Strategy (MSTR) The Bitcoin Machine

MSTR is often treated as a leveraged way to own bitcoin. That description is incomplete.

Owning MSTR means owning the residual equity of a capital structure built around a bitcoin treasury.

Strategy Inc holds 840,447 bitcoins, acquired at an aggregate purchase price of $63.36 billion — an average of $75,385 per coin.

At a bitcoin price of $64,279, that position is worth approximately $54.0 billion and carries an unrealized loss of roughly $9.3 billion. The position is underwater against its own cost basis.

But the bitcoin is not the common shareholder's claim.

Ahead of MSTR common sit approximately $6.75 billion of senior unsecured convertible debt and $15.43 billion of perpetual preferred stock across five series. Strategy also maintains a $4.65 billion USD Reserve intended to support preferred dividends and interest on its outstanding indebtedness. The operating business generated approximately $9.85 million of operating cash flow during the first half of 2026.

That hierarchy matters.

Bitcoin is the asset. Debt and preferred securities are the senior claims. MSTR common is the residual.

And the residual is where the economics become interesting.

Strategy's capital structure currently carries approximately $1.725 billion of annual preferred dividends, plus roughly $35 million of convertible-note interest. The preferred capital has no scheduled maturity; the note interest does. The resulting funding burden is therefore not simply a debt problem. It is a capital-market-access problem.

That distinction is the central subject of this paper.

The mechanism works in both directions.

When MSTR trades at a sufficiently attractive valuation, Strategy can raise capital, acquire bitcoin, increase bitcoin per share, increase the economic value of the common equity and improve its access to further capital.

That is the flywheel.

But the same mechanism can run in reverse.

If bitcoin falls, MSTR can fall with it. mNAV can compress. The economics of new issuance can deteriorate. Financing capacity can narrow. If external financing becomes insufficient to meet the capital structure's cash requirements, bitcoin sales can become part of the funding mix. Holdings then decline, bitcoin per share can fall, and the mechanism that amplified the upside begins to amplify the downside.

That is the reverse.

The important question, therefore, is not simply:

How far can bitcoin fall before MSTR becomes insolvent?

That is the wrong first question.

There is no margin call, no borrowing-base test and no contractual mechanism by which a decline in bitcoin's market price alone forces liquidation of the treasury. Convertible holders do have contractual put rights, but those arise on specified dates independently of the bitcoin price.

The more important question is:

At what point does the financing mechanism stop working?

Under the closed-window stress framework developed in this paper, bitcoin would have to fall approximately 96.1% before the modeled bitcoin-and-reserve pool becomes insufficient to cover the convertible notes, and approximately 67.5% before it becomes insufficient to cover the most junior preferred claim. These are analytical thresholds, not forecasts or predictions of default.

The distinction is crucial.

Strategy's principal vulnerability is not simply asset coverage. It is the preservation of capital-market access required to carry a perpetual preferred burden while maintaining the bitcoin-per-share economics that underpin the equity valuation.

This is why the paper examines MSTR as a capital structure, rather than as a bitcoin price proxy.

It reconstructs the hierarchy of debt, preferred securities and common equity from the governing documents; defines the relevant NAV and mNAV measures; separates bitcoin-per-share accretion from common-equity accretion; examines the convertible put schedule; quantifies the annualised capital-structure cash burden; and stress-tests the point at which the financing loop changes direction.

No probability is assigned to a bitcoin price path. No price target is offered.

The objective is narrower and, in my view, more useful: to show the arithmetic of the machine, identify the thresholds at which its economics change, and establish which observable market signals would confirm or refute the thesis.

A note on method. Throughout, claims are separated into three kinds. FACT denotes something stated in a filing or governing document. CALCULATION denotes arithmetic performed here on those facts. INTERPRETATION denotes an analytical conclusion drawn from them, which a reasonable reader may reject while accepting everything above it. The labels appear where the distinction carries weight rather than on every sentence.

The stub and the mark reconcile exactly to the reported figure. The software business lost $15.6 million in the second quarter of 2026 — and $15.5 million in the second quarter of 2025. On a six-month basis, $30.5 million against $30.9 million. It has been consistently unprofitable across both periods, and the trend is flat rather than improving.

Revenue grew 6.9% year over year. Gross margin fell 2.1 percentage points, from 68.8% to 66.6%. The explanation is a mix shift: subscription services grew 54.0% to become 51.4% of revenue at a 63.8% gross margin, while product support fell 22.7% at an 86.0% margin. The growing line is displacing a more profitable one, so the transition is dilutive to blended margin while it runs.

Two further observations. General and administrative expense of $39.9 million exceeds research and development of $23.1 million by 1.73 times. The filings do not disaggregate G&A, so no attribution of that spending to particular activities is made here. And deferred revenue fell 18.5% over the six months, from $277.6 million to $226.1 million; during a subscription transition this may reflect changes in billing cadence rather than demand, and it is recorded here as an observation to track rather than as evidence about either.

