mNAV explained: the number that decides if a treasury company works

What is mNAV actually trying to measure?

A digital asset treasury company — a DAT — is a listed company whose main business is holding a cryptocurrency. Strategy Inc holds bitcoin. BitMine Immersion Technologies holds ether. Metaplanet, listed in Tokyo, holds bitcoin. When you buy the share you are not buying the coin: you are buying a residual claim on a pile of coins, wrapped in a corporate structure that also contains debt, preferred stock, cash, sometimes an operating business, and always a management team. If you are new to the category, start with how digital asset treasury companies work and what digital assets actually are.

mNAV asks one question: are you paying more or less than the assets are worth? Take the total market value of the company's shares and divide it by the market value of the coins it holds. If the answer is above 1.0x, the market is paying a premium — every £1 of share price buys you less than £1 of underlying coin. If it is below 1.0x, the shares trade at a discount, and £1 buys you more than £1 of coin. A reading of 2.0x means you are paying two pounds for a pound of bitcoin. A reading of 0.70x means you are paying seventy pence for it. That is the whole idea. Everything else on this page is an argument about how to count.

The name is a problem before the arithmetic even starts. "mNAV" stands for multiple of net asset value, which sounds like the net asset value a fund administrator calculates every evening under a published valuation policy. It is not that. Strategy's own SEC filing says so explicitly: although mNAV "incorporates the label 'NAV', it is not equivalent to 'net asset value' or 'NAV' or any similar metric in the traditional financial context" (Strategy Inc, Form FWP filed 24 August 2026). That is the issuer telling you the word is borrowed, not earned.

Read this first: what you are actually holding

Before any of the arithmetic, the risk. These shares are common equity in a leveraged company that owns a volatile asset. They can fall, and have fallen, very much further than the coin they hold.

• You are last in the queue. Strategy's own filing puts it flatly: "MSTR is common equity, junior to all debt and preferred stock. It is not a spot bitcoin ETF and shareholders have no direct claim on specific bitcoin." (Filed, Form FWP, 24 August 2026.)
• The leverage is disclosed, not hidden — and it grows as the coin falls. Strategy publishes a measure it calls Amplification, defined as its gross bitcoin reserve divided by its net reserve. It was 1.30x as at 21 August 2026 (company metric). By construction that number rises as bitcoin falls, because the senior claims are fixed in dollars and the assets are not.
• The historical record is brutal. Between 31 December 2024 and 30 June 2026, bitcoin fell 37.1% and Strategy's shares fell 70.0% (our computation from filed KPI tables and exchange closes) — despite the company growing its bitcoin-per-share by 32.7% over the same period. Metaplanet closed at ¥305 on 24 August 2026, roughly 83% below its ¥1,895 peak of 16 June 2025 (market data).
• The volatility is not a bug the company is trying to fix. Strategy's 21 August 2026 snapshot reports a negative 65% one-year return, 76% one-year realised volatility and 70% implied volatility, and its own list of "what MSTR is not designed to provide" includes "a stable mNAV or low volatility" (filed).
• There is no safety net. These are ordinary listed shares. The Financial Services Compensation Scheme does not compensate you for investment losses, and it does not cover the underlying cryptoassets at all. The FCA's position on cryptoassets is that they are high risk and largely unregulated. A company failing is not a mis-selling event you can claim for.
• The endgame has already happened to at least one UK company. Satsuma Technology, a UK-listed bitcoin treasury company, was forced by its own shareholders in July 2026 to liquidate its coins and delist. See what happens below NAV for the sequence.

The structural risks of the wrapper — the preferred dividend stack, the convertible maturities, the forced-seller dynamic — are set out at length in the risks of bitcoin treasury companies. Nothing on this page is a recommendation, a valuation, or a view on where any price is going.

Why does mNAV have no agreed definition?

Because nobody has the authority to set one. mNAV is not an accounting standard, not a regulatory definition, and not an industry convention with a standards body behind it. It is a shorthand that companies, data vendors and traders each compute their own way — and the differences are not rounding errors. They routinely exceed the day's move in the underlying coin.

Three answers from one filing, 21 August 2026

Every input below comes from the same document: Strategy Inc's Form FWP filed 24 August 2026 (SEC accession 0001193125-26-363557), containing the "MSTR Investor Briefing August 23, 2026" with market data as at the 21 August 2026 close. All figures filed.

The filed inputs, so you can redo it: MSTR price $119.25; market capitalisation $49.600bn on basic shares (which implies 415,932,914 shares); 840,447 BTC valued at $77,004, giving a gross bitcoin reserve of $64.718bn; USD assets $6.69bn; debt $6.754bn notional; preferred stock $14.966bn notional; net bitcoin per share $118.31.

• Basic: 49.600 ÷ 64.718 = 0.766x.
• Enterprise value: (49.600 + 6.754 + 14.966 − 6.69) ÷ 64.718 = 64.630 ÷ 64.718 = 0.999x. Strategy states an enterprise value of $64.635bn on the same page, which gives 0.9988x — the same number.
• Strategy's own mNAV: $119.25 ÷ $118.31 = 1.008x, which the company reports as 1.01x.

Nothing about the market changed between those three answers. Only the definition did.

