96. The two numbers move together because they are the same number.

A sevenfold adjustment is not a technicality. It is the signature of a ratchet that responds to every issuance and compounds on each one. Weighted-average anti-dilution clauses typically produce adjustments in the low single digits across a full cycle. Repeat expansions of this magnitude are characteristic of full-ratchet behavior, or of a weighted-average formula applied frequently enough that compounding does the work of a full ratchet.

Either way, the effect on the cap table is identical and easy to state. Each equity raise funded bitcoin purchases. Those purchases grew the balance sheet. The balance sheet supported the story. The story supported the share price. The share price enabled the next raise. And every raise fed the conversion ratio. The insiders' claim on the company expanded in direct proportion to the company's fundraising success. That is not a side effect of the compensation design. It is the compensation design.

Consider what the ratchet did in dollar terms at a representative share price. A 10-yen strike against a share price in the low hundreds of yen means the holder's intrinsic spread per share is nearly the full market price. At 696 shares per right, a single Series 10 right carried intrinsic value equivalent to 696 shares. At 410, it carries the value of 410. Multiply the difference by the roughly 459,000 rights outstanding and you arrive at the $220 million the company says it erased. The figure is not rhetorical. It is the difference between two multipliers applied to a fixed count of instruments.

There is a second structural defect, and the company has effectively conceded it. The pool was sized as a percentage of fully diluted capital rather than as a fixed grant. That single design choice converts a compensation plan into a pro-cyclical index. A fixed grant of 46 million shares remains 46 million shares no matter how many times the company issues. A percentage-of-fully-diluted pool grows every time the denominator grows. Raise, dilute, grow the pool, raise again. The loop has no internal brake.

I stress-tested liquidity pools in 2020 against precisely this failure mode, and the result generalizes. The invariant held comfortably in the middle of the parameter range and failed at the tail, because the failure condition was endogenous — the act of using the mechanism changed the parameters the mechanism depended on. A percentage-of-fully-diluted compensation pool has the same property. The incentive to raise capital and the incentive to enlarge the insider claim are not merely correlated. They are the same instruction expressed twice.

Now the shareholder payoff, which should be stated as plainly as the problem. The company says bitcoin per fully diluted share rises about 8.8% without buying a single coin. That is real, and it is the strongest argument the board has. Removing 131.3 million potential shares from the denominator raises the per-share claim on a static coin stack.

Working backward from the 8.8% figure implies something about the base. If coin holdings are unchanged and the fully diluted count falls by 131.3 million, then the ratio of the old count to the new count is 1.088. Solving that gives a pre-announcement fully diluted base of roughly 1.6 billion shares. That number is the correct context for the concession: a 41% cut to one instrument produced an 8.8% improvement to the metric, because the instrument was only one of several claims on the same base.

That gap — 41% against 8.8% — is the honest measure of what this rollback returns. It is meaningful. It is not transformative. And it does not touch any of the other instruments in the stack, which continue to compete for the same denominator.

It is worth being precise about why fully diluted is the right denominator here. A treasury vehicle's equity is a claim on a pooled asset. What a holder owns is not a share certificate; it is a proportional claim on the coin stack. Every instrument that can convert into a share reduces that proportion. Options, warrants, restricted stock, convertible notes — the legal form does not matter. Only the conversion schedule matters. A warrant count that grows from 46 million to 319.5 million shares is not a compensation disclosure. It is a change in the ownership structure of the balance sheet.

The exercise schedule compounds the point. The remaining warrants become exercisable in thirds in 2029, 2030 and 2031, with shares received on exercise locked up until August 2031. That is not a removal of the overhang. It is a deferral with a calendar date attached. The supply is scheduled, not eliminated, and it lands after the current cycle has resolved in one direction or the other.

The company sold the reset as a response to shareholder pressure, and it was. But the sequence is informative. The stock dropped roughly 17% over two sessions the week before the announcement, following an initial response from CEO Simon Gerovich that failed to satisfy investors. The 41% cut is the second attempt. Proposal, backlash, inadequate response, larger concession — that ordering is the standard progression of a governance dispute resolved by negotiation rather than by design. The number moved because the holders pushed. It did not move because the structure was rethought.

The deeper issue is that the accretion math and the compensation math were pulling in the same direction for eighteen months. As long as the equity traded at a premium to net asset value, issuance raised bitcoin per share and enlarged the insider pool simultaneously. Both parties were winning, and the arrangement looked aligned. The moment the premium compressed, the two interests diverged: further issuance would dilute per-share coin count while the pool continued to grow on the percentage formula. The September 2025 boundary is the date that divergence became undeniable — and it took a shareholder revolt to get the company to name it in writing.

Contrarian: What the Disclosure Does Not Say

Here is what the disclosure does not say, which is where audit attention belongs.