A note on the two different "operating" figures

Two figures in this article look contradictory and are not. The software business records an operating LOSS of $15.6 million for the quarter. The Company reports POSITIVE net cash provided by operating activities of $9.85 million for the half-year. Both are accurate, and the difference is not a rounding artefact — it is the gap between accrual earnings and cash movement.

The balance sheet shows where the cash came from. Accounts receivable fell from $205.7 million at December 31, 2025 to $123.8 million at June 30, 2026 — an $82.0 million collection, which is a cash inflow without being revenue. Working against it, deferred revenue and advance payments fell from $277.6 million to $226.1 million, a $51.5 million outflow. Those two movements net to roughly $30.5 million of cash released, against a six-month software operating loss of $30.5 million. Non-cash charges account for the remainder.

The distinction matters for the rest of the analysis. A receivable can only be collected once. Positive operating cash flow driven by a working-capital release is not evidence that the business has begun funding itself, and it should not be projected forward. Readers evaluating this should review the full statement of cash flows rather than relying on the summary here.

Before any valuation framework is introduced, the asset itself. Balance sheet figures are as of June 30, 2026; the USD Reserve is as of August 9, 2026; market data is as of August 13, 2026.

The basic relationship, before any adjustment for claims:

Bitcoin Value = Bitcoin Holdings × Bitcoin Price

One fact belongs here rather than in a later section, because it conditions everything that follows: at $64,279 against an average cost of $75,385, the position is 14.7% underwater, an unrealized loss of approximately $9.3 billion. Strategy is not sitting on embedded gains that could absorb a shock. It is sitting on a loss.

This is the centrepiece of the analysis. What follows establishes legal seniority as set out in the governing documents, not economic assumptions about who bears loss.

Preferred amounts are stated notional; the aggregate liquidation preference is $15.46bn. Rank 1 is the most senior claim.

The Debt

Six tranches of senior unsecured convertible notes, all issued between March 2024 and February 2025. Principal amounts are as of June 30, 2026. Conversion prices are initial conversion prices.

The 2029 tranche was $3.00 billion at issuance. In May 2026 the Company repurchased and cancelled $1.50 billion of principal for $1.38 billion in cash, recognising a $113.9 million gain on extinguishment. A further $39.8 million of other long-term secured debt brings total debt notional to $6.753 billion.

Every tranche is deeply out of the money. The nearest conversion price is the 2030A at $149.77 against a share price of $97.33 — MSTR must rise 53.9% before the closest tranche reaches parity. The notes were not convertible during the six months ended June 30, 2026 and no conversion requests were received.

The Preferred

Five perpetual series. None has a maturity date. None is collateralised by the bitcoin. All rank junior to indebtedness.

Rank 1 is indebtedness. The euro conversion for STRE is approximate.

The waterfall, established from the governing documents rather than inferred:

Sources for the waterfall: the FY2025 Form 10-K and Exhibit 4.2 for STRF; the STRC Stock Annex for STRC; the STRE offering terms for STRE; and the STRD terms for STRD.

Contractual seniority, structural subordination, economic exposure

These are three different things and the distinction matters.

Contractual seniority is the table above. Structural subordination is separate: claims at operating subsidiaries rank ahead of claims at the parent against those subsidiaries' assets, and the preferred is expressly structurally junior to subsidiary liabilities.

Economic exposure is where the analysis diverges from the legal ranking. Absent a contractual default or payment event, changes in the value of the bitcoin position are first absorbed economically by the common residual, because the debt and preferred claims are fixed in amount and are not marked to the asset. That absorption occurs without any mechanism by which the senior claims can accelerate to realise their position.

Treating 'the preferred' as one instrument is the most common analytical error made about this company. The five series differ in rank, in dividend mechanics, in call protection and in what happens when a payment is missed. Three terms in particular are widely misreported.

Note the shape before the detail: STRC alone is 68% of the preferred notional, and the senior-most series, STRF, is only $1.28 billion. This is not a thin senior tranche protecting a thick junior one. It is one dominant mezzanine instrument with small tranches above and below it.

STRF — the senior series

Pays 10.00% cumulative on a $100 stated amount, with compounded dividends accumulating on any unpaid amount. Ranks senior to every other equity security for both dividends and liquidation. Its protection is its rank; at $1.28 billion of notional it sits above $14.1 billion of other preferred.

STRC — the dominant series, and the one to watch

Pays a cumulative dividend at a rate currently set at 12.00% per annum, on a $100 stated amount, payable semi-monthly. At $10.49 billion of notional it is 68% of the preferred stack and 72% of the current annualised capital-structure cash cost.

The critical term is almost universally misdescribed. STRC is called a variable-rate instrument, which implies a reference rate and a spread. It has neither.

Per the FY2025 Form 10-K, the Company has the right, at its sole and absolute discretion, to adjust the STRC regular dividend rate in the manner set out in the certificate of designations, and the offering documents carry an explicit risk factor that unilateral reduction could cause STRC to accumulate dividends at rates below comparable instruments and depress its trading price.

