Digital asset treasury companies: how they work and how they break

What a digital asset treasury company actually is

A digital asset treasury company — a DAT, or a "DATCO" in index-provider language — is a company with an ordinary stock market listing whose principal purpose is to hold a cryptoasset on its balance sheet, funded by issuing securities to investors. Three parts, all of them load-bearing:

• A listed vehicle. Ordinary shares, an ordinary exchange, ordinary settlement. Usually not a fresh listing: the standard route into this sector is a reverse merger into an existing small-cap whose original business had stalled.
• A cryptoasset as the principal asset. Not a side allocation of surplus cash. The coin is the balance sheet. At Strategy Inc, total revenues were $122.4m in the quarter to 30 June 2026 against an $8.32bn mark-to-market swing on bitcoin in the same quarter (Form 10-Q, filed 3 August 2026). Any description of it as a software company with a bitcoin sideline has the proportions wrong by about three orders of magnitude.
• Securities issuance as the funding engine. The money does not come from operations. It comes from selling shares, convertible notes and preferred stock, and the proceeds are spent on the asset.

That third part is what separates a DAT from an operating company that happens to hold some bitcoin, and from a fund or exchange-traded product. A fund holds assets for unitholders inside a structure built for the purpose. A DAT is a corporation: it has creditors, preferred shareholders, tax, a board, a listing to defend, and its own securities to service. If the underlying vocabulary is new, start with what digital assets actually are.

The shell pattern

You can verify the reverse-merger pattern yourself in SEC EDGAR, which keeps every registrant's former names. A few verified chains, as at 26 August 2026: Riot Platforms was Riot Blockchain (2022), before that Bioptix (2017), Venaxis (2016) and AspenBio Pharma (2012) — a veterinary diagnostics company that became a bitcoin miner. American Bitcoin Corp was Gryphon Digital Mining (2025), before that Akerna Corp, a cannabis compliance software business (2024), before that a blank-cheque SPAC. Forum Markets was ETHZilla (2026), before that 180 Life Sciences, a biotech (2025), before that KBL Merger Corp IV, a SPAC (2020). Empery Digital was Volcon, an electric powersports manufacturer (2025). Tron Inc was SRM Entertainment, a toy maker (2025). Nakamoto Inc was KindlyMD (2025). BitMine Immersion was Sandy Springs Holdings (2021).

The naming trap. At least eight companies in this cohort renamed themselves between mid-2025 and mid-2026, and in several cases the ticker did not move with the name: MicroStrategy became Strategy but remains MSTR; Asset Entities became Strive but remains ASST; SharpLink Gaming became Sharplink but remains SBET; Helius Medical Technologies became "Solana Co" and still trades as HSDT, a ticker that reflects a medical-device past. Others changed both: ETHZilla (ETHZ) became Forum Markets (FRMM) on 2 March 2026. One vanished: Semler Scientific's SMLR ticker was withdrawn on 16 January 2026 when Strive acquired it — Semler is an acquiree, not a continuing peer, which is the single most commonly garbled fact in coverage of this sector. Any list of these companies more than about three months old will contain names that no longer exist and tickers that now point somewhere else.

How big the sector is

Sector totals differ wildly by methodology, so treat all of them as approximate and dated:

• Bitwise counted 174 public companies holding a combined 1,187,898 BTC as at 12 May 2026, of which Strategy alone held 818,869 BTC — around 69% of the entire corporate pool (analyst dataset, via secondary reporting).
• The tracker bitcointreasuries.net displayed 215 public companies holding $121bn of digital assets when we accessed it on 26 August 2026. That page carries no as-at date, so it is a live snapshot rather than a measured figure (third-party tracker).
• The Block's corporate tracker showed 119 companies as at 28 July 2026 (third-party tracker). Same sector, a very different count, because the trackers define membership differently.

The concentration matters more than the total. Arithmetically, almost any statement about "corporate bitcoin holdings" is a statement about one company.

Why would anyone buy the share instead of the coin?

Three arguments are made for the share. Each has an immediate cost attached, and the cost is usually left out of the pitch.

1. Access through a wrapper

An ordinary listed share can be held in the ordinary places a UK investor already has — a general investment account, and, depending on the platform and the listing, a pension or ISA. Coins held directly cannot be. That convenience is real, and it is why a great many DAT shareholders are DAT shareholders at all. What you give up: a wrapper does not change what you own. You own equity in a leveraged corporation, not the asset it holds. If wrapper-eligible crypto exposure is the actual goal, how crypto ETNs are treated inside a UK ISA is a separate mechanism with a separate risk set — an ETN is a debt security issued by a provider, not a share in an operating business, and the two fail in different ways.

2. Leverage

A DAT can borrow and issue preferred stock, so it controls more coin than its shareholders' money alone would buy. When the coin rises, the equity rises faster. What you give up: gearing is symmetric and the ranking is not. At Strategy, as at 30 June 2026, roughly $6.75bn of debt and $15.46bn of aggregate preferred liquidation preference sat ahead of the common stock (Form 10-Q, filed 3 August 2026). Between early October 2025 and early June 2026, bitcoin treasury companies shed roughly $62bn of combined market value, falling from about $134bn to about $72bn, and in many cases the share losses exceeded the losses on the bitcoin actually held (reported: Bloomberg, 5 June 2026). That gap is the definition of a company moving from a premium to a discount.

3. The accretion claim

This is the argument that built the sector. If the share trades above the value of the coins attributable to it, the company can issue new shares, buy coins with the money, and end up with more coins per share than it started with. Existing holders are told they got richer in coin terms without buying anything. What you give up: the claim is entirely conditional on the premium existing. Below it, the identical action destroys coins per share. And the metric used to advertise the accretion generally ignores the debt and preferred claims that funded it — see the KPI section, which is the part of this page that matters most.

The machine, step by step

Strip away the branding and the model is a five-step loop. It has one precondition, and the precondition is the whole thing.

