A listed company whose main activity is raising capital to buy and hold a crypto asset is one of the defining corporate structures of this cycle, and one of the least understood. The two numbers that matter most, the premium the shares carry over the assets behind them and the rate at which new shares are printed, appear nowhere in the audited accounts. They live in weekly announcements, marketing documents and third-party trackers, and almost every company defines them differently.
What the programme actually scheduled
The Paris programme of 2–3 June 2026 put the structure on the timetable twice inside ninety minutes on the first day, once as a cross-examination and once as an investment case. On Day 1, Tuesday 2 June 2026, the X Ventures Masterclass Stage carried a workshop from 12:00 to 13:00. Our agenda lists it without speakers; its registration listing gives the title as "On Trial: The Bitcoin Treasury Company Model". The framing is the useful part: not what the model is, but whether it stands up. On a programme where speaking slots are not for sale, a session can be scheduled to interrogate a structure much of its own audience is invested in.
Ten minutes after that workshop ended, the Taostats Stage ran a fireside from 13:10 to 13:30, "The Role of Digital Asset Treasuries and the Investment Case Ahead". Nico Pasquariello, billed at the edition as a digital assets specialist at Cantor, and Hyunsu Jung, billed as chief executive of Hyperion DeFi, took the stage; Hadley Stern moderated, billed as independent non-executive chairman of DeFi Development Corporation UK PLC. On Day 2 the Roundtable Lounge held a Digital Asset Treasuries Roundtable from 10:30 to 11:30, with Maja Vujinovic, billed as chief executive of FGNexus, and Jung again.
Every billing here is as it stood in June 2026; that roster is archived, and several titles were out of date within a month. For the full roll-call see the institutions on stage, and for why this post goes to filings rather than quotations, the public record of Paris 2026. What follows reads the mechanism from primary filings and puts both arguments at full strength, date-stamped, because much of it moved between June and September 2026. It is a reading of public documents rather than advice; companies are named only because their disclosures illustrate the structure clearly.
The mechanism, in the order the money moves
Strip the vocabulary away and there are four steps. Raise capital against the listed shell. Buy the asset. Publish a per-share measure of it. Repeat, provided the shares still trade above the assets behind them.
The raising is the part that varies. Strategy's Q2 2026 10-Q shows all three channels running at once in the first half of 2026: 58.5 million common shares issued for $8.24bn net, 75.3 million preferred shares for $7.52bn net across five perpetual series, and $6.71bn of convertible notes outstanding, coupons from zero to 2.25%, nearest maturity September 2028. Twenty One Capital assembled the same engine through a blank-cheque merger that closed in December 2025, funded in part by 1.0% convertible notes due 2030, $486.5m of principal at closing.
Perpetual preferred is the cleverest piece and the heaviest. It never matures, so a falling asset price cannot trigger a refinancing. It does carry a coupon: Strategy's income statement carries $630.2m of dividends on preferred stock for the first half of 2026 alone. A company whose principal asset produces no cash must find those dollars from an operating business, a cash reserve, share sales, or sales of the asset itself. Which of the four it uses, in which week, is the most informative thing in its filings.
Ether-based versions differ structurally here rather than cosmetically. BitMine's release of 31 August 2026 reports 5,067,309 ETH staked, worth $12.7bn, at a 2.67% annualised seven-day yield. The asset pays something towards the cost of holding it. Bitcoin does not.
Premium, discount and the arithmetic of dilution
The shorthand for the premium is mNAV, and it is not one metric. Strategy's free writing prospectus of 24 August 2026 defines it as the market price of a Class A share divided by Net Bitcoin Per Share, warns that it "is not equivalent to 'net asset value' or 'NAV' or any similar metric in the traditional financial context", and states that before 23 July 2026 the term "referred to a different metric" so that earlier figures "are not comparable". Its figure was 1.01x on 21 August 2026. DeFi Development Corp computes the ratio differently, as market capitalisation divided by net asset value, and reported 0.8x on 12 August 2026. None of these appear in audited financial statements. An mNAV figure without its source, date and formula attached is not a number.