The rate is therefore an issuer decision, not a market outcome. Its path is disclosed: 10.75% for the month ended December 31, 2025; 11.50% for monthly periods from April 1, 2026; and 12.00% currently. The Company has stated that management does not intend to recommend a change from 12.00% until STRC demonstrates sustained trading at or near its $100 stated amount.

STRC is also cumulative — confirmed in the 10-K, and frequently reported otherwise. Suspension defers the obligation; it does not remove it.

STRE — the euro series, with a forced-issuance provision

Pays 10.00% cumulative on a €100 stated amount, quarterly. Two terms distinguish it, and neither appears in general coverage.

First, unpaid dividends escalate. If a regular dividend is not paid, compounded dividends accumulate quarterly at an initial rate of 10% plus 100 basis points, and that rate increases by a further 100 basis points for each subsequent dividend period until paid in full — to a maximum of 18% per annum. Deferring STRE is progressively more expensive the longer it continues.

Second, and more consequential: failure to declare an STRE dividend by the record date constitutes the issuance of a notice of deferral. Upon that notice, Strategy will use commercially reasonable efforts over the following 60-day period to sell STRK stock, STRD stock, class A common stock and/or other junior stock to raise proceeds sufficient to cover the deferred dividends plus compounded dividends — with application of those proceeds subject to accumulated dividends being paid in full on STRF and STRC first.

Missing an STRE dividend requires Strategy to use commercially reasonable efforts to raise proceeds through junior securities, including common stock. That is an efforts obligation rather than an absolute covenant to issue — but it is written into the instrument, not left to management discretion.

STRK — the convertible series, and the one that can be paid in stock

Pays 8.00% cumulative on a $100 stated amount, the lowest rate in the stack, reflecting an embedded conversion right. The initial conversion rate is 0.1000 shares of class A common stock per share of STRK — an initial conversion price of $1,000.00 per class A share. Against $97.33, that option is 927% out of the money and carries essentially no value in the present regime.

Two further terms distinguish STRK, and both bear directly on the capital-structure cash cost analysis in Section XII.

First, declared regular dividends are payable at the Company's election in cash, in shares of class A common stock, or a combination, with shares valued at 95% of daily VWAP. So $112 million of what would otherwise be a cash obligation is potentially settleable in equity.

That option is impaired rather than free. The number of shares deliverable per share of STRK is capped at the dividend amount divided by a floor price, initially $119.03. MSTR trades 18.2% below that floor, so the cap binds, and any resulting deficiency must be paid in cash or accumulates as compounded dividends. The lever narrows as the share price falls.

Second, and more constraining, there is a mandatory payment trigger tied to the equity programme. If the Company sells class A common stock for cash through a registered public offering — the terms name at-the-market offerings explicitly — during the 90 calendar days preceding a regular record date, then, subject to the terms of any dividend senior stock, it will declare and pay the regular dividend due on the next payment date, to the extent those net proceeds suffice.

Strategy sells common stock through its ATM nearly every week. Suspending the STRK dividend is therefore not freely available while the equity programme runs: issuing common creates a contractual obligation to pay the STRK dividend to the extent the applicable net proceeds suffice.

The STRK liquidation preference has its own history, and it has been resolved — a point on which any source published before mid-2026 is stale.

FACT. On July 7, 2025 the Company filed the STRK Amendment in Delaware, introducing the ratcheting liquidation preference described below. On July 21, 2025 a purported class action was filed in the Delaware Court of Chancery alleging that, under Section 242 of the DGCL, common stockholders were entitled to vote on that amendment. On March 12, 2026 the parties stipulated to dismissal as moot, the Company agreeing to seek stockholder ratification under Section 204 of the DGCL at its next annual meeting and to pay $550,000 of plaintiff's fees.

FACT. That vote was held on June 8, 2026 as Proposal 4. Final results filed June 10 record 302,968,907 votes for, 35,304,265 against and 470,695 abstentions, with 73,095,003 broker non-votes. The proxy provides that abstentions and broker non-votes count as votes against, and that if approved the ratification is retroactive to July 7, 2025, unless otherwise determined in an action brought under Section 205 of the DGCL.

CALCULATION. Votes for represent 73.6% of all shares represented at the meeting and exceed the combined against, abstain and broker non-vote total by 2.8 to one.

INTERPRETATION. The ratcheting liquidation preference should therefore be treated as governing STRK, retroactive to July 2025, rather than as a pending uncertainty — and STRK joins the other four series in carrying that mechanism. The residual risk is narrow and specific: Section 205 preserves the possibility of a contrary court determination, so the matter is settled as a corporate act rather than foreclosed as a legal question.

STRD — the junior series, and the only non-cumulative one

Pays 10.00% non-cumulative on a $100 stated amount. Undeclared dividends never accrue — they are permanently forgone. It ranks senior only to common stock.