1. Start from a premium. The shares must trade above the value of the coins attributable to them. The market shorthand for this is mNAV — market capitalisation divided by the value of the coins held. Above 1.0x, the loop works. At or below 1.0x, it does not. This one ratio governs every decision that follows, and it is worth understanding properly: see mNAV explained.
2. Raise capital. Three main channels, with very different risk profiles. At-the-market (ATM) equity: a continuous drip of new shares sold straight into the market. Convertible notes: cheap coupons — often zero — bought by investors who want the equity option, not the interest. Perpetual preferred stock: a fixed or variable dividend, no maturity date, no repayment obligation, ranking ahead of the common.
3. Buy the asset. Immediately, publicly, and in a size designed to be reported. Weekly 8-K disclosures and press releases are part of the funding mechanism, not merely compliance.
4. Report more asset per share. The company publishes a per-share accumulation metric — Strategy's "BTC Yield", DeFi Development Corp's "SOL Per Share", Twenty One Capital's "Bitcoin Per Share", American Bitcoin's "satoshis per share". The number goes up, and the company describes the raise as accretive.
5. The rising metric supports the premium; the premium makes the next raise accretive. Go to step 2.

Two features of the loop deserve stating plainly. First, the premium is the product. The company is not manufacturing anything; it is converting a market's willingness to pay more than $1 for $1 of coins into extra coins. Second, issuing non-convertible preferred is the most flattering way to run step 2, because those shares never enter the per-share denominator — the metric improves while permanent senior cash claims are added. Strategy's own filings say so explicitly, and we cover that admission below.

The reverse gear, in the same detail

Any explanation of the flywheel that stops here is defective. The same machine runs backwards, step for step, and the sector spent 2026 demonstrating it.

1. The coin falls, and the share falls further. Leverage amplifies the move and the premium compresses at the same time. Bitcoin was marked at $58,714 at 30 June 2026 against $87,515 at 31 December 2025 (Strategy 10-Q, Coinbase Level 1 input). MSTR's one-year total return to 21 August 2026 was approximately minus 65% (company-stated, in an SEC-filed marketing document).
2. The premium becomes a discount. By late December 2025, of roughly 195 treasury companies then in existence, only a handful traded above the value of their coins; Strategy had moved from a substantial premium to a 21% discount, Metaplanet from a 237% premium in July 2025 to 10% in December 2025, and Nakamoto to a 63% discount (reported: DL News, 28 December 2025). The framing used at the time was that the premium era was over.
3. Step 2 inverts. Below 1.0x, issuing shares to buy coins destroys coins per share. The rational action flips from issuing to repurchasing. Sharplink said this out loud: it now only issues equity at a premium to net asset value, and has otherwise repurchased 4,071,223 of its own shares for approximately $41.7m since August 2025 (Q2 2026 results, filed 10 August 2026). CleanSpark's treasury stock rose from 11,759,935 shares ($145.0m) to 42,365,391 shares ($608.161m) over the nine months to 30 June 2026 — it was buying its own stock, not more coins (Form 10-Q, filed 6 August 2026).
4. But the fixed charges do not invert. Preferred dividends and note coupons are contractual cash amounts, unrelated to the coin price. This is the hinge of the entire model and it has its own section below.
5. So the cash has to come from somewhere. There are only three places: issue more equity anyway (dilutive, at exactly the wrong price), sell the asset (reduces the thing shareholders bought the share for), or borrow against the asset (adds a forced-sale trigger). All three have been used in 2026.
6. Where the borrowing is secured, a price fall becomes forced selling. This is the difference between an uncomfortable company and a broken one. Nakamoto Inc had 3,805 of its 4,467 bitcoin pledged to a Kraken loan at 30 June 2026, sold approximately 600 BTC and applied $45.0m to partially repay it, and had pledged a further 688 BTC on 5 February 2026 to meet collateral maintenance requirements (Form 10-Q, filed 13 August 2026). BTCS Inc borrows on the AAVE protocol against its ether — approximately $53.0m of USDT against roughly 46,531 ETH as at 21 August 2026, under a board-approved 50% loan-to-value cap, subject to automatic liquidation if its "health factor" falls below one (Form 8-K, filed 25 August 2026). On-chain, there is no lender to telephone.

Where the reverse gear ends

Four documented end states, all from 2026:

• Reverse split to keep the listing. Nakamoto executed a 1-for-40 reverse split effective 22 May 2026, cutting shares outstanding from approximately 696.1m to approximately 17.4m, after a Nasdaq minimum-bid-price notice in December 2025 and a 99% fall from peak. That pre-split share count is the dilution, made visible. Strive did a 1-for-20 reverse split on 6 February 2026 (reported; company 8-K filings on EDGAR).
• Delisting notice. Upexi received a Nasdaq minimum bid price deficiency notice on 30 July 2026 after 30 consecutive sessions below $1.00, with until 26 January 2027 to regain compliance (Form 8-K, filed 31 July 2026).
• Abandon the strategy and re-brand. ETHZilla sold roughly $114m of ether, ran buybacks, and became Forum Markets Inc effective 25 February 2026. Empery Digital sold roughly 1,400 bitcoin at about $62,200 each — around half its stack — in July 2026 to fund an AI data centre (reported: CoinDesk, 11 July 2026). Riot Platforms and CleanSpark both now describe themselves primarily as data centre businesses. The "permanent, never-sell reserve asset" language of 2025 did not survive contact with a drawdown.
• Liquidate. Satsuma Technology plc shareholders voted on 21 July 2026 to sell all 668 remaining bitcoin, return all capital and cancel the listing — 90.63% for the capital return and 90.59% for the delisting (reported: The Block, 21 July 2026). It had raised £163.6m in convertible notes in August 2025 and sold 579 of 1,199 BTC in December 2025 for around £40m to retire £78m of those notes.

Galaxy Research warned during 2026 that at least five digital asset treasury companies could be forced to sell assets, merge or shut down that year, noting how many trade below 1.0x mNAV. We could not obtain the primary report, so treat that as an analyst view reported secondhand, not as fact. For the failure modes in full — encumbrance, forced sale, dividend suspension, index exclusion — see the risks of bitcoin treasury companies.