What the ratio governs is dilution. Above parity, issuing shares to buy the asset raises the asset held per existing share, and holders get more of it without doing anything. Below parity, the same act does the reverse. DeFi Development Corp's quarterly letter contains both halves: roughly 478,000 shares issued for $1.4m through its at-the-market programme, which the company says cut SOL per share by about 1.4%, set against fully converted SOL per share of 0.066, up about 24% year on year. Growth over the year, erosion in the quarter, from one machine.
Below parity a second gear opens, and it is genuinely accretive: retiring your own paper at a discount. That letter records $7.9m of 2030 convertible principal repurchased for $5.0m year to date, a 38% weighted-average discount to par. Strategy did the same at scale, buying back $1.5bn of its zero-coupon 2029 notes for $1.38bn in the second quarter and booking a $113.9m gain.
What a drawdown does
At 30 June 2026 Strategy held approximately 846,000 bitcoin at a fair value of $49.67bn against a cost basis of $63.94bn, roughly $14.27bn below cost, on a balance sheet carrying $52.56bn of total assets. Since the adoption of ASU 2023-08 on 1 January 2025 those holdings are marked to quoted exchange prices each period, and the swing runs through the income statement on a line called "Unrealized loss (gain) on digital assets", inside operating expenses. The charge was $22.77bn for the half.
Then the coupon comes due in dollars. Strategy's 8-K of 1 June 2026 discloses that between 26 and 31 May it sold 32 bitcoin at an average of $77,135, and that "Proceeds from the bitcoin sales are expected to be used to fund distributions on preferred stock." Thirty-two coins is a rounding error against the 843,706 it still held on 31 May. As a disclosed act by a company built on accumulation, it is not.
Two consecutive weeks in August 2026 show the machine in both gears. In the week of 17–23 August it sold 18,261,118 common shares for $2,006.5m net and bought no bitcoin at all: $136.4m went to repurchasing its own STRC preferred, $300.0m into the USD Reserve, the rest into USD Cash. The following week it sold 4,531,421 shares for $602.8m net and split it four ways: $369.7m into 4,603 bitcoin at an average of $80,318, $50.7m to fund STRC dividends, $151.8m to buy that preferred back, $30.0m into USD Cash.
Note what that does to the simplest assumption about these companies, that the pile only grows. The filed figure was approximately 846,000 bitcoin at 30 June 2026, 840,447 at 23 August, 845,050 at 30 August. It is not a ratchet.
The case for, at full strength
The strongest version of the argument is about financing rather than price. Perpetual preferred has no maturity date and therefore no refinancing cliff, the failure mode that historically destroys leveraged holders of volatile assets. Strategy carries $15.46bn of liquidation preference against only $6.71bn of convertible principal, several tranches at a zero coupon, and reported a USD Reserve of $5.10bn earmarked for preferred dividends and interest plus $1.61bn of USD Cash at 30 August 2026. That is a balance sheet built with the drawdown in mind, and the discount repurchases above show the structure adding per-share value on the way down too.
Where the asset yields, the case improves further. BitMine's August release reports holdings of 5,901,112 ETH, 4.9% of a 120.7 million supply against a stated goal of 5%, bought in every one of 65 consecutive weeks. Even where deals have failed the sponsors report demand: BSTR Holdings, announcing in August 2026 that its route to a public listing had closed, said it continues "to see substantial demand for return on Bitcoin".
The definitional objection is fair, too. A report of 1 September 2026 on a letter from Strategy's executive chairman and chief executive to MSCI says the company argued that "operating" and "non-operating" have no clear definition under US GAAP, IFRS or securities law, that it reports its bitcoin treasury as an operating segment and books the related fair-value swings inside operating expenses, and that such a screen would spare asset-heavy businesses like property trusts while landing on treasury companies. That accounting point is verifiable in the 10-Q.
The case against, at full strength
Start with the arithmetic already on the page. A $22.77bn unrealised loss in a half-year is not a footnote, and it sits in operating expenses. Coupons are paid in a currency the asset does not generate, which is why 32 bitcoin were sold in May 2026 to fund preferred distributions, and why one week in August produced $2bn of share sales and no bitcoin purchases. At an mNAV of 1.01x, the accretion argument is close to nothing.
Below parity it inverts, and companies do not necessarily stop. DeFi Development Corp reported 0.8x in August 2026, disclosed the per-share erosion from its own at-the-market sales, and on 31 August announced its intention to offer a further variable-rate perpetual preferred series. Its letter observes that "Performance among DATs has struggled since August 2025".
Several companies have already stopped. FG Nexus's board resolved on 24 June 2026 to exit the digital-asset business. The Q2 2026 10-Q records that it sold every digital asset before 30 June 2026 and held none at the period end, against $119.4m of ETH at 31 December 2025, and booked a $43.6m net loss from the discontinued operations over the half. It is redeploying towards income-producing affordable housing.
In London the exit was involuntary. Satsuma Technology Plc's announcement of 20 July 2026 records shareholders voting 90.63% to return capital and 90.59% to cancel the listing, against the board's express recommendation: weeks earlier it had urged a vote against while holding 668 bitcoin. The board then moved to prepare to close trading and sell the bitcoin.
Corporate identities have proved provisional. The London-listed company in Hadley Stern's Paris billing gave up the DeFi Development Corporation UK PLC name weeks after the edition, becoming Cykel AI PLC with its ticker moving from DFDV to CYKL at 08:00 on 14 July 2026, and a stated strategic focus on artificial intelligence instead.
Routes to a listing can close as well. On 20 August 2026 BSTR Holdings and Cantor Equity Partners I terminated the agreement that would have taken BSTR public, citing "significant pricing pressure amid challenging market conditions, contributing to dislocations in capital markets". The blank-cheque company's own filing records $15.0m in cash payable to it on termination, the intended withdrawal of the registration statement, and a renewed search for a target. Precision matters: nothing was delisted; a listing route closed. Twenty One Capital, meanwhile, carried its 43,514 bitcoin at $2.55bn against a $3.69bn cost basis at 30 June 2026, with approximately 16,116 of them pledged as collateral.
The index question is live and unresolved
Gatekeeper risk is not hypothetical, and it has already run once. MSCI put Metaplanet and Capital B on a preliminary list of digital asset treasury companies in October 2025 and froze increases to their share counts and inclusion factors, then announced on 6 January 2026 that it would not exclude digital asset treasury companies from its Global Investable Market Indexes, while keeping those restrictions in place. It would consult instead on non-operating companies generally.
That second consultation, published in August 2026, describes the type about as sharply as anyone has managed: firms that "Create value by accumulating and holding non-operating assets", have "Performance driven by market movements, not other revenue-generating activities", and are "Reliant on external capital, not their own operations, to grow". The proposed test admits any company whose operating assets exceed half of total assets, then screens the rest against five ratios covering operating-asset intensity, expense intensity, negative operating cash flow, fair-value intensity and capital dependence, with ineligibility at four of the five. Applied to the ACWI IMI as of May 2026, MSCI's own impact table produces three deletions, Strategy, Yellow Cake and Metaplanet, plus a watchlist of three more. Yellow Cake holds uranium, a useful check on the assumption that the screen is aimed only at crypto.
All of that is a proposal run against historic data, not a decision, and it is deliberately slow: an existing index constituent would have to fail the screens in two consecutive filings before removal. Feedback closes at the end of September 2026, MSCI expects to announce results by 16 October 2026, and any change would be implemented at the November 2026 index review. Exchange-level scrutiny is older and thinner. In September 2025 Nasdaq was reported to be requiring shareholder approval before some listed companies issue equity to fund crypto reserves; in November 2025 Japan Exchange Group was reported to have halted three listings by firms with large crypto holdings while it reconsidered disclosure standards. Neither account carried a rule number.
What is worth checking before forming a view
Five questions get most of the way, and all are answerable from filings. Which instruments fund the treasury, and do any of them mature. Where the next dollar of coupon comes from. Whether shares are being issued above or below the company's own stated net asset value per share, under which formula and on which date. Whether any of the treasury is pledged. And whether the per-share metric quoted sits in an audited statement or only in a marketing document, which for every company here is the latter.
The workshop had the right frame in its title. A structure that has produced both real per-share gains and multi-billion unrealised losses, sometimes inside one company in one year, is a matter for evidence rather than allegiance. The same method reads the tokenisation track. The inaugural Gulf edition follows on 3–4 December 2026 at Louvre Abu Dhabi in the Saadiyat Cultural District, after the date MSCI expects to answer its own consultation. Admission is by application and review.