It is also non-callable for life, except in certain adverse tax scenarios or once fewer than 25% of the issued shares remain outstanding, and carries a fundamental-change put at the stated amount plus declared and unpaid dividends. Strategy cannot simply retire it when convenient.

The series are chained: a missed senior dividend blocks every junior one

The seniority set out in Section V is not merely a liquidation ranking — it is enforced through the dividend terms. If the Company fails to declare and pay full dividends on STRF, STRC or STRE, it is prohibited from paying dividends on STRK. If it fails to pay STRK in full, it is prohibited from paying dividends on class A common stock and other junior securities, subject to limited exceptions.

There is therefore no configuration in which a senior series goes unpaid while a junior one keeps receiving cash. Suspension propagates strictly downward, which means the first missed payment on STRC — the largest and most expensive series — automatically stops everything beneath it.

The liquidation preference can exceed the stated amount

This is the term most consistently reported wrong across all series, and it inverts a standard assumption.

FACT. Per the FY2025 Form 10-K, the liquidation preference per share of STRF, STRC, STRE and STRD generally approximates the greater of the trading price per share or the stated amount. Following ratification of the STRK Amendment, the same mechanism governs STRK, so all five series now carry it. The STRE certificate specifies the mechanism precisely: adjusted effective immediately after the close of each business day to the greatest of (i) the stated amount, (ii) where a sale transaction has been executed in the preceding ten trading days, the last reported sale price on the immediately preceding trading day, and (iii) the arithmetic average of last reported sale prices over the preceding ten consecutive trading days.

The liquidation preference is floored at the stated amount and can rise above it when the contractual market-price test produces a higher value. It is a greatest-of formula recalculated each business day, not a permanent high-water mark: the preference reflects whichever of the three inputs is highest on the relevant date, so it can fall back toward the floor if quoted prices decline, but it cannot fall below the stated amount. Two consequences follow, and they point in opposite directions:

While a series trades below its stated amount, the floor governs and notional is the correct measure. That is the current position for the series where price has been disclosed.

If a series trades above the stated amount, the claim measured on that date exceeds notional. Management's stated objective of returning STRC to sustained trading at or near $100 sits precisely at the threshold where this term becomes operative on $10.5 billion of notional, so the Company's own success condition is also the point at which the measured claim can begin to exceed the figure used in this article.

The balance sheet already shows the mechanism in miniature: aggregate liquidation preference of $15.462 billion against $15.430 billion of notional, a $32 million excess reflecting series trading above the floor.

The preferred tax treatment depends on the Company's earnings and profits position

Preferred distributions have been treated as a non-taxable return of capital, which materially improves the after-tax yield to holders. The reason is disclosed, and it is not a structuring achievement: the Company states that it does not have any accumulated earnings and profits and does not expect to generate current earnings and profits in the current year or the foreseeable future, so distributions are generally not expected to qualify as dividends for U.S. federal income tax purposes.

The distinction matters: this is a tax characterisation turning on earnings and profits as determined for U.S. federal income tax purposes, not a mechanical consequence of GAAP losses. E&P is computed separately from book income and the two can diverge.

INTERPRETATION. The tax efficiency the preferred bid depends on is tied to the Company having no earnings and profits to distribute. A sustained E&P position would begin converting those distributions into taxable dividends.

The offering documents disclose a further risk under the fast-pay stock regulations: if dividends on a preferred instrument are economically a return of the holder's investment rather than a return on it, the stock may be characterised as fast-pay stock — treated as a listed transaction carrying annual reporting obligations and potential penalties for issuer and holders alike. The Company states it does not intend to issue such stock and has obtained advice of counsel, while acknowledging that the rules are unclear in certain respects and that the IRS could disagree.

Missing dividends transfers board seats

One term omitted from essentially all coverage. For STRK and STRF, if less than the full accumulated and unpaid regular dividends are declared and paid in respect of four or more consecutive dividend payment dates, the authorised number of directors automatically increases by one and holders elect a director; at eight or more consecutive, the same again.

Missing preferred dividends does not merely cost money and reputation. It transfers board representation. For a company whose strategy is inseparable from the conviction of its founder-chairman, that is a categorically different consequence than a deferred coupon at an ordinary issuer.

A convertible note has two possible economic identities. At low share prices it behaves as debt: a fixed claim, senior to equity, repayable in cash. At sufficiently high share prices the conversion feature dominates and the obligation can extinguish into equity. Which identity applies is not a modelling choice — it is determined by the share price against the conversion price.

In the present regime the answer is unambiguous. Every tranche is out of the money, the nearest by 53.9%, so the conversion feature contributes little and the instruments function as what they now are: zero and near-zero coupon senior unsecured paper.

The Company's own fair value measurements provide an accounting-date indication of how the notes were valued at quarter end. These are fair value estimates prepared for financial reporting, not verified exchange quotations, and the distinction matters for the inferences drawn from them:

Fair values are the Company's measurements as of June 30, 2026. In aggregate they imply a value of 93.2% of principal. This matters for the put options discussed in Section XII: where an instrument is carried below par, a holder exercising a contractual right to be repaid at 100 would receive more than the reported fair value.