Strategy (MSTR): the case study

Strategy Inc is the template for the entire sector and much the largest example, so it is the right case study — not because it is representative in size, but because it discloses more, and in more detail, than anyone else. Everything below is from SEC filings, with the filing named.

The two-stage rename

Most write-ups get this wrong by collapsing two separate events. Stage one was branding only: the Form 8-K of 5 February 2025 was filed by "MicroStrategy Incorporated d/b/a Strategy". The legal entity was unchanged; only the trading-as name moved. Stage two was legal: on 11 August 2025 the company filed a Certificate of Amendment in Delaware changing its legal name from MicroStrategy Incorporated to Strategy Inc, effective at 4:01 p.m. ET that day and effective for Nasdaq trading at the open on 12 August 2025. Board approval alone; no stockholder vote was required under Delaware law. CUSIPs, corporate structure and the MSTR ticker were all unchanged, and EDGAR's own former-names record still shows "MICROSTRATEGY INC" running to 11 August 2025. So: rebrand February 2025, legal rename August 2025.

Holdings and average cost

As at 23 August 2026, per the Form 8-K filed 24 August 2026: 840,447 BTC, aggregate purchase price $63.36bn, average purchase price $75,385 per bitcoin, inclusive of fees and expenses. That same holdings figure was reported unchanged on 9, 16 and 23 August 2026 — the accumulation had stopped.

The audited-quarter position is more revealing than the weekly headline. At 30 June 2026, per the Form 10-Q filed 3 August 2026: approximately 846,000 BTC, cost basis $63.94bn, fair value $49.67bn — carried $14.27bn below cost. The comparatives at 31 December 2025 were 672,500 BTC, cost $50.44bn, fair value $58.85bn. The bitcoin price used was $58,714 at 30 June 2026 against $87,515 at 31 December 2025.

Note the two numbers do not tie, and should not be expected to: the weekly 8-K is a gross holdings count on a Sunday, the 10-Q is a balance sheet on a quarter end. Reconciling them is the first thing to do with any DAT.

The four funding layers

1. At-the-market equity — the dominant channel by a distance. Net ATM proceeds were $15.757bn in the first half of 2026 alone (class A common $8.236bn on 58,455,506 shares; STRC preferred $7.518bn on 75,307,642 shares), against $10.288bn in the first half of 2025. Roughly $20.319bn was raised year-to-date through 23 August 2026 — $12.796bn common and $7.524bn preferred. In the single week 17–23 August 2026 the company sold 18,261,118 MSTR shares for $2,006.5m net. Critically, in the first half of 2026 approximately $1.52bn of common-stock proceeds went into the company's USD Reserve, approximately $546.0m paid preferred dividends and approximately $17.3m paid note interest. That is equity issued to service securities, not to buy bitcoin.
2. Convertible notes — $6,713.659m of principal across six issues at 30 June 2026, all senior unsecured, weighted coupon under 0.5%, all deep out of the money. Coupons run from 0.000% to 2.250%; initial conversion prices from $149.77 to $672.40. Total cash coupon for the whole first half of 2026 was $17.297m — trivial. The real feature is the holder put options: approximately $6.71bn becomes callable for cash between 15 September 2027 and 15 June 2029 if the shares stay below the conversion prices. In May 2026 the company repurchased and cancelled $1.50bn of its 0% 2029 notes for $1.38bn cash — roughly an 8% discount to par — booking a $113.9m gain on extinguishment. That was only possible because the notes trade below par, which is a direct consequence of the equity trading far below the conversion prices.
3. Perpetual preferred stock — five series, aggregate liquidation preference $15.462bn at 30 June 2026. This is the layer that changed the company's character, and it gets its own section below.
4. Other secured debt — $40.0m at 30 June 2026, secured on non-bitcoin assets. At both 30 June 2026 and 31 December 2025 all of the bitcoin itself was unencumbered. There is no margin call in this structure. There are, instead, dates.

What the operating business contributes

Effective the second quarter of 2026 the company reports two segments, "Software" and "Bitcoin" — bitcoin treasury operations became a separate reportable segment, having previously been shown as a non-operating category. That reporting change formalises what the company is: a financing vehicle with an attached software business. Second-quarter 2026 total revenues were $122.4m (up 6.9% year on year) with gross profit of $81.6m; first-half revenues $246.7m; operating cash flow for the half $9.85m. Set that against roughly $1.70bn a year of fixed cash charges and the proportions speak for themselves. Its credit rating is B- from S&P Global Ratings, below investment grade, assigned 27 October 2025 and affirmed in December 2025.

The five perpetual preferred series

There are five, not four, and the fifth is not on Nasdaq. They are all "STR-" words, they are routinely confused with each other, and the differences between them are exactly the differences that matter in a drawdown. Terms below are as at 30 June 2026 from the Form 10-Q filed 3 August 2026, except the STRC rate, which is the rate effective for record dates on or after 1 July 2026.

Four points that apply across all five:

• Every series is perpetual. There is no maturity date, so there is no date on which the company can fail to repay. That cuts both ways: it removes a default trigger, and it makes the dividend obligation permanent.
• Dividends are payable only "when, as and if declared" by the board out of funds legally available. The board may choose not to declare, subject to limited exceptions for STRF and STRK.
• Delaware law adds a constraint that cash cannot solve. Dividends may be paid only out of surplus — assets less liabilities at fair value, less statutory capital — or out of current or prior-year net profits. A large enough drawdown can legally block a dividend even with a full cash reserve. The filings state this explicitly.
• Failure to pay the preferred in full blocks any dividend on the common, and STRK dividends may be paid in class A common stock instead of cash, which dilutes common holders and hands STRK holders share-price risk in place of a cash payment.