The fourth answer: a tracker running on last month's share count

On 25 August 2026 the public tracker at bitcointreasuries.net showed Strategy at a market cap of $48.7bn, basic mNAV 0.74x, EV mNAV 1.01x, and bitcoin per share (basic) of 0.002185 (reported — vendor dashboard, retrieved 25 August 2026). That last figure is the tell. Divide 840,447 BTC by 0.002185 and you get about 384.6 million shares — which is almost exactly the cover page of Strategy's Form 10-Q for the quarter ended 30 June 2026, filed 3 August 2026: 364,585,501 class A plus 19,640,250 class B = 384,225,751 shares as of 24 July 2026.

The problem is that Strategy sells shares every week. Its 8-K filings disclose at-the-market sales of 3,011,361 class A shares (27 July–2 August), 6,585,682 (3–9 August), 3,458,866 (10–16 August) and 18,261,118 (17–23 August) — 31,317,027 shares in total, about 8% of the base (filed). Strategy's own 21 August briefing implies 415,932,914 basic shares, which confirms the reconstruction. Correcting the share count alone, on the tracker's own price and coin inputs, lifts its basic figure from 0.74x to roughly 0.80x (our computation). Our own independent calculation on the 25 August close, set out in full below, gives 0.796x.

Six percentage points of "discount" that exist only because a database had not been refreshed. In a metric that companies explicitly use to decide whether to issue shares or buy them back, that is not a cosmetic error.

And the issuer can change the definition whenever it likes

Strategy redefined its own mNAV on 23 July 2026 and says so in the filing, verbatim: "Prior to July 23, 2026, the Company's use of the term mNAV referred to a different metric so references to the Company's mNAV calculated prior to that date are not comparable to the Company's mNAV calculated after that date." The identical warning is attached to its Amplification metric (filed). Any chart of Strategy's company-reported mNAV that runs across July 2026 is therefore splicing two different measurements together, and we could not establish from the filings we retrieved what the earlier definition actually was.

This is not unique to Strategy. Twenty One Capital changed the denominator of its own bitcoin-per-share metric, excluding roughly 216 million class B shares its registration statement had previously included, on the reasoning that class B "carries no economic rights and is not publicly tradable". The change raises the reported figure by around 62% at a stroke (filed, Form 10-Q for the period ended 30 June 2026). The justification looks sound on the charter language. That is precisely the point: the denominator of every per-share DAT metric is an issuer choice, not an accounting rule.

The four families of formula

Strip away the branding and there are four families in active use. The difference between them is entirely about where senior claims — convertible notes, perpetual preferred stock — are allowed to sit.

Families 2 and 3 are not the same sum

People assume that adding a claim to the top and subtracting it from the bottom must come out the same. They do not. Take a company with assets of 100, debt of 30, no cash, and a market capitalisation of 80:

• Enterprise-value mNAV = (80 + 30) ÷ 100 = 1.10x
• Market cap over net assets = 80 ÷ (100 − 30) = 80 ÷ 70 = 1.14x

Four points of difference on identical facts, and the gap widens as leverage rises. If you ever see two commentators disagree about a DAT's mNAV without either of them saying which family they are in, this is usually why.

How Strategy defines its own — and why the method is better than most

Strategy's formal definition is family 3: the Nasdaq market price per class A share, divided by net bitcoin per share in dollars. Net Reserve is the gross bitcoin reserve, minus the notional of out-of-the-money convertible notes and other debt-like instruments, minus the notional of outstanding perpetual preferred (excluding in-the-money STRK), plus USD assets. In-the-money instruments are instead converted into shares, and the methodology assumes conversion "followed immediately by a hypothetical liquidation" (filed).

That construction avoids a common and serious error: counting a convertible bond twice, once as debt in the numerator and again as shares in the denominator. Many third-party "diluted mNAV" figures do exactly that, or the opposite — treating an out-of-the-money convertible as neither debt nor shares, so that it disappears from the calculation entirely.

Strategy's method has its own soft spots, and the filing names some of them. The preferred notional "may not be equivalent to its liquidation preference or redemption amount, nor are any accrued and unpaid dividends included". Non-dollar preferred is converted at exchange rates fixed at 12:30pm New York time on the most recent Friday. And the "Assumed Diluted Shares Outstanding" figure Strategy uses elsewhere explicitly "does not take into account any vesting conditions... the exercise price of any stock option awards or any contractual conditions limiting convertibility" — it is not a treasury-stock-method calculation (filed, Form 10-Q for the period ended 30 June 2026).

A worked calculation you can reproduce

Here is Strategy Inc's mNAV as at the close on 25 August 2026, computed three ways, with every input dated and sourced. This is our computation, not a company figure. Redo it and you should get the same answers; do it a week later and you will not, which is the point.

Step one: value the coins

840,447 × $78,814.59 = $66.239bn of bitcoin.

Step two: value the equity

415,542,778 × $126.84 = $52.707bn market capitalisation.

Step three: run it three ways

• Basic: 52.707 ÷ 66.239 = 0.796x. A 20% discount to the coins.
• Enterprise value: (52.707 + 6.714 + 14.945 − 6.69) ÷ 66.239 = 67.676 ÷ 66.239 = 1.022x. A 2% premium.
• Look-through net assets per share: (66.239 + 6.69 − 6.714 − 14.945) = $51.270bn of assets attributable to common, ÷ 415,542,778 shares = $123.38 per share of net bitcoin-plus-cash. Against a $126.84 price that is 1.028x. A 3% premium.