This article therefore treats the downside case without any assumption of conversion, and notes that the upside case requires a 53.9% rally before the nearest tranche even begins to matter. That separation is deliberate — modelling conversion in a downside scenario, or ignoring it in an extreme upside scenario, produces internally contradictory results.

Only now, with the assets established and the claims enumerated, can a measure of common equity value be constructed.

This article uses an explicitly defined analytical term. It is deliberately narrow, and its name should not be read as a full net asset value:

Common NAV = Assets − Debt − Preferred − Other senior claims

Common NAV per share = Common NAV ÷ Common shares

This is an analytical valuation measure. It is not a GAAP balance-sheet figure, and it is not Strategy's proprietary bitcoin NAV metric — a distinction the Company itself insists on, describing its own outputs as not traditional NAV or liquidation value.

Two specification decisions have to be made explicit, because each moves the answer materially.

Which preferred measure? This article uses the liquidation preference, for the reasons set out in Section IV. Using the mezzanine carrying amount would overstate common equity by $1.02 billion.

Which share count? Basic shares outstanding were 384,225,751 as of July 24, 2026 — 364,585,501 class A and 19,640,250 class B. The Company uses an assumed fully diluted count of approximately 398.2 million. Given that every conversion price is at least 53.9% out of the money, the basic count is arguably the more honest denominator in the present regime. Both are shown below.

This excludes any value for the software business, discussed in Section XVI, and excludes other corporate assets. It is a bitcoin-and-reserve measure. Note that the reproduction of the Company's own formula in Section IX uses that formula's own inputs, including its stated out-of-the-money preferred notional, and therefore yields a slightly different per-share figure.

With Common NAV defined, the market's own metric can be introduced — and separated from it.

Three distinct concepts circulate under overlapping names:

The difficulty is that mNAV has no standardised definition, and the versions in circulation produce materially different answers from identical inputs.

Strategy's August 13, 2026 published formulation

Strategy publishes its arithmetic, which is to its credit. From the August 13, 2026 briefing, net reserve per share is computed as bitcoin holdings multiplied by the assumed bitcoin price, plus the USD Reserve, less out-of-the-money notional debt of $6.754 billion, less out-of-the-money notional preferred of $15.239 billion, divided by approximately 398.2 million fully diluted shares.

That yields a stated net reserve per share of $92.11 against a share price of $97.33, and a stated mNAV of 1.06x. Reproducing the formula independently from the filings gives $91.64 — a 0.5% variance explained by the Company using $15.239 billion of preferred notional against the $15.430 billion June 30 balance, reflecting STRC repurchases executed since.

Note what this means: in the August 13 briefing the headline mNAV is share price divided by net reserve per share. It is a per-share measure, not an enterprise value ratio.

That distinction requires care in attribution. Strategy has presented mNAV in more than one way, including formulations built on an enterprise value framework that treat debt and preferred as capital-structure components — which is also the form most common in market practice. This article uses the August 13 formulation when reproducing the Company's stated figure, for comparability with the number the Company currently publishes, and does not present it as the sole or definitive corporate definition. Readers comparing across sources should establish which formulation each is using before treating the outputs as equivalent.

Three definitions, one company, one day

Holding every input constant — the same bitcoin price, share price, share count and balance sheet:

Among the three EV/market-cap formulations A–C, the spread is 0.33x. Definition A places enterprise value slightly above bitcoin NAV, while B and C place it substantially below. Definition D is the Company’s per-share formulation and produces 1.06x. No input differs within A–C — only which capital-structure claims are subtracted.

Definitions A and B differ by exactly one item: whether the $15.43 billion of perpetual preferred is treated as a claim against the bitcoin or as capital alongside the common. That single choice moves the multiple by 0.29x, and it is the largest driver of dispersion between published figures. Definition C differs from both more broadly, since it also excludes debt and the reserve.

Why the treatment is genuinely contestable

Subtracting the preferred at notional treats it as a fixed claim repayable at face — the treatment appropriate to debt. But the preferred is perpetual. There is no date on which $15.43 billion must be found, and no collateral securing it.

Subtract it at notional and you charge the common with a repayment that will never contractually occur, which flatters the multiple by shrinking the residual against which the share price is measured.

Exclude it and you ignore the current $1.725 billion annualised preferred-dividend burden generated by perpetual preferred capital, even though individual dividends can be deferred or, for STRD, permanently skipped.

There is a third treatment that no published mNAV applies, and it is worth carrying out rather than merely naming. If the preferred is a perpetuity, value it as one: the annual dividend divided by a required return.

This is a mechanical sensitivity, not a valuation of the preferred securities: the underlying dividend streams are not fixed perpetuities. STRC's rate is adjustable at the issuer's discretion, STRD is non-cumulative, STRK can be settled in equity, and STRE escalates when unpaid. The exercise isolates one variable — the rate at which a level stream is capitalised — while holding those features aside.