Strategy's stack is not unique. Strive lists a Variable Rate Series A Perpetual Preferred (SATA) alongside its ASST common; BitMine has a 9.50% Series A Perpetual Preferred (BMNP); and The Smarter Web Company completed a £210m capital reduction confirmed on 15 July 2026 to create the distributable reserves for what would be the UK's first bitcoin-backed sterling perpetual preferred (reported via RNS coverage). The sector-wide shift from pure equity issuance to yield-bearing perpetual preferreds moves risk onto income-seeking buyers, and it is worth being blunt about where that income comes from: at Strategy in 2026 it has come, in part, from selling bitcoin.

What do "BTC Yield" and "BTC Gain" actually measure?

This is the most important section on the page. "BTC Yield", "BTC Gain" and "BTC $ Gain" are invented, company-defined, non-GAAP measures. They are not accounting figures, they are not audited in the way the financial statements are, and they are not comparable to anything with a similar name elsewhere in finance. Understanding precisely what they measure — and what they conspicuously do not — is the difference between reading these companies and being marketed to by them.

The company's own definitions

From the "Key Performance Indicators" discussion in Strategy's Form 10-Q for the quarter ended 30 June 2026, with the same wording repeated in the earnings release:

• BPS (in Sats) — the ratio between the company's gross bitcoin holdings and its Assumed Diluted Shares Outstanding, expressed in satoshis.
• BTC Yield — the percentage change in BPS (in Sats) from the beginning of a period to the end of that period.
• BTC Gain — the gross number of bitcoins held at the beginning of a period, multiplied by the BTC Yield for that period.
• BTC $ Gain — the BTC Gain multiplied by the market price of bitcoin on the Coinbase exchange at the applicable measurement time.
• Assumed Diluted Shares Outstanding — basic shares plus all shares from assumed conversion of every convertible note and convertible preferred, plus all options, RSUs and PSUs. Explicitly not the treasury method: it ignores vesting conditions, option exercise prices and contractual conversion conditions.

Read that chain and the structure becomes clear. All three headline metrics are functions of one ratio: coins on top, a share count underneath. They are per-share asset accumulation metrics, and nothing else.

What these metrics do not capture

Four things, each of which is decisive:

• Not the price paid. BPS rises if the company buys a coin at any price whatsoever. Buying bitcoin at $120,000 and buying it at $60,000 have identical effects on BTC Yield. Nothing in the metric family registers the quality of the purchase.
• Not the cost of the capital raised. Raising $7.5bn of 12% perpetual preferred and spending it on coins improves the metric. The 12%, forever, does not appear anywhere in it.
• Not the senior claims. The numerator is gross bitcoin. Debt and preferred liquidation preference are simply absent. The company states this itself: the KPIs do not take into account that its assets are subject to all of its existing and future liabilities and to the preferential rights of its preferred stockholders, and they assume all indebtedness will be refinanced or converted into common stock. It adds that different assumptions would produce materially different results and that the metrics "may overstate or understate" the bitcoin available after senior claims.
• Not shareholder return. The company's disclaimer is unusually explicit and worth quoting because it is unimpeachable: BTC Yield "is not equivalent to yield in the traditional financial context", and it is not a measure of the return shareholders may have achieved. Likewise, the filings state that BTC $ Gain does not represent a fair-value gain on the bitcoin, and may be positive in a period in which the company incurred fair-value losses. BPS, they say, does not represent the ability to satisfy financial obligations, and is not book value per share. The filings go so far as to say the KPIs should be used only by sophisticated investors who understand their limited purpose and many limitations.

The mechanical flaw the filings admit

Here is the part that deserves to be better known. Issuing non-convertible preferred stock — STRF, STRC, STRE, STRD — mechanically increases BTC Yield, because those shares never enter the Assumed Diluted Shares Outstanding denominator. The coins bought with the proceeds land in the numerator; the permanent senior cash claim created to buy them appears nowhere.

Strategy says so in its own filings: because those securities rank senior to the class A common stock, bitcoin acquired with their proceeds may not accrete to common shareholders; and buying bitcoin with proceeds from non-convertible instruments, or from convertibles struck above the current share price, has the effect of increasing BPS, BTC Yield, BTC Gain and BTC $ Gain while simultaneously increasing indebtedness and senior claims in a way those metrics do not reflect.

The filings also narrow the word "accretive" to near-uselessness: references to a transaction being accretive or dilutive refer only to the effect on the specified KPI under the stated assumptions, and do not mean the transaction is accretive to earnings, cash flow, book value, enterprise value, intrinsic value or the trading price of the company's securities. When a treasury company calls a raise accretive, that is the meaning it is using.

The metrics can, and do, fall

Reported values, all company metrics: BTC Yield of 5.0% for the second quarter of 2026 and 8.1% for the first half, against 8.7% and 19.7% in the equivalent periods of 2025. BTC Gain of 37,733 bitcoin for the quarter and 54,625 for the half. BTC $ Gain of $2.215bn and $3.207bn. Underlying BPS ran 158,826 sats at 31 December 2024, 194,986 at 31 December 2025 and 210,824 at 30 June 2026, while Assumed Diluted Shares Outstanding grew from 344,897 thousand to 401,283 thousand over the first half of 2026.

Then, by 26 July 2026, the year-to-date BTC Yield had fallen to 4.5% — down from 8.1% at 30 June. The metric went backwards inside a single month. The company's own stated reasons are instructive: the class A common trading at a lower premium to the underlying bitcoin; an increase in sales of class A common for purposes other than purchasing bitcoin, including funding the USD Reserve and paying preferred dividends and interest; and sales of 1,395 bitcoins. We have deliberately not published a later figure: the year-to-date number after 26 July 2026 is not company-published, and the definitive figure will appear in the third-quarter Form 10-Q.