First, the adjustment clause. Metaplanet reset the conversion ratio to 410. It did not announce the removal of the mechanism that produced 696. Those are different acts with different consequences. If the clause remains live, then the next discounted issuance — for any reason, including a bitcoin purchase — re-inflates the ratio from the new base. Freezing a ratchet at a lower level is not the same as disarming it. The disclosure does not state whether the clause was amended, suspended, or left untouched. Silence in the logs speaks loudest, and this log has a hole in it.

Second, the reconciliation of the canceled employee tranche. The company says a plan to move 20% of the warrants into a new employee incentive pool was scrapped, with those rights canceled as part of the 41%. But 41% is exactly, arithmetically, what the 696-to-410 reset produces on its own. The reset alone takes the pool from 319.5 million to 188.2 million — a 41.1% reduction with nothing else applied. Either the employee tranche is embedded in that calculation in a way the disclosure does not illustrate, or the framing attributes a larger concession to a smaller set of actions. A reader cannot determine which. That is a disclosure gap rather than necessarily a misstatement, but a gap of that kind is precisely what a reconciled breakdown exists to eliminate.

Third, the related-party channel. The announcement does not address the CEO's economic interest in MMXX Ventures. This is not an accusation; it is an observation about scope. Governance failures rarely sit inside the disclosed perimeter. They sit at its edge, in the entities and arrangements the disclosure schedule does not reach. I spent 2021 mapping ERC-721 royalty enforcement and found the same structural blind spot from the opposite direction: protocol-level royalty logic was absent on roughly 30% of marketplaces, so enforcement migrated entirely off-chain, into contractual arrangements no block explorer could verify. The disclosure perimeter behaves identically. What is inside is verifiable. What is outside is asserted.

Now the personal accounting, which deserves precision rather than gesture.

Gerovich recused himself as a Series 10 holder. He retains the 64 million shares he received through an August 28 exercise under the old terms, plus the right to acquire another 49.1 million. Against a post-reset pool of 188.2 million potential shares, a retained right to 49.1 million is roughly a quarter of what remains. That is a material concentration, disclosed without being emphasized.

VanEck's head of digital assets research, Matthew Sigel, calculated that the CEO gives up approximately 79 million shares worth about $123 million, and described the package as a meaningful realignment of management and shareholder interests. That is a fair description of the direction of travel. It is also a description that stops short of claiming the structure was fixed.

The two disclosed dollar figures do not quite agree, and the disagreement is instructive. The company's $220 million of erased warrant value across 131.3 million shares implies roughly $1.68 per share. Sigel's $123 million across 79 million shares implies roughly $1.56. Both are defensible against their own reference dates. The difference is a reminder that warrant value is a function of the date you mark it, and that any headline figure for value given up is a snapshot, not a settlement. Trust is verified, never assumed — including when the verifying party is a fund whose business depends on the asset class it is describing.

And there is the concession that was not made. Some investors wanted more than they got. The disclosure does not quantify the ceiling of their demand, but the shape of the outcome is legible: the pool was cut, the CEO's retained position was not restructured, the related-party arrangement was not addressed, and the replacement program was deferred to an outside consultant. Each of those is a real item. Collectively they define the boundary of what shareholder pressure achieved here — a boundary drawn at the point where the insiders' own position began.

Takeaway

The forward-looking question is not whether this rollback is real. It is. The per-share coin metric improves, and that improvement is measurable without deploying a single yen of new capital.

The forward-looking question is what replaces the structure, and whether the replacement reproduces the defect. Metaplanet says it will design a new compensation program with an outside consultant. The single most important design decision in that program is already visible in the failure it is replacing: a fixed grant in absolute shares does not scale with dilution, and a percentage-of-fully-diluted pool does. If the consultant delivers a percentage pool with a ratchet, the 696 problem returns under a new series number, and the next rollback gets negotiated at a worse price.

There is a structural condition underneath all of this that no compensation reform addresses. Metaplanet's equity is down more than 43% this year against a roughly 15% decline in bitcoin and a 20% decline at Strategy. That is a downside beta above two to the underlying asset. The premium that makes the model accretive is the same premium that makes the drawdown violent. Liquidity is a mirror, not a moat — it reflects the premium on the way up and the absence of it on the way down.

When the multiple sits below one, the engine inverts. Every raise makes the per-share coin count worse, and the rational move is to stop raising. The company's own September 2025 admission says it has already stood at the edge of that condition once — and the warrant ratchet was still expanding while it stood there.

Stability is engineered, not emergent. A cap table is a system, and systems fail at their boundaries rather than at their centers. The next securities report will contain one line that determines whether this rollback was a correction or a deferral: whether the adjustment clause still exists. Everything else in this story is commentary on that.