The break-even is 11.18%, and that number is not a coincidence: it is the weighted average coupon of the stack. The identity is exact, because the annual dividend equals notional multiplied by the coupon, so dividing that dividend by the coupon returns the notional.

Subtracting the preferred at notional is arithmetically identical to valuing the dividend stream as a perpetuity discounted at its own coupon of 11.18%.

That reframes the choice usefully. Subtracting notional is not a neutral convention — it embeds the assumption that the market's required return on this capital equals the rate the Company is contractually paying. If investors require more than 11.18%, which preferred trading below stated amount would imply, notional overstates the economic claim. If they require less, it understates it.

Neither conventional treatment is obviously correct. What is incorrect is quoting a single mNAV without saying which one produced it.

Only now is it possible to answer the question that motivates ownership: why hold MSTR rather than bitcoin?

The mechanism is a flywheel, and it is genuinely powerful when it turns:

The critical asymmetry: the flywheel requires issuance above net asset value to operate. Below parity, the same mechanical process runs in reverse — issuing equity to buy bitcoin reduces bitcoin per share rather than increasing it.

Strategy published guidance keyed to precisely this in August 2025: below 1.0x mNAV it would consider issuing credit to repurchase MSTR; between 2.5x and 4.0x it would opportunistically issue MSTR to acquire bitcoin; above 4.0x, actively so.

Observed behaviour in 2026 matches the first regime, not the third. In the second quarter the Company repurchased $1.5 billion of its 2029 convertible notes at approximately an 8% discount, authorised $1.0 billion repurchase programmes for both STRC and MSTR common, and sold $218.4 million of bitcoin year to date — explicitly, per the filing, to help fund preferred dividends. A further 1,638 coins were sold in the single week to August 2, 2026.

The reverse cycle is not a scenario in this analysis. It is the documented 2026 behaviour of the company.

Three concepts travel under the word accretive, and they are not the same:

Bitcoin-per-share accretion — does the transaction increase bitcoin attributable to each common share?

Common-equity accretion — does it increase economic value attributable to each common share?

Liquidity preservation — does it preserve bitcoin and maintain the ability to meet obligations?

A transaction can satisfy the first and fail the second. Issuing preferred stock to buy bitcoin raises bitcoin per common share while adding a permanent senior claim that reduces the residual — accretive on one measure, dilutive on the other.

Strategy discloses this limitation itself, noting that its bitcoin yield and bitcoin gain metrics do not fully reflect debt, preferred stock and other senior claims. The Company also publishes both a gross and a net per-share bitcoin figure, and the gap between them quantifies exactly the point:

The same briefing discloses an amplification factor of 1.47x, noting explicitly that it amplifies downside as well as upside, and states that MSTR is a corporate equity rather than a bitcoin tracker, listing financing costs, preferred and debt costs, share issuance, credit spreads and changing mNAV among the reasons returns diverge from bitcoin.

An issuer voluntarily publishing its own downside leverage multiple represents a meaningful change in posture, and is more probative than any external characterisation.

The capital structure generates two distinct obligations running on separate clocks. Most analysis addresses one and misses the other.

Clock A — the perpetual dividend strip

A terminology note, because the label matters. The $1,759 million total below is the current annualised cash cost of the capital structure, not a fixed charge in the sense that term carries for debt. Note interest is contractually unavoidable while outstanding. Cumulative preferred dividends can be deferred but accumulate, and on STRF and STRE they compound. STRD is non-cumulative, so an omitted dividend is permanently forgone rather than deferred. This article therefore uses the aggregate as a measure of the cash the structure currently consumes, not of what must be paid under every circumstance or for every future year.

STRC alone accounts for 73% of preferred dividends, and 71.5% of the current annualised capital-structure cash cost once note interest is included. Operating cash flow of $9.85 million for the half-year annualises to roughly $19.7 million — about 1.1% of the current annualised cash cost. The preferred capital has no maturity, while note interest ends as the notes mature or are extinguished. That distinction is precisely why the duration of the preferred burden matters.

Clock B — the convertible put concentration

Every convertible tranche carries a holder put: a noncontingent right to require repurchase for cash at 100% of principal. Each date below has been corroborated against the completion announcement for the relevant offering. Mapped chronologically, the schedule is not what the maturity ladder suggests.

88.1% of the convertible principal is puttable within a single twelve-month window running from September 2027 to September 2028.

On a maturity basis the same debt appears comfortably staggered from 2028 to 2032. The put schedule collapses that four-year ladder into one year. Given the Company's reported fair values below par, assuming little or no exercise would require the analysis to assume that holders forgo a contractual right to be repaid at 100 on instruments carried below that level. INTERPRETATION. The Company's May 2026 repurchase of the 2029s at a discount demonstrates that management is actively managing the convertible stack well ahead of contractual maturity.