When the definition itself moves

Three redefinitions in this family inside eight months, each of which breaks comparability:

• 1 January 2026 — KPI calculation methodology. Interim-period BTC Yield now uses BPS at the start of the reference period (the fiscal year), less yield already booked, rather than the start of the sub-period, so quarterly figures now sum to the annual one. The company states that when BPS is increasing the new method generally produces higher interim figures, and when BPS is declining it may produce lower, more negative ones. Annual, year-to-date and first-quarter figures are unaffected; other interim figures are not comparable to previously reported ones.
• 23 July 2026 — mNAV. Redefined as market price per class A share divided by "Net Bitcoin Per Share (in USD)", where the net figure deducts out-of-the-money convertible notes and other debt-like instruments and the notional amount of outstanding perpetual preferred, and adds USD assets. The company states plainly that pre-23 July references to its mNAV are not comparable, and that although the metric carries the label NAV it is not equivalent to net asset value. Because the denominator moved from gross bitcoin per share to net bitcoin per share after deducting large senior claims, dividing by the smaller figure produces a materially higher multiple.
• 23 July 2026 — "Amplification". Redefined on the same date, with the same non-comparability warning.

The size of that mNAV effect is worth seeing. At the 21 August 2026 snapshot the company reported an mNAV of 1.01x on its new definition. On the conventional market convention — market capitalisation divided by the gross value of the bitcoin — the same company-stated inputs give roughly 0.77x ($49.60bn market capitalisation over $64.718bn of bitcoin; our arithmetic, not a published figure). Third-party trackers were reporting a basic mNAV of about 0.68x in early August 2026 (secondary and contested). The three numbers disagree because they measure different things; the direction, though, is not in dispute. Whenever you see an mNAV quoted, the first question is whose definition it uses, and the second is what date the definition took effect.

The criticism, stated fairly

Two things are true at once, and both belong in an honest explainer.

In the company's favour: it publishes these metrics alongside an unusually explicit list of their limitations, in its SEC filings, in language a critic could not improve on. Nothing above is a gotcha — most of it is quoted or paraphrased directly from the issuer's own "Important Information about KPIs" disclosure. It also publishes three further invented measures — BTC Hurdle ARR, BTC Breakeven ARR and BTC Floor ARR (10.74%, 2.63% and minus 15.64% respectively at the 21 August 2026 snapshot) — which amount to a candid statement that bitcoin must compound above the hurdle indefinitely or the preferred stack consumes the common equity. The company's own word for what may be required below the floor is "restructuring".

Against it: the metrics are nonetheless presented as headline achievements, in press releases and marketing decks, using three words — yield, gain, NAV — that every investor already associates with something else entirely. The disclaimers live in the filings; the numbers live in the headlines. And the strongest criticism of the family is structural rather than rhetorical: a measure that improves when you issue 12% perpetual preferred, that ignores the 12%, that ignores the price paid for the coins, and that its issuer says is not a measure of shareholder return, is not a scorecard for a shareholder. It is a scorecard for the accumulation programme.

One caveat on our own reporting: we found no independent critique of these metrics from an accounting body, regulator or peer-reviewed source, and we did not check whether the SEC has ever queried the presentation through a comment letter. The criticism here rests on the issuer's own disclosures, which is defensible for an explainer but means there is no external authority behind it.

The same pattern across the cohort

Every DAT of any size has a version of this metric, and the same objections apply to each. DeFi Development Corp reports "SOL Per Share" of 0.066 as at 12 August 2026 against a stated long-term target of 1.0 by December 2028 (company metric, via its results release). Twenty One Capital reports "Bitcoin Per Share" and "Bitcoin Return Rate". American Bitcoin reports satoshis per share, up roughly 11% from about 9,943 to about 10,989 over the quarter to 30 June 2026. Each of these can rise while shareholders lose money, and none of them says anything about leverage, encumbrance or the discount to net asset value.

The fair-value accounting change and its tax tail

If you look at a DAT's income statement and cannot understand how a company with $122m of quarterly revenue posted an $8bn quarterly loss, this is why.

What ASU 2023-08 changed

Until the end of 2024, US GAAP treated cryptoassets as indefinite-lived intangibles under a cost-less-impairment model: write the holding down on any dip, never write it back up, no matter how far the price recovered. A one-way ratchet. ASU 2023-08 replaced that with fair-value measurement, marking holdings both ways through the income statement.

Strategy adopted it on 1 January 2025. The transition entry is instructive: cumulative impairments of $4,058,875 thousand were reversed as part of a total fair-value uplift of $17,881,048 thousand, taking the carrying value of 447,470 BTC from $23,909,373 thousand to $41,790,421 thousand overnight, with a $12.75bn cumulative-effect credit to opening retained earnings net of deferred tax (Form 10-K for FY2025, filed 19 February 2026). The standard does not permit retrospective restatement, so pre-2025 results are not comparable to anything published since.

What it did to reported earnings

Violent, entirely non-cash swings. All figures from the Form 10-Q filed 3 August 2026 and the FY2025 Form 10-K:

The second-quarter 2026 net loss attributable to common shareholders was $8,620,289 thousand, after $400,661 thousand of preferred dividends — the dividends in that single quarter alone are larger than a full year of software gross profit. Against a $122.4m revenue line, the accounting result is essentially a leveraged bitcoin price chart with a small software business attached.

The asymmetry other DATs face. Directly held crypto is now fair-valued both ways, but liquid staking tokens can still be treated as intangibles subject to impairment that is never reversed. Sharplink states this plainly about its $76.1m impairment on LsETH and weETH in the second quarter of 2026: the charges reduce carrying value under US GAAP and are not reversed for subsequent market recoveries, while not reducing the number of tokens held (Q2 2026 results, filed 10 August 2026). Reported book equity for a company holding staked or wrapped tokens is therefore a downward ratchet, which can push it into negative equity even if the token recovers.

The tax consequence

Marking an unrealised position through profit and loss creates a deferred tax position on gains nobody has banked. Strategy's swung from a $2.42bn deferred tax liability at 31 December 2025 to a $4.12bn deferred tax asset on unrealised bitcoin losses (plus $24.4m on realised capital losses from sales) at 30 June 2026 — and then wrote the asset off in full with a $4.60bn valuation allowance against all domestic net deferred tax assets, judged more likely than not not to be realised. The first-half 2026 tax benefit was $1.92bn on a $22.68bn pre-tax loss: an effective rate of just 8.5%, against a 28.0% provision rate on first-half 2025 income.

