Four of Twenty: The Treasury-Company Flywheel Reversed Before Anyone Published a Warning

On 24 September, DWF Ventures published a count. Of the top twenty publicly traded digital asset treasury companies, four trade above the value of the assets they hold. Sixteen trade below.

Twenty issuers. One strategy. A 20% hit rate.

I have spent enough of my career inside capital structures to recognize what that ratio is. It is not a valuation opinion. It is a mechanism failure. When a primitive breaks in sixteen of twenty implementations while the underlying asset sits unchanged, the defect is not in the asset. The defect is in the primitive — and it was present from the first deployment.

The treasury-company model is a state machine with two branches. Sixteen of twenty issuers are now executing the branch that destroys shareholder value on every iteration. Their operators did not choose it. The arithmetic chose it for them.

Context: what a DAT actually is

The template came from Strategy, formerly MicroStrategy, in 2020. A listed operating company designates a crypto asset — almost always bitcoin, occasionally ether — as its primary balance-sheet reserve. The equity is the wrapper. The coin is the collateral. The mechanism connecting the two is market-to-NAV.

mNAV is market capitalization divided by the marked value of the reserves. Above 1.0, the equity is a premium instrument. Below 1.0, it is a discount instrument. That single ratio governs every decision the company can make.

Above 1.0, issuing shares buys coin at a discount to what the market pays for the wrapper. Coin-per-share rises. This is accretion, and for these vehicles it is the entire product. Below 1.0, issuing shares hands existing holders' coin value to new buyers. Coin-per-share falls. This is dilution.

Accretion and dilution are not narratives. They are arithmetic, and the arithmetic is public. We do not build for today, and this structure was built entirely for a premium that was never contractually guaranteed.

The first-generation pitch — buy the stock, receive leveraged coin exposure, avoid the custody problem — was coherent exactly as long as mNAV stayed above 1.0. Sixteen issuers now fail that condition.

The arithmetic of a compressing premium

Let V be reserve NAV, S the share count, M the mNAV. Issue new equity equal to n·V, with n expressed as a fraction of NAV. The change in NAV per share is:

(1 + n) / (1 + n/M) − 1

At n = 0.10:

| mNAV | NAV per share change | |------|----------------------| | 2.00 | +4.76% | | 1.50 | +3.13% | | 1.25 | +1.85% | | 1.10 | +0.83% | | 1.00 | 0.00% | | 0.90 | −1.00% | | 0.80 | −2.22% |

Read the top half. A premium compression from 2.0x to 1.1x does not reduce per-issuance accretion by 45%. It reduces it by 83%. The flywheel does not slow proportionally to the premium; it collapses non-linearly toward the point where the sign flips. A company that raised capital comfortably at 2.0x is, at 1.1x, running a machine that returns 0.83% per turn — before fees, before the underwriter, before the board's own dilution.

Read the bottom half. Below 1.0, the only rational action is to stop. Which is precisely the problem.

while mnav > 1:
    shares  += issue_atm(price)
    reserve += buy(coin, proceeds)
    mnav = market_cap() / nav(reserve)   # assumes premium holds
# mnav < 1: loop body is value-destructive
# rational action: break

Breaking the loop stops accretion. Stopping accretion removes the growth profile that justified the premium. Removing the growth profile compresses the premium further. This is a reflexivity spiral with a sign flip in the middle, and sixteen of twenty issuers are already past the flip.

I audited Parity's multi-sig library in 2018 — three weeks on version 2.1 — and the finding that mattered was never the external call itself. It was that the call returned before the ownership state variable was finalized. The reentrancy lesson from 2018 does not change with asset class: the recursive call always arrives before the state variable is written. Here the recursive caller is the market. It repriced the wrapper before the issuer could update its capital structure, and it will reprice again on the next disclosure.

Note what the largest holder did next. According to the same reporting, Strategy stated it can continue buying with existing cash. Read that operationally. An issuer switching from financing-driven accumulation to cash-flow-driven accumulation is telling you it no longer trusts the ATM window. That is a defensive posture, and the largest player in the category adopted it first.

The layer nobody audits: custody and the convert stack

In 2021 I published a report on ERC-721 metadata after watching 60% of popular collections degrade when gateway caching policies changed. The title was "The Illusion of Ownership." Token holders believed they owned art. They owned a URI, a gateway, and a promise.

DAT shareholders hold a claim on a corporate balance sheet: a custodian, a board, an auditor, a share count, and a debt stack sitting between them and the reserves. The metadata is off-chain. So is the coin. The art is the hash; the value is the proof — and in this structure the proof is a quarterly filing, not a block.

DWF's dataset does not disclose the convertible debt terms of any of the twenty issuers. That is the most important missing field in the entire report. Convertible notes are the standard funding instrument for this model, and a conversion price above the prevailing share price is not a feature. It is a maturity wall. In a sustained discount regime, an issuer has three exits: refinance on dilutive terms, sell reserve assets, or default. The first destroys the pitch, the second hits spot markets, the third is terminal.

The arbitrage between NAV and price looks free on a screen. It is not. T+1 equity settlement, lockups, restricted stock, borrow cost and position limits all sit between the mispricing and the trade. mNAV is an oracle feed with a multi-day lag and a non-trivial execution cost. I have watched the same gap in lending markets: the feed is correct, the liquidation is late, and the loss lives in the interval. Sixteen trading below NAV is not sixteen opportunities. It is one dislocation with sixteen different liquidity profiles.

The blind spot in the consensus read

The comfortable interpretation is that this is a healthy clearing — weak imitators die, strong incumbents survive. I do not buy it.

The surviving four's premium is partly a function of the other sixteen's failure, not of their own structural quality. When a category collapses, flow concentrates into whatever still has liquidity and index eligibility. That is not a moat; it is a single point of failure wearing a moat's clothing. If one of the four loses its premium — a convert maturity, an accounting reclassification, one bad quarter — there is no breadth in the sector to absorb it.

Second blind spot: the Investment Company Act of 1940. The purer the balance sheet becomes, the closer the entity is to a pure holdco for a reserve asset, and the closer it sits to the statutory definition of an investment company. The selling point is purity. Purity is the regulatory trigger. Nobody in the marketing deck mentions this.

Third: fair-value accounting. Mark-to-market reserves produce reported earnings that look like operating performance and behave like beta. The income statement becomes a price feed with a legal cover page. The compliance wrapper these vehicles sell is also thinner than advertised; I have watched KYC regimes where the entire cost lands on honest users while a handful of wallet movements routes around the check. An ETF does not ask you to trust a CEO's timing.

Takeaway

Stop watching the mNAV screen. Start watching the convertible maturity ladder. The premium is a quote, and quotes get withdrawn without notice. The debt is a contract, and contracts have dates printed on them. The first issuer to disclose a conversion problem in a discount regime will not be announcing it — it will be filing it, quietly, on a Friday.

The sector has not found a floor. It has only stopped reporting where the floor is.