The two clocks collide

The reserve covers the dividends or the puts. It does not cover both, and the two overlap. This is a scenario rather than a forecast — holders may not exercise, and the Company may repurchase further notes at a discount or issue new instruments. But it frames the question the rest of the analysis answers: roughly $4.9 billion must be sourced between now and September 2028 from a business that does not generate it.

Where the money currently comes from

Three economically material sources. First, securities issuance: at-the-market raises in the second quarter of 2026 totalled $8.41 billion, comprising $2.95 billion of MSTR common and $5.47 billion of STRC. Second, the USD Reserve — itself built from issuance proceeds. Third, bitcoin sales.

FACT. Strategy sold $5.47 billion of STRC in the second quarter of 2026. FACT. The Company states that the USD Reserve is intended to support payment of dividends on its preferred stock and interest on its indebtedness, and discloses that issuance proceeds have been applied to build that reserve. FACT. In the week to August 16, 2026, $52.4 million of net proceeds from class A common sales were applied directly to STRC dividends and $132.2 million to STRC repurchases.

INTERPRETATION. The structure therefore carries a circular funding characteristic: the securities issued most heavily fund a reserve whose stated purpose is servicing those same securities — and in the most recent week, the path from equity issuance to preferred dividends is direct rather than inferred.

This is not a criticism of the mechanism's legality or disclosure, both of which are clear. It is a description of a structure whose stability depends on STRC continuing to clear near par — which, as Section VI established, requires a dividend rate the issuer sets, whose cash cost rises precisely when confidence falls.

This section answers three separate questions. They are often run together, which is why stress tests of this company tend to produce confident conclusions that contradict each other. Kept apart, they are:

If everything stopped today, is there enough to pay everyone? — a wind-up test, answered in 13.1.

What does it cost to keep the structure running? — a cash test, answered in 13.3.

If the money runs short, what actually happens, and to whom first? — answered in 13.4.

No probabilities are assigned to any bitcoin price path. None can be sourced to a primary document, and assigning weights would introduce the only unsupported assumption into an analysis where every other figure traces to a filing.

The first numeric column is the position today. The three columns to its right are explained in 13.2 and are the more important ones.

The convertible debt moves most sharply in proportional terms — from $2,502 to $9,165 after five years at $40,000, nearly four times higher — because the debt is the layer most protected by the cash reserve, and the reserve is the first thing to go.

The final column is a static illustration only: it holds the price constant, assumes no issuance, and ignores the cash reserve. It is shown to convey the shape of the problem, not to predict a date.

For the static illustration, the annualised dollar cost is held constant, so a cheaper coin means more coins sold. At $150,000 the structure costs 1.40% of the position a year. At $20,000 it costs 10.47% — seven and a half times as much. The burn accelerates precisely when the asset is weakest, and that acceleration is the mechanism driving the rising thresholds in 13.2.

At spot, servicing the capital structure costs 3.26% of the bitcoin position per year if funded entirely by sales. That figure should be held against any claim about long-term accumulation.

Two interactions deserve emphasis. Suspending STRE does not merely defer a payment: its terms require the Company to use commercially reasonable efforts over the following 60 days to raise proceeds through junior securities, including common stock — pointing toward dilution at precisely the moment common issuance is least attractive.

The STRK terms cut against suspension from the opposite direction. Because selling class A common through the ATM within 90 days of a record date compels payment of the STRK dividend, the Company cannot simultaneously run its equity programme and defer that dividend. The two levers it uses most — issue common, and stop paying — are partially mutually exclusive by contract.

Rows of numbers are less useful than the transitions between states. The stress test resolves into four regimes, distinguished not by price but by what the binding constraint becomes.

One clarification on Regime 4. It is sometimes framed as a test of whether the operating business and other assets can support the senior claims. On the figures in Section II, they cannot — the software business loses money and produced $9.85 million of operating cash flow in a half-year against $1.759 billion of annual charges. Regime 4 is a question about the saleability of bitcoin, not about the software business.

The interesting transition is between Regimes 2 and 3, and it is not marked by a bitcoin price. It is marked by the terms on which Strategy can issue.

The listed preferred series constitute a continuously priced market opinion on Strategy's capital structure, layer by layer. This is more informative than any single-point valuation exercise, because it is a spread structure: the difference in required return between series that differ only in rank and dividend mechanics isolates what the market thinks about subordination within this specific issuer.

Four questions the preferred market answers, and how to read each:

1. Discount or premium to stated amount

A series quoted below its $100 stated amount indicates that buyers require a higher return than the coupon provides at par. Because the liquidation preference is floored at the stated amount, a series quoted below that level carries a claim greater than its price, so a yield computed on price will exceed the yield computed on the claim. Any yield figure quoted should specify which denominator it uses.

2. The spread between series

The theoretically clean comparison is STRF against STRD. Both pay 10.00% on a $100 stated amount. STRF is senior to all other equity and cumulative with compounding; STRD is junior to everything except common and non-cumulative. Identical coupon, opposite ends of the waterfall. The spread between them is the cleanest within-issuer observable measure of the compensation required for additional subordination, though it also incorporates differences in dividend mechanics, compounding, governance rights, liquidity and investor base.