Same day, same company, same filings: a 20% discount or a 3% premium, purely on definition. A reader told "MSTR trades at a 20% discount to its bitcoin" and a reader told "MSTR trades at a small premium" have both been told something true, and neither has been told enough.

The arithmetic that matters

This is the part worth learning properly, because it is exact, it is short, and it explains almost every capital-markets decision these companies make.

Let the company have N shares and assets worth A, so assets per share are a = A ÷ N. Its mNAV is m, which means the market capitalisation is m × A, and so the share price is p = mA ÷ N. Now issue a fraction f of new shares — f = 0.10 means a 10% increase in the share count — at that market price, and spend every penny raised on more of the same asset.

• Cash raised = fN × p = fN × (mA ÷ N) = fmA
• New assets = A + fmA = A(1 + fm)
• New share count = N(1 + f)
• New assets per share = A(1 + fm) ÷ N(1 + f) = a × (1 + fm) ÷ (1 + f)

That multiplier is the whole flywheel. It is greater than 1 whenever m > 1, exactly 1 when m = 1, and less than 1 whenever m < 1 — regardless of f. There is no momentum term, no memory, no management skill in it. It is purely the ratio of the price you sell equity at to the value of the assets you buy with it.

Issuance, worked in full

Take a toy company — illustrative arithmetic, not market data — with 100,000,000 shares and 10,000 BTC, bitcoin at $80,000. Its bitcoin NAV is $800m, or $8.00 of bitcoin per share. It issues 10,000,000 new shares (f = 10%) and spends the proceeds on bitcoin at $80,000.

• At 2.00x mNAV (share price $16.00): raises $160m, buys 2,000 BTC. Now 110m shares hold 12,000 BTC. Bitcoin per share goes from 0.00010000 to 0.00010909 — +9.09%. Check against the formula: (1 + 0.10×2.00) ÷ 1.10 = 1.20 ÷ 1.10 = 1.0909.
• At 1.00x mNAV (share price $8.00): raises $80m, buys 1,000 BTC. Now 110m shares hold 11,000 BTC — bitcoin per share is unchanged at 0.00010000. 0.00%. (1.10 ÷ 1.10 = 1.0000.)
• At 0.80x mNAV (share price $6.40): raises $64m, buys 800 BTC. Now 110m shares hold 10,800 BTC — 0.00009818 per share, −1.82%. (1.08 ÷ 1.10 = 0.9818.) Existing holders have handed the new buyer a slice of their coins for less than the coins were worth.

Above 1.0x the new shareholder subsidises the existing ones. Below 1.0x the existing shareholders subsidise the new one. That single sentence is the reason companies issue equity relentlessly at a premium and stop abruptly at parity — and the reason a company that keeps issuing below parity is transferring value away from the people who already own it.

Two caveats the formula hides. First, it assumes you sell at the market price and buy the asset at the market price with no fees, no slippage and no market impact; real at-the-market programmes push the share price down as they sell and can push the coin price up as they buy. Second, it assumes the proceeds actually go into the asset. In August 2026 Strategy raised $20.319bn year-to-date through 23 August — $12.796bn of common and $7.524bn of preferred (filed) — but the recent proceeds went into a dollar reserve, dollar cash and preferred buybacks rather than into bitcoin. An at-the-market programme does not stop working when mNAV falls below 1.0x. It simply stops being accretive.

The buyback inversion

Run the same machine backwards. Repurchase a fraction f of the shares at the market price p = mA ÷ N, funding it by selling assets:

• Cash needed = fN × p = fmA
• Remaining assets = A − fmA = A(1 − fm)
• Remaining shares = N(1 − f)
• New assets per share = a × (1 − fm) ÷ (1 − f)

The inequality flips. This multiplier is greater than 1 whenever m < 1 and less than 1 whenever m > 1. Buybacks funded by asset sales are accretive below 1.0x and destructive above it — the exact mirror image of issuance.

Back to the toy company: 100m shares, 10,000 BTC, bitcoin $80,000, $8.00 of bitcoin per share. It buys back 10,000,000 shares (f = 10%).

• At 0.50x mNAV (price $4.00): costs $40m, funded by selling 500 BTC. Left with 90m shares and 9,500 BTC = 0.00010556 per share, +5.56%. Check: (1 − 0.10×0.50) ÷ 0.90 = 0.95 ÷ 0.90 = 1.0556.
• At 1.00x mNAV (price $8.00): costs $80m, sells 1,000 BTC. Left with 90m shares and 9,000 BTC = 0.00010000. Unchanged.
• At 1.50x mNAV (price $12.00): costs $120m, sells 1,500 BTC. Left with 90m shares and 8,500 BTC = 0.00009444, −5.56%. The company has destroyed coin per share by buying its own stock at a premium.

This is why buyback announcements cluster below 1.0x and equity issuance clusters above it, and why Metaplanet has a stated policy of considering repurchases when its mNAV falls under 1.0x (reported). It is also why the 1.0x line is dangerous: at exactly parity both actions do nothing, and a small methodological change to how you count preferred stock can flip a company's stated policy from "issue equity and buy coin" to "sell coin and buy equity". NYDIG's Greg Cipolaro has made precisely this criticism.

Accretion in coin terms is not a return in money terms

The flywheel is a real mechanism and it genuinely works — and it can still leave you with a catastrophic loss. Strategy's own filings decompose the point exactly. Between 31 December 2024 and 30 June 2026:

• Bitcoin per share rose from 158,826 to 210,824 sats: +32.74% (filed KPI tables)
• Bitcoin fell from $93,390 to $58,714: −37.13% (filed, Coinbase 4pm ET)
• Diluted mNAV fell from 1.95x to 0.70x: −64.10% (our computation)
• The shares fell from $289.62 to $86.93: −69.98% (market data)

Multiply the first three: 1.3274 × 0.6287 × 0.3590 = 0.2996. The actual share return was 0.3002. It is an identity, to rounding. Your return as a shareholder is coin-per-share growth × the coin's return × the change in the multiple. Management can only influence the first term. Over this period the other two buried it.

What drives the premium up?

A rational premium needs a reason why the wrapper is worth more than its contents. Several have been offered, and they vary enormously in durability.

• Wrapper access. Some buyers can hold a listed share but not a coin — mandates that permit equities only, brokerage accounts with no crypto access, tax wrappers, retirement accounts. Historically this was the strongest argument, and it is the one that decays fastest as spot exchange-traded products spread.
• Embedded leverage. Convertible notes and perpetual preferred stock let the company buy more coin than the equity alone could, without immediately diluting the common. The equity becomes a geared claim on the asset. In a rising market that gearing is worth paying for.
• The accretion flywheel itself. If the market believes the premium will persist, it should price in the future coin-per-share growth that issuing at that premium will generate — which is a genuine, arithmetically real source of value, and also perfectly circular. The premium justifies issuance; issuance justifies the premium.
• Index and passive flows. Inclusion in equity indices brings buyers who are not making any judgement about mNAV at all.
• Optionality and the operating business. Some DATs have real businesses, lending ambitions, or venture stakes attached. Twenty One Capital's CEO has set out plans including acquiring operating businesses on a Berkshire model and building a credit business (filed shareholder letter, 11 August 2026).
• Narrative, float and flow. Retail momentum, a heavy options market, constrained float and short interest can all support a price with no reference to assets at all. NYDIG calls this the "memetic premium", and the word is doing honest work: it names something real without dignifying it.

Notice that only the first two are structural. The rest are either circular or sentiment. That asymmetry is what makes the collapse so fast when it comes.

What collapses the premium

• A cheaper wrapper arrives. Spot crypto ETPs do the access job with lower fees, daily transparency and no capital structure. Cipolaro's formulation is hard to argue with: "if all the DAT is doing is issuing stock to buy crypto, it should trade at no premium to NAV — these are called ETFs."
• The company issues into its own premium. Continuous at-the-market selling is, mechanically, a permanent supply of stock at the market price. The flywheel that pays existing holders is also the thing capping the share price.
• The senior stack grows. Every dollar of preferred and convertible debt raised to buy coin sits ahead of the common. Strategy's own 10-Q concedes the consequence in plain language: because those instruments "rank senior to our class A common stock... additional bitcoin acquired using the proceeds from the sale of such instruments may not accrete to common shareholders" (filed). That is the issuer stating the defect in gross bitcoin-per-share and basic mNAV.
• Leverage rises automatically when the asset falls. Fixed dollar claims against a falling asset means the residual equity claim shrinks faster than the asset does. Strategy's Amplification metric was 1.30x at 21 August 2026 and rises as bitcoin falls, by construction (company metric).
• The fixed cash cost becomes a forcing function. As at 30 June and 24 July 2026 Strategy's expected annual preferred dividends plus interest were approximately $1.76bn — around $4.8m a day (filed). The same 10-Q states: "We do not expect cash and cash equivalents generated by our software operations to be sufficient to satisfy our short-term or long-term liquidity needs", and "We may sell bitcoin to satisfy our short-term or long-term liquidity needs... even when other sources of liquidity are available to us." A discount is an inconvenience. A discount plus a non-negotiable daily cash bill is a mechanism.
• Reflexivity in reverse. Once the premium goes, so does the accretion story, which removes a reason to pay a premium, which deepens the discount. The gear runs both ways at the same speed.
• Supply of imitators. Lee, Shleifer and Thaler documented in 1991 that new closed-end funds get launched precisely when seasoned ones trade at premiums. The 2024–25 wave of DAT launches maps onto that pattern almost exactly.

The historical record shows sub-parity trading is not a novelty. MicroStrategy traded at roughly 0.72x basic mNAV at 30 June 2022 — 129,699 BTC at $19,785 against a market capitalisation of about $1.857bn (our computation from the Form 10-Q for the period ended 30 June 2022 and exchange closes). On an enterprise-value basis, with roughly $2.4bn of debt outstanding, the same company was far above 1.0x. The definitional fork visible in 2026 was already there in 2022.

What actually happens when a company trades below NAV?

Not much, for a while. Then, quite suddenly, the company starts doing the opposite of everything it said it would do.

The documented descent

The table below reconstructs Strategy's quarter-end diluted mNAV entirely from SEC filings — assumed diluted shares outstanding, bitcoin held and the bitcoin price at each period end from the KPI tables — plus exchange closing prices. Our computation. It validates: the bitcoin-per-share figures implied by these inputs reproduce Strategy's own published series in sats exactly (158,827 computed against 158,826 filed, then 176,266, 190,100, 194,986, 201,170 and 210,824 all matching), which confirms the share counts and holdings used.

1.97x to 0.70x in twelve months. Basic mNAV crossed below 1.0x intraday on 12–13 November 2025, reported at 0.97x, the first time since January 2024 (reported). Note also that the widely repeated November 2024 "peak premium" is cited at both 3.4x and 3.89x for the same date and cannot be settled from filings, because daily share counts are not disclosed. Treat any single-figure peak claim as definition-dependent.

The reverse gear, written into policy

On 29 June 2026, with the shares near 0.70x, Strategy's board formally authorised a "Digital Credit Capital Framework" in an 8-K filed that day. Five components (all filed):

1. A board-approved USD Reserve policy, usable only for preferred dividends and debt interest, with a minimum of twelve months of those obligations held. Balance $2.55bn at 28 June 2026.
2. A revised dividend policy for its STRC preferred — including the notable statement that the company "will not necessarily increase the STRC dividend rate solely because STRC trades below its stated amount".
3. A $1.0bn programme to repurchase its own preferred stock.
4. A $1.0bn programme to repurchase class A common stock.
5. A "BTC Monetization Program" authorising bitcoin sales to raise up to $1.25bn, to fund preferred dividends and interest "when management determines that it is more advantageous than issuing class A common stock", and to fund repurchases.

A bitcoin treasury company writing bitcoin sales into its capital policy is the reverse flywheel formalised. And it was used. Strategy sold bitcoin five times in 2026, beginning with 32 BTC on 26–31 May 2026 at an average $77,135 — its first sale since 2022 — followed by the four larger disposals below (all filed, Forms 8-K):

Of the 27 July–2 August tranche, $52.4m funded preferred dividends and $52.3m funded preferred repurchases. Holdings went from roughly 846,000 BTC at 30 June 2026 to 840,447 at 23 August 2026 (filed), the difference — 5,553 coins — being exactly the three sales above; no bitcoin was purchased in the period.

Retiring senior claims at a discount

Two of Strategy's below-parity responses are textbook and defensible. In Q2 2026 it repurchased $1.50bn principal of its 2029 convertible notes for $1.38bn cash, an approximate 8% discount to par, booking a $113.9m gain on extinguishment, funded from cash reserves (filed). Buying back your own discounted senior claims with cash is genuinely accretive to the residual equity.

The second is more contested. Between 20 July and 23 August 2026 Strategy repurchased 5,173,025 STRC preferred shares for $483.4m — an average of $93.45 against a $100 stated amount, leaving $516.6m of the $1.0bn authorisation (filed, weekly 8-Ks). Retiring a $100 senior claim for $93.45 removes a claim at a discount. But some of that was funded by selling bitcoin below cost, and some by issuing common stock at roughly 0.77x basic mNAV. As at 23 August 2026 no class A common stock had been repurchased under the $1.0bn common programme at all.

What others have done at the same crossroads

• Metaplanet authorised a ¥75bn (about $500m) buyback of up to 150 million shares — 13.13% of issued shares — on 28–29 October 2025, running to 28 October 2026, financed by a bitcoin-collateralised credit facility (reported; we could not retrieve the primary TSE disclosure). Financing a buyback with debt secured on the coins keeps the coin count intact but adds leverage, which raises the enterprise-value mNAV even as it lowers the basic one. The same action, two opposite-looking metrics.
• Empery Digital (formerly Volcon) committed publicly in February 2026 to selling bitcoin in daily lockstep with share repurchases: "Sales of bitcoin on any given day will be made in coordination with contemporaneous share repurchases, to the extent practicable, to effectively lock in the accretion" (filed 8-K exhibit, 2 February 2026). As at 30 January 2026 it had repurchased 14,041,998 shares at an average $6.94 under a $150m programme — funded "primarily... by borrowings" under $150m of facilities of which $105m was drawn. Borrowing against a falling asset to buy back stock is a real strategy with a real failure mode.
• Consolidation. Strive acquired Semler Scientific in an all-stock deal completing 16 January 2026, each Semler share converting into 21.05 Strive class A shares, taking combined holdings to roughly 12,798 BTC at close (filed / company releases). Buying another DAT's coins at a discount to spot is the acquisitive form of the same below-parity arbitrage. Strive shares fell 12–15% on shareholder approval.
• Exits. Prenetics liquidated about 510 BTC in May 2026 and adopted a no-further-purchases policy; Genius Group liquidated its entire treasury in early April 2026 to repay about $8.5m of debt; Bitdeer emptied its 943 BTC treasury by 20 February 2026 to fund an AI data-centre pivot; Sequans sold $100m of bitcoin to repay debt; ETHZilla sold about $40m of ether from 24 October to fund buybacks while trading at a 30% discount to NAV (all reported, not individually checked against filings).

The completed UK endgame: Satsuma Technology

Satsuma Technology PLC is the case a UK reader should look at hardest, because it finished. The regulatory announcement trail runs: a requisition notice from shareholders (28 January and 2 February 2026), a circular and notice of general meeting (19 February), a board update (16 March), a cost reduction programme and a further bitcoin purchase (2 April), another bitcoin acquisition (24 April), a proposed return of capital and delisting (24 June), the result of the general meeting (20 July) and an update on the reduction of capital (4 August 2026).

Reported outcome: two special resolutions passed with 90.63% and 90.59% of votes cast, against the majority board's recommendation; roughly 668 BTC to be sold; approximately £163.6m originally raised through convertible notes; expected return to shareholders of £26.8m–£30m after £2.7m of wind-down costs, or £66m–£70m including a December 2025 sale of 579 BTC for £40m; listing cancellation scheduled for mid-September 2026. These figures are from secondary reporting — we could not retrieve the primary RNS text, only the regulatory index confirming the dates and headlines, and they should be checked against the RNS before being relied on.

Two details deserve emphasis. The board was still buying bitcoin in April 2026 while the shares traded at a fraction of the value of the coins already held — every one of those purchases was, by the arithmetic above, transferring value away from existing shareholders if funded by equity. And it took an activist holder and a 90% shareholder vote against the board to stop it.

The counterweight from inside the industry is worth reading too. Twenty One Capital's CEO wrote in an SEC-filed shareholder letter dated 11 August 2026: "at the time of writing, the stock trades at a material discount to the Bitcoin we hold. That gap could be viewed as a misallocation of capital; we share that view." And: "Twenty One is not a substitute for Bitcoin. Investors who want pure Bitcoin exposure should understand that Bitcoin itself is the cleanest expression of that view." (Filed.)

The closed-end fund analogy, and where it breaks

Every British investor already has the right mental model for this, and it is not a crypto model. It is the investment trust.

An investment trust is a closed-ended company holding a portfolio. Its share price is set by supply and demand for the shares, and its net asset value is set by the portfolio. The two are related but not tied, and the gap between them — the discount or premium — is a permanent feature of the sector, not a malfunction. A DAT is structurally the same object with one asset in the portfolio. If you want the UK-specific mechanics, see investment trust discounts and premiums explained.

Fifty-three years of UK discounts

The Association of Investment Companies published a 53-year dataset in January 2026 covering 29 December 1972 to December 2025. It is the single best antidote to the idea that a discount is a temporary anomaly that must close.

The widest average discount recorded was 41% in October 1976; the narrowest was 2% in December 1993. The longest unbroken run of double-digit average discounts lasted sixteen years and six months, from December 1972 to June 1989. A second run lasted from June 1997 to January 2001, and a third from May 2022 until April 2026. The average discount was 9.6% at 31 May 2026, the first single-digit reading in nearly four years, having peaked at 19% in October 2023. Sector dispersion is wide — UK residential property trusts sat at −37.4%.

The academic anchor is Lee, Shleifer and Thaler, "Investor Sentiment and the Closed-End Fund Puzzle", Journal of Finance 46(1), March 1991, pp. 75–109. It documents a lifecycle that will feel uncomfortably familiar: funds float at roughly a 10% premium, the premium disappears within about 120 days, and the shares then trade at an average 10% discount. And new funds are launched when seasoned funds trade at premiums.

Where the analogy breaks

It breaks in both directions, which is why it is worth stating carefully rather than deploying as a rhetorical move.

1. The missing arbitrage. This is the important one. A spot exchange-traded product creates and redeems shares at NAV, which is the mechanism that forces price towards value. Strategy's own risk factors draw the distinction: spot bitcoin ETPs are "able to closely track the price of the underlying assets they hold through share creation and redemption", are statutory trusts operating under a trust agreement with stated investment objectives, and provide daily transparency as to holdings and NAV. Strategy is a Delaware corporation, does none of these things, and states it is "not required to provide daily transparency as to our bitcoin holdings or our daily net asset value" (filed). No create/redeem means nothing forces the gap shut. That is exactly why investment trust discounts persist for sixteen years.
2. DATs have more levers than a fund. A closed-end fund can buy back shares. A DAT can issue equity, repurchase equity, repurchase its own debt at a discount, sell assets, borrow against them, or buy a rival. A discount is more actionable here than in a traditional trust — which cuts against the pure-sentiment reading.
3. DATs carry a capital structure that most trusts do not. Perpetual preferred with double-digit dividends and convertible maturities mean a DAT discount can be structurally caused by the stack rather than by sentiment. A trust at a 15% discount and a DAT at a 15% discount are not the same object.
4. The NAV itself is not the same kind of number. A trust's NAV is calculated and published daily by an administrator under a written valuation policy. A DAT's "NAV" is whatever the calculating party decides it is — which is where this whole page started.
5. Operating businesses. A pure asset-to-market-cap comparison ignores whatever else the company owns, from Strategy's software business to BitMine's venture stakes.

Is mNAV a good metric at all?

Partly. It answers a real question that nothing else answers as directly, and the (1 + fm)/(1 + f) arithmetic it feeds is genuinely important. But the criticisms are serious, and the most forceful ones come from inside the industry. Here they are, consolidated.