The mechanism is worth stating in one sentence: when the coin is up, the company books a real deferred tax liability on paper gains it has not realised; when the coin is down, the resulting tax asset is written off. The asymmetry is unfavourable in both directions. The company notes the valuation allowance could be released if fair value exceeds cost basis again.

The CAMT question

The 2022 Inflation Reduction Act imposes a 15% corporate alternative minimum tax where average annual adjusted financial statement income over any three consecutive years exceeds $1bn. Because ASU 2023-08 pushes unrealised gains through the income statement, those gains fed that measure — and the company had previously disclosed it expected it could become subject to CAMT. On 30 September 2025 the US Treasury and IRS issued interim guidance clarifying that a corporation may disregard unrealised gains and losses on digital assets when computing adjusted financial statement income, and signalled revised proposed regulations along the same lines. Strategy now states it does not expect to become subject to CAMT on unrealised bitcoin gains. Caveat for readers: that rests on interim guidance and regulations that have been intended rather than issued. Separately, the Act's 1% excise tax on net stock repurchases may apply to the company's new buyback programmes.

All of the above is US GAAP and US federal tax, and describes the company's position, not yours. A UK investor holding a US-listed share is taxed under UK rules on that share — dividends as dividend income, disposals under the normal capital gains rules, with the usual complications where a US withholding tax applies. Note too that Strategy expects its preferred distributions to be treated as non-taxable return of capital for US federal purposes, because it does not expect accumulated or current earnings and profits for the foreseeable future — a direct consequence of those enormous losses. That is a statement about US tax treatment for US holders and says nothing at all about how a UK holder would be taxed.

What must be paid in cash, whatever the coin does

Everything above concerns numbers that move with the market. This section concerns numbers that do not. It is the hinge of the whole model.

The annual bill

Strategy's own figure for annual preferred dividends plus debt interest was $1.76bn at both 30 June and 24 July 2026 (assuming a 12.00% STRC rate), falling to $1.703bn at 23 August 2026 — the fall being a consequence of retiring STRC below par, not of any easing in the terms. Against that:

• Cash preferred dividends actually paid: $400.2m in the second quarter of 2026 and $629.2m in the first half — STRC $394.3m, STRD $70.1m, STRF $64.2m, STRK $56.1m, STRE $44.5m. The comparative for the first half of 2025 was $58.1m. That is a more-than-tenfold increase in twelve months. Cumulative preferred dividends paid stood at approximately $1.06bn as at 26 July 2026, across eighteen consecutive months of payments.
• Convertible note cash coupon: $17.297m for the entire first half of 2026. Trivial by comparison — the converts were never the cash problem.
• Software revenues: $246.7m for the half. Operating cash flow: $9.85m for the half.

The Form 10-Q states the position without euphemism: the company does not expect its enterprise analytics software business to generate sufficient cash flow from operations to satisfy its financial obligations or liquidity needs over the next twelve months.

So where does the cash come from?

From issuing securities. In the first half of 2026, approximately $1.52bn of common-stock proceeds funded the USD Reserve, approximately $546.0m paid preferred dividends and approximately $17.3m paid note interest. In the week to 23 August 2026, $300.0m of the $2,006.5m of ATM proceeds went straight into the Reserve. When the premium is gone, that is common shareholders funding preferred shareholders' income, at a share price the company itself describes as trading at a lower premium to the underlying bitcoin.

The cash reserves, and what they are not

Balances at 23 August 2026: USD Reserve $5.10bn and a newly created USD Cash pool $1.59bn — $6.69bn combined, which the company describes as roughly 3.9 years of coverage of preferred dividends and interest, against 16 months at 30 June 2026 and 26 months at 24 July 2026 on the then-smaller reserve. The Reserve policy, adopted June 2026, permits use only for preferred dividends and debt interest and requires a minimum of twelve months' coverage. The USD Cash pool, established 24 August 2026, has no minimum and much broader permitted uses — buying bitcoin, dividends, interest, repurchasing common or preferred, repaying converts, or topping up the Reserve.

Both are internal board policies, not contractual protections. The risk factors say so directly: the USD Reserve policy is an internal policy, does not create contractual rights in favour of holders of the company's securities, is not subject to any contractual mandate or lien, and may be modified, suspended or terminated. A preferred holder has no enforceable claim on it. Anyone reading "reserve" as "ring-fenced" is reading something that is not there.

No covenants, no collateral — but hard dates

Strategy discloses no financial maintenance covenants and no bitcoin collateral; the bitcoin was unencumbered at both 30 June 2026 and 31 December 2025. There is therefore no margin call and no ratio-triggered forced sale. The forcing mechanisms are dates instead: convertible holder put options at 100% of principal on 15 September 2027 ($1.01bn), 1 March 2028 ($2.00bn), 1 June 2028 ($1.50bn), 15 September 2028 ($1.404bn combined) and 15 June 2029 ($0.80bn) — approximately $6.71bn callable for cash inside roughly three years if the shares stay below the conversion prices — plus fundamental-change repurchase rights across the notes and all five preferred series. The company flags that if converts mature or are redeemed without converting it may have to sell common stock or bitcoin, which would decrease BPS and BTC Yield: an effect its own KPIs do not contemplate.

What happens to the preferred in a deep fall

No default is triggered — the preferreds are perpetual equity, so there is no maturity to miss. Instead the dividends simply stop, through two independent routes: board discretion (payable only when, as and if declared) and the Delaware surplus test, which can legally block payment even with cash in the Reserve. The consequences then differ sharply by series. STRF and STRK accumulate, and holders gain the right to elect one director each after four and then eight consecutive missed payment dates. STRC and STRE accumulate but carry no board rights. STRD is non-cumulative: a skipped dividend is permanently forgone, with no board rights to show for it. If you take one thing from the preferred table, take that.

The index dimension

One under-discussed pressure point. Strategy was added to the Nasdaq-100 in December 2024 and retained at the December 2025 reconstitution. It qualified on the published criteria for the S&P 500 in 2025 but was not selected at the rebalance announced 5 September 2025 — index committees never publish reasoning, so every explanation offered for that is speculation and should be treated as such. In October 2025 MSCI proposed excluding from its Global Investable Market Indexes companies whose primary business involves digital asset treasury activities and whose digital assets exceed 50% of total assets, which the company confirms could have included it. On 6 January 2026 MSCI announced it would not implement the proposal for the February 2026 review, but opened a broader consultation and, in the interim, permitted no increases to share counts or inclusion factors for listed treasury companies. The consequence is easy to miss: passive MSCI-tracking money does not absorb the new ATM issuance, removing a large price-insensitive buyer at exactly the moment tens of millions of shares a month are being printed. We could not confirm from an MSCI source dated after 6 January 2026 whether the freeze remains in force; the second-quarter 2026 risk factors still describe it as current, which is the best available evidence.

The documented turn: selling coins to pay the preferred

This is the single most instructive sequence of events in the sector's short history, and it is all in the filings.

On 29 June 2026 Strategy's board authorised a Digital Credit Capital Framework with five components: the USD Reserve policy; a revised STRC dividend rate policy; a $1.0bn preferred repurchase programme across STRF, STRC, STRK and STRD, with STRC the initial priority; a $1.0bn class A common repurchase programme; and a BTC Monetization Program. That last item authorised the company to sell bitcoin for three purposes: to raise up to $1.25bn for the USD Reserve, to fund preferred dividends and debt interest, and to fund repurchases of its own securities.

A company that had spent five years telling the market it would never sell had just given itself formal authority to sell — and then used it.

Company-stated total bitcoin sold year-to-date 2026 was approximately $218.4m as at 26 July 2026. Every sale from June 2026 onwards was below the $75,385 average purchase price; the first, 32 BTC on 26–31 May 2026 at $77,135, was marginally above it. The company was realising losses on the asset in order to service and retire the securities it had issued to buy that asset.

The other half of the trade is the buyback. As at 23 August 2026, $516.6m of the $1.0bn preferred authorisation remained — approximately $483.4m used, all of it on STRC — while the full $1.0bn common authorisation remained unused. The disclosed STRC tranches: 288,930 shares for $25.0m at $86.53 (20–26 July); 912,143 shares for $81.2m (27 July – 2 August, implying roughly $89.02 — our arithmetic); and 1,431,212 shares for $136.4m (17–23 August, implying roughly $95.30 — our arithmetic). All against a $100 stated amount, and all consistent with a company buying back its own preferred at a discount to par.

How do you read one of these companies?

In this order. The order matters, because each step changes how you read the next one, and because the sequence is designed to reach the uncomfortable questions before you have formed an opinion.

1. Identify the registrant, not the brand. Look up the company in SEC EDGAR and read the "former names" record. It will usually tell you what the shell used to be, and it will tell you whether the ticker still matches the story. Names and tickers move independently in this sector.
2. Find the latest dated holdings disclosure — and the second-latest. The weekly 8-K or press release gives the headline count; the 10-Q or 10-K gives the balance sheet. Write down both dates. Reconcile the two: Strategy reported 840,447 BTC as at 23 August 2026 in an 8-K and approximately 846,000 BTC at 30 June 2026 in a 10-Q, and understanding why those differ is the beginning of understanding the company.
3. Ask how much of it is actually unencumbered. This is the most under-reported risk in the sector. Read the digital assets note and every collateral line. Dated examples, all from filings for the quarter to 30 June 2026: MARA Holdings 9,270 of 35,577 BTC (26%) loaned or pledged; Riot Platforms 5,821 of 11,380 BTC (51%) pledged against a $200m Coinbase Credit facility; Nakamoto 3,805 of 4,467 BTC (85%) pledged to Kraken; CleanSpark 1,719 BTC posted to derivative counterparties and carried as a receivable rather than as bitcoin. DeFi Development Corp is the sharpest case: a headline of 2,294,576 SOL around 12 August 2026 against 888,000 SOL units carried at fair value on the balance sheet, with the remainder split across pledged collateral, locked SOL bought below spot, and — separately and importantly — 621.9 thousand SOL delegated by customers, which are not company assets at all.
4. Read the capital structure note in full. For each convertible: principal, coupon, maturity, conversion price, and the holder put date. For each preferred: rate, cumulative or not, liquidation preference, redemption terms, and whether missed dividends can ever be recovered. For any secured debt: what secures it, and at what loan-to-value.
5. Total the annual cash charges. Then divide by operating cash flow. If the answer is a large number, the company is funded by the capital markets, not by its business, and its solvency is a function of market access rather than of the coin. Strategy: roughly $1.70bn a year against $9.85m of first-half operating cash flow.
6. Work out the discount yourself. Market capitalisation divided by the value of the coins held. Then read the company's own mNAV definition and its effective date, and see whether they agree. If a company redefined the metric, find the old definition — the change is usually informative about which direction the number was heading.
7. Read the KPI limitations disclosure. Whatever it is called — Strategy's is "Important Information about KPIs" — this is where the issuer tells you what its own headline metrics do not mean. It is generally the most useful page in the filing.
8. Compare cost basis to spot. A company sitting far below cost has less room to sell into strength and a weaker story to raise on. Dated examples: CleanSpark's cost basis was $90,188 per bitcoin against a $58,524 mark at 30 June 2026 (filed). Metaplanet's average purchase price was reported at approximately $95,209 (secondary; we could not retrieve the company's own dashboard and this needs a primary check). The Smarter Web Company's net average purchase price was reported at £82,886, given as approximately $111,548 (reported and contested).
9. Check listing and index status. Minimum-bid-price notices, reverse splits, and index-provider treatment all determine whether the funding channel stays open.
10. Read the liquidity and going-concern language in the company's own words. Not the press release. The MD&A. Companies in this sector are often strikingly candid in filings and strikingly promotional everywhere else, and where the two conflict, the filing is the one signed under liability.