3. STRC against par

STRC is arguably the most informative preferred-market price for the company, and the reason is structural rather than analytical. STRC is 68% of the preferred notional and 72% of the annualised capital-structure cash cost. Its rate is set by the issuer rather than by a reference rate, and its ability to clear near par determines whether the largest funding channel remains open.

INTERPRETATION. In the framework developed here, the STRC price is arguably the most informative real-time indicator of whether the largest preferred funding channel remains viable — more so than the bitcoin price. This article does not state a STRC price series, so the claim is offered as an analytical conclusion rather than as an observation.

The Company has disclosed the objective directly: management does not intend to recommend a change from the 12.00% rate until STRC demonstrates sustained trading at or near its $100 stated amount. That converts an abstract question into an observable test.

4. The convertible market alongside it

Reading the preferred market alone is incomplete. The Company's reported June 30 fair values imply 93.2% of principal in aggregate for the convertible stack, with the 2029 notes at 86.5% and the 2030B notes at 88.8%. Debt carried below par ahead of a put concentration, alongside the preferred market's relationship to stated amount, would be the same indication observed at two levels of the structure. Reading them together gives a capital-structure view rather than an instrument view — though note that the debt figures are accounting measurements and the preferred figures would be market quotations, which are not the same kind of observation.

The common stock can now be compared against everything established above:

The framing question is usually: what is the market paying a premium for? In the current regime that question needs inverting. On two of three definitions the market is not paying a premium at all. Among the EV/market-cap formulations A–C, the spread is 0.33x: definition A places enterprise value slightly above bitcoin NAV, while B and C place it substantially below. The Company's own per-share formulation, D, produces 1.06x and therefore a modest equity premium. The full range across all four definitions is 0.35x. Measured against the bitcoin alone, MSTR trades at a 29% discount. Measured against Common NAV — bitcoin and reserve less all senior claims — it trades at a 5% premium.

That pair of numbers is the most compact statement of what MSTR is. The market is not pricing MSTR as a simple bitcoin proxy; it is pricing the residual equity claim created by the capital structure, and pricing it at roughly parity.

Components that could justify a premium to Common NAV, and what can be said about each:

Future bitcoin accumulation — requires issuance above NAV to be accretive. Currently near or below parity, so this component is weak by the Company's own published framework.

Capital-market access — demonstrably real: $8.41 billion raised in a single quarter. Also the component most at risk if STRC moves away from par.

The software business — annualising second-quarter revenue gives approximately $489.5 million. Because the business is currently loss-making, a revenue multiple provides one simple way to size it; other approaches would apply equally. At two to five times revenue the implied value is $0.98 billion to $2.45 billion, between 2.6% and 6.4% of market capitalisation. At roughly 3.5 times, the entire software business is equivalent to roughly one year of preferred dividends.

Volatility and convexity — the Company discloses realised volatility and options open interest, so this can be evidenced rather than asserted. It is a genuine component: a levered, optionable bitcoin proxy is not replicable by holding spot.

Embedded optionality on the capital structure itself — if bitcoin rallies far enough, conversion could become live and up to $6.71 billion of convertible principal could ultimately convert into equity if the relevant conditions are met on each tranche. That requires a 53.9% move on the nearest and considerably more on the rest, and the tranches convert at different prices on different dates rather than as a single event.

These are not one risk. They occur at different points, in a specific order, and the sequence matters more than the list. Ordered by the transmission mechanism rather than by severity:

INTERPRETATION. The bitcoin price does not appear as a separate line because it acts at every step rather than at one. A decline makes each transition faster and each lever more expensive. The ordering above is an analytical claim about where deterioration becomes observable first — not a claim that bitcoin cannot initiate the sequence, which it plainly can, through the path from bitcoin to the share price to financing capacity.

Note also that item 10 — the one most coverage treats as the central risk — is both the most remote and the least actionable, because no counterparty possesses a contractual right to force it.

At what point does Strategy stop being a bitcoin accumulation machine and become a balance-sheet management problem?

That question is more useful than asking what happens to MSTR if bitcoin falls, because it captures the interaction between the asset, the capital markets and the capital structure rather than treating the first as the only variable.

INTERPRETATION. On the 2026 evidence, this article argues the transition has already occurred, and can be dated from filings rather than forecast.

A company in accumulation mode issues securities and buys bitcoin. In the first half of 2026 Strategy did buy $13.67 billion of bitcoin. But in the same period it also repurchased $1.5 billion of its own convertible notes at a discount, authorised $1.0 billion of repurchase programmes across STRC and MSTR common, and sold $218.4 million of bitcoin to help fund preferred dividends — with sales continuing into August.

Repurchasing your own debt at a discount, buying back your own preferred, and selling the reserve asset to pay dividends are not accumulation activities. They are balance-sheet management. The Company's published guidance describes exactly this behaviour as the below-parity regime.

Strategy is no longer only a bitcoin accