The Denominator Decides

Companies that hold cryptocurrency on the balance sheet, and how they are valued.

We rebuilt five listed crypto-treasury companies from their own SEC filings — 34 source documents and 5 EDGAR submission indexes — and measured each with identical arithmetic at one instant. Coin counts and share counts are filed figures. Every price is market data, labelled as such and struck at 2026-09-08 16:00 ET. No vendor, no aggregator, no consensus estimate anywhere.

The finding

One company, one afternoon, one set of filed inputs — and three defensible answers.

The first charges the common equity against the whole coin stack, including the $21.318bn of it that belongs to preferred and note holders. The third deducts what the common does not own. Neither corrects the other — the gap between them is exactly the size of the capital structure, which is 32.16% of the coin stack at this company and 0.25% at another in the same table.

The industry quotes whichever of the three supports its argument. A reader shown one of them has not been shown a valuation. They have been shown a choice of denominator, and it was not named.

The half the rally hides

Between the last quarterly filing's cover date of 2026-07-24 and the close of 2026-09-08, the company's coin count rose 0.151% and its share count rose 9.33% — 35,848,448 shares issued for $3,886.7M net. So:

A holder who bought the shares for exposure to the coins owns 8.40% less bitcoin per share, while the share price rose 48.93% over the same days. Over fourteen months the fall is 9.28%, and almost all of it happened in the last six and a half weeks of that period. No market price enters that arithmetic, which is why it survives a rally or a fall — and why it is the figure nobody quotes.

And an obligation with a date on it

$5,913,659k of convertible note principal — 88.1% of the $6,713,659k outstanding — can be put back to the company at par by 2028-09-15. The maturity column of the same table reads 2028 to 2032 and is the wrong column to read. 73.0% falls inside calendar 2028 alone.

Beside it, the annual cash call, rebuilt from five preferred series and six note tranches rather than quoted: $1,762,548,028 at 2026-06-30, against the company's own "approximately $1.76 billion" — agreement to 0.145%. It has since fallen to $1,660,054,984 as preferred is retired, covered 36.9 months over by a $5.10bn reserve against a board floor of twelve.

What we could not explain, and did not pretend to

The obvious story is that leverage causes the discount. The cross-section refuses it. One company in the same table carries $5.572M of total liabilities against $2.207bn of ether — no debt, no preferred, no dividend — and trades at 0.868. The levered one, with $21.318bn ranking ahead of it, trades at 0.865. They are indistinguishable.

So this report states the measurement finding and names no cause. Where the capital structure does show up is the enterprise measure, on which the levered company is the more expensive of the two.

Why this is not a report about a premium

The research was opened on a claim about where one company's ratio sat. That number moved 32.9% in six and a half weeks on price action alone, from 0.651 to 0.865, with nothing about the business changing. Finnacl publishes under the Advisers Act publisher's exclusion, whose third limb requires a genuine, regular publishing cadence. A measure that decays in days cannot carry a report published on one. The perishable version was dropped before a word of it was drafted, and the durable findings were rescoped into this one.

Three corrections we printed against ourselves

1. A fund weight we published was wrong. An earlier screen recorded one crypto-equity fund as 18.24% treasury companies. It is 20.32% — the earlier pass read the published top ten only, and a third treasury company sits at position 15.
2. A fifth preferred series exists, euro-denominated and listed in Luxembourg, so it never appears on the US cover page. Treating €775,000k as dollars understates the obligation stack by $109,120k.
3. A percentage whose basis was wrong even though its answer was right. The put-wall share paired a principal numerator with a carrying-value denominator. Recomputed on one basis it is still 88.1% — but its companion figure moved from 73.1% to 73.0%, and that is what we print.

The three numbers

What is in the full report

The eight quarterly markers and their readings. The full three-denominator arithmetic, including what happens when the preferred is marked at market rather than at par. The complete capital structure — five preferred series, six note tranches, every put date — and the annual obligation rebuilt line by line. The cross-section of five vehicles measured at one instant, with the one that refuses a leverage story. What 20.32% of one $463,132,759 fund is invested in, and what that fund's own page does and does not disclose. A page on each company. Every source that was gated and not circumvented. 448 verification checks, none failing; 165 defects planted across three mutation rounds, and 73 of 73 caught in the final one.

Disclosures. Finnacl Research publishes impersonal investment research, identical for every purchaser and tailored to no one's circumstances. It is not individualised investment advice, and Finnacl Research is not an investment adviser (publisher's exclusion, Investment Advisers Act §202(a)(11)(D)). No statement here says what any reader should do. No recommendation, no price target, no forecast. Every share price and coin price is market data struck at 2026-09-08 16:00 ET and was stale within a day. The author holds no position in any security or fund named. © Finnacl Research 2026.