1. No standard. At least four formulas, producing 0.766x to 1.008x for one company on one day — and far wider ranges for smaller ones.
2. Basic mNAV ignores senior claims entirely. The issuer's own filing concedes that assets bought with preferred proceeds "may not accrete to common shareholders".
3. Notional versus market value. An enterprise-value mNAV using face value overstated Strategy's debt claim by about $454m at 30 June 2026. Preferred notional may differ from redemption value, and accrued unpaid dividends are typically excluded.
4. Convertible double-counting. An out-of-the-money convertible counted both as debt in the numerator and as shares in the denominator is charged twice. Counted as neither, it vanishes. Strategy's own method avoids this; many third-party "diluted" figures do not.
5. Denominator opacity. No major data vendor documents what goes into its diluted share count. Blockworks Research's dashboard notes say fully diluted mNAV "accounts for shares that may be issued via warrants (e.g. pre-funded or stapled), RSUs, stock options, or convertible debt" and that share counts are "updated as frequently as the Company's announcements allow" — neither of which tells you how out-of-the-money convertibles are handled.
6. Stale data. Trackers lag at-the-market issuance, and these companies issue weekly. See the 8% share-count error above.
7. Non-crypto assets are treated inconsistently. BitMine's 23 August 2026 release blended 5,847,611 ETH at $2,440 with 210 BTC, a $180m stake in Beast Industries and an $89m stake in Eightco — described as "moonshots" — plus $308m of cash and securities, into a headline $14.9bn. The release contains no share count and no mNAV at all (filed). Whether BitMine's mNAV is about 1.05x on pure ether NAV or about 1.005x on the broader figure depends entirely on what you count, and we could not confirm the size of its listed Series A preferred, so any enterprise-value figure for it is incomplete.
8. Issuer discretion. Strategy redefined mNAV on 23 July 2026. Twenty One redefined its per-share denominator. Neither series is comparable across the change, and both said so.
9. It ignores the cost of the capital stack. Two companies at identical mNAV can carry wildly different fixed obligations. Strategy's $1.76bn a year sits against a software business the company itself says cannot cover it.
10. It does not compare across companies. Twenty One Capital was quoted by one tracker on 25 August 2026 at 0.66x basic and 1.24x diluted simultaneously — because the "diluted" count included 215,736,011 class B shares that the company's own 10-Q says carry no economic rights: "holders of Class B Common Stock will not have any economic rights" (filed). On the 346,636,211 economically entitled class A shares at $6.51, against 43,514 BTC at $78,814.59, the basic figure is 0.658x (our computation). A 34% discount or a 24% premium, same day, same company. Add ADS ratios, pre-funded warrants and post-merger share bases and naive cross-company comparison stops meaning anything.
11. It is a price, not a value. mNAV tells you what the market is paying relative to one line of the balance sheet. It contains no information whatsoever about whether that price is right.

The scale of point 5 deserves its own illustration. DefiLlama's published methodology shows the same stock screening from 0.06x to 5.27x depending purely on which share count is used across three lenses — shares existing today, plus in-the-money dilution, and everything issuable. Sonnet BioTherapeutics screened at 0.06x on its then 6,754,352 shares and 5.27x on the real post-merger base of 562,862,667. The same article documents Yahoo Finance simultaneously showing one company a float of 904.93m, shares outstanding of 3.95m and implied shares of 476,760 — three numbers on one page giving mNAVs of 119.54x, 0.52x and 0.063x, before an ADS ratio of 1:2,500 is even applied (reported, trader-facing source, but the arithmetic is checkable).

NYDIG's Greg Cipolaro has argued publicly that DATs should abandon mNAV altogether. His charges: ambiguous definition with no industry standard; an "apples-to-oranges problem" where companies use fully diluted shares for coin-per-share but basic shares for mNAV; overstatement of enterprise value by using face rather than market value of debt and preferred; total omission of operating businesses; and — most dangerously — that near 1.0x, "small methodological changes can reverse strategic decisions from issue equity to buy bitcoin to sell bitcoin to buy back equity". His proposed replacement is an Equity Premium (Discount) to NAV: basic shares, but every asset and liability accounted for under a proper NAV framework, with operating businesses valued at enterprise value. On convertibles he sets out two defensible treatments — as straight cash obligations, or with assumed conversion at embedded premiums — and warns against mixing them. That last warning is the practical takeaway even if you keep using mNAV.

Who is telling you the number, and what do they own?

Almost everything written about mNAV is produced by someone with a position. That does not make any of it wrong, but it changes how you should read it, and the sector's own promotional language often needs translating.

• Issuer filings are marketing, even when they are primary sources. A Form FWP is filed with the SEC and the company is responsible for it — and it exists to sell securities. BitMine's August 2026 release forecast that a 30% weekly ether gain has historically preceded moves of +167% and +170%. That is a statement by an interested party in a promotional register, filed or not.
• Company-published mNAVs can exclude the company's own liabilities. SUI Group Holdings published a full mNAV calculation in an 8-K exhibit dated 21 May 2026 — 80,896,554 shares including pre-funded warrants at $1.56, giving a $126.2m market cap against a stated NAV of $138.8m, for 0.91x — with its own caveat that the metrics have "inherent limitations including not taking into account that our assets are subject to all existing and future liabilities" (filed). A company-published mNAV that explicitly ignores liabilities is not comparable to an enterprise-value mNAV, and putting the two in the same table is an error.
• Data vendors are commercial products with undocumented methods. The trackers everyone quotes are useful and fast. None of them publishes enough for you to reproduce their diluted figures.
• The critics have books too. NYDIG is a bitcoin financial services firm. VanEck issues crypto investment products. The most-cited "what is mNAV" explainer is written by a pseudonymous analyst at a DeFi analytics firm, for traders who trade the spread. Being right and being interested are not mutually exclusive, but you should know which you are reading.
• Corporate identity is often newer than the ticker. Many large DATs are renamed shells whose price charts and "shares outstanding" fields on retail data sites span the change without adjustment: Strategy Inc traded as Strategy from February 2025 and legally became Strategy Inc on 11 August 2025; Nakamoto Inc was Kindly MD until December 2025; Strive was Asset Entities until September 2025; Empery Digital was Volcon, an electric powersports company, until July 2025; DeFi Development Corp was Janover until April 2025; SUI Group was Mill City Ventures III until August 2025, and Poker Magic before 2012; FG Nexus has been through four names since 2020 (all filed, EDGAR former-name records). A five-year chart of some of these is a chart of a different company.
• And us. UK Tax Drag takes no affiliate revenue, runs no sponsored content and sells nothing on this page. We are not a broker, not an adviser, and we have no view on where any of these prices go. Our funding is disclosed in full.