Four red flags worth naming

• An undated holdings figure. In a sector that discloses weekly, a number without a date is either careless or convenient.
• A metric with a new definition and an old comparison. If the definition changed and the chart did not restate, the chart is wrong.
• Directional market commentary in issuer communications. Several companies in this cohort publish price framing and historical analogies in their own press releases — BitMine's are an example. That is promotional issuer communication, not analysis, and should be read as marketing regardless of who signs it.
• "Never sell" as a permanent commitment. Empery Digital, Bitdeer, Satsuma, Smarter Web and Strategy itself have all sold. A policy that can be changed by a board vote is a policy, not a property of the asset.

What this means for a UK investor

Most of the filings quoted above are American, and the figures are in dollars because the sources are US filings — we have not converted them, and you should be suspicious of anyone who does so silently at an unstated rate. Four points translate directly to a UK holder.

Protection, and the absence of it

There is no compensation scheme standing behind the value of these shares. The FSCS can pay out where a regulated UK firm fails owing you money or assets; it does not restore investment losses on a share you chose to buy, and cryptoassets themselves sit outside its scope entirely. Holding the share inside an ISA or SIPP changes the tax treatment of gains, not the risk of the underlying company.

Currency and listing venue

A US-listed DAT gives a sterling investor a dollar-denominated share tracking a dollar-priced asset. That is two layers of currency effect between you and any outcome. Strategy's fifth preferred series, STRE, is euro-denominated and listed in Luxembourg, not on Nasdaq — a reminder that the same issuer can span three currencies and two continents.

The UK-listed examples

Two are directly relevant, and both are cautionary rather than exemplary:

• Satsuma Technology plc (SATS), formerly TAO Alpha Plc. Raised £163.6m in an oversubscribed convertible note round in August 2025 and received investor contributions of 1,097 BTC as part of the fundraise. Sold 579 of 1,199 BTC in December 2025 for around £40m to retire £78m of notes maturing that month. Shares lost more than 99% from their June 2025 peak by April 2026. On 21 July 2026 shareholders voted to sell all 668 remaining bitcoin, return all capital and cancel the listing, with a UK High Court confirmation hearing set for 8 September 2026, listing cancellation for 14 September 2026 and payments by 28 September 2026 (reported; and note that sources disagree on whether the quotation was on the London Stock Exchange or Aquis — check the company's own RNS before relying on the venue).
• The Smarter Web Company Plc. A web design business founded in 2009 that adopted a bitcoin treasury policy after an April 2025 IPO, moved from the AQSE Growth Market to the London Stock Exchange Main Market with first trading on 3 February 2026 — the first new Main Market listing of that year — and fell a reported 22.5% in early trading to 31.25p, sitting 27.3% below its 43p debut price within three days (reported). Its holdings figure is genuinely contested: mid-2026 reports include 2,712, 2,805, 2,830, 2,840 and 2,878 BTC, and a 3 August 2026 announcement of a purchase taking the total to 2,712 is lower than earlier reported totals, implying disposals in between. We are not publishing a single number for it, because we cannot verify one. Get it from the company's own RNS.

The Smarter Web capital reduction of £210m, confirmed 15 July 2026, was made to create distributable reserves for a planned sterling perpetual preferred — the same instrument type discussed above, aimed at UK income buyers. Anyone considering that class of security when it arrives should read the cash obligations section first, and should be clear that a perpetual preferred with an attractive headline rate is not a deposit, not a bond, and not covered by anything.

Where to go next

If you want the roster rather than the mechanism — who holds what, on what date, in which currency, on which exchange — see the list of digital asset treasury companies, which carries the dated holdings and the naming changes. If you want the failure modes examined one at a time, the risks page linked earlier takes each in turn.

Sources and methodology

Every company figure on this page carries a date and names its filing. The primary sources are SEC filings, retrieved from EDGAR: Strategy Inc's Form 10-Q for the quarter ended 30 June 2026 (filed 3 August 2026), its Form 10-K for FY2025 (filed 19 February 2026), the weekly Forms 8-K of 3, 10, 17 and 24 August 2026, the rename 8-Ks of 5 February 2025 and 11 August 2025, and the Free Writing Prospectus filed 24 August 2026. Cohort figures come from Strive's Form 8-K of 24 August 2026, Riot Platforms' Form 10-Q, CleanSpark's Form 10-Q, Nakamoto's Form 10-Q, DeFi Development Corp's Form 10-Q, BTCS Inc's Form 8-K of 25 August 2026, Upexi's Form 8-K of 31 July 2026, Sharplink's Q2 2026 results, MARA Holdings' Q2 2026 shareholder letter and BitMine Immersion's weekly release of 24 August 2026. Corporate lineages come from EDGAR's own former-names records.

What we did not verify, and have therefore labelled or omitted. Metaplanet's holdings and average cost, the European treasury companies' figures, Bitdeer's and Sequans' reported sales, and The Smarter Web Company's current holding all rest on secondary sources we could not tie to a primary filing; each is marked as reported and none is presented as a settled figure. Twenty One Capital's holdings figure appears identically in datasets three months apart, so we have not used it. We have not published a year-to-date BTC Yield figure later than the company's own 4.5% as at 26 July 2026. Every calculation we performed ourselves — the roughly 0.77x conventional mNAV, and the implied STRC repurchase prices — is labelled as our arithmetic.

Editorial position. This page is educational. It contains no recommendation, no price expectation and no view on the merits of any company, share or coin. It is not regulated financial advice, and a decision to buy or avoid any of these securities is one to take with a qualified adviser, or not at all. Our methodology page and corrections policy set out how we source and fix things; if you can show us a filing that contradicts anything above, we will change it.

Related digital asset guides

• What are digital assets? A UK primer
• mNAV explained: the ratio the whole sector runs on
• Bitcoin treasury company risks, one at a time
• The digital asset treasury companies list
• Crypto ETNs and the UK ISA rules
• Crypto CGT in the UK (2026/27)
• Investment scam checklist