How to read any mNAV figure you are shown

Five questions. If a figure cannot answer all five, it is decoration.

1. Which formula? Basic, enterprise value, net-assets-per-share, or net treasury. They are not the same number and the gap grows with leverage.
2. Whose share count, and as at when? Basic or diluted; what is in the dilution; and what date the count comes from. Companies with active at-the-market programmes can move 8% in a month.
3. Debt and preferred at face or at market? And are accrued dividends in or out?
4. What is the timestamp on both prices? The coin price and the share price were taken at different moments. Ask which moments.
5. Who produced it, and what do they hold? Issuer, vendor, sell-side, or someone with a position in the spread.

The UK tax angle, briefly

A share in a treasury company is an ordinary share for UK tax purposes, not a cryptoasset — which means it is taxed like any other equity rather than under HMRC's cryptoasset rules. Gains on disposal fall within Capital Gains Tax, with a £3,000 annual exempt amount for 2026/27 and rates of 18% and 24% for non-property gains; any dividends are dividend income; US-listed shares held in a general account will have US withholding applied unless a W-8BEN is in place. Whether you can hold one at all inside an ISA depends on your platform and on the listing venue — the general wrapper trade-off is covered in ISA versus general investment account. If instead you hold the coin directly, the rules are entirely different and are set out in crypto CGT in the UK and HMRC's Cryptoassets Manual.

What we could not verify

Publishing the gaps is part of the point of a page like this.

• We found no auditable count of how many DATs trade below 1.0x. Figures of "nearly 40%" and "more than 200 DATs" circulate without a named tracker or a date attached. We are not repeating them as fact.
• Metaplanet's primary Tokyo Stock Exchange disclosures and its own analytics dashboard could not be retrieved, so every Metaplanet figure here is corroborated secondary reporting of company statements. Metaplanet's own definition of mNAV is therefore unconfirmed.
• Satsuma Technology's primary RNS text could not be retrieved — only the regulatory index confirming dates and headlines. The vote splits and the expected capital return are secondary reporting.
• BitMine's Series A preferred is outstanding but we could not confirm its size, so any enterprise-value mNAV for BitMine understates its senior claims.
• Strategy's exact share count for 21–25 August 2026 is a reconstruction, not a filed figure, for the reasons set out above.
• Strategy's pre-23-July-2026 mNAV definition is flagged as different in the filing but not described. Anyone charting the company-reported series across that date needs to find and state what the earlier definition was.
• The November 2024 peak premium is cited at both 3.4x and 3.89x for the same date and cannot be settled from filings.

Sources and methodology

Every Strategy Inc figure on this page traces to a filing on SEC EDGAR: the Form FWP filed 24 August 2026 (accession 0001193125-26-363557, containing the investor briefing with 21 August market data and the glossary of defined terms); the Form 10-Q for the period ended 30 June 2026 (accession 0001050446-26-000044, filed 3 August 2026); the Form 8-K filed 29 June 2026 setting out the Digital Credit Capital Framework (accession 0001193125-26-286871); the weekly 8-Ks of 6 July, 27 July, 3, 10, 17 and 24 August 2026 covering bitcoin transactions, at-the-market share sales and preferred repurchases; and the Form 10-Q for the period ended 30 June 2022 for the 2022 comparison. Twenty One Capital's Form 10-Q for the period ended 30 June 2026 and its 11 August 2026 shareholder letter, BitMine's 8-K of 24 August 2026, SUI Group's 8-K exhibit of 21 May 2026 and Empery Digital's 8-K exhibit of 2 February 2026 can be found through EDGAR company search. Former-name records come from EDGAR's submissions data.

UK investment trust discount history is from the Association of Investment Companies, whose January 2026 release covers 29 December 1972 to December 2025. The closed-end fund lifecycle is from Lee, Shleifer and Thaler, "Investor Sentiment and the Closed-End Fund Puzzle", Journal of Finance 46(1), March 1991, pp. 75–109. The regulatory framing for UK readers is the FCA's cryptoassets pages and the FSCS's investment cover rules; UK tax treatment is from gov.uk and HMRC's Cryptoassets Manual.

Prices marked as market data were taken on 25 August 2026: bitcoin $78,814.59 (Coinbase spot, 20:06 UTC), ether $2,454.70, and US closing prices including MSTR $126.83–$126.84, XXI $6.51 and BMNR $24.82. Metaplanet closed at ¥305 on 24 August 2026. Every mNAV on this page moves daily and none of them should be republished without its date. Our general approach is documented on the methodology page, and we correct errors publicly under our corrections policy.

Related guides

• Digital asset treasury companies explained
• The risks of bitcoin treasury companies
• What are digital assets? A UK primer
• Investment trust discounts and premiums
• Investment trusts versus ETFs
• ISA versus general investment account
• Crypto CGT in the UK
• UK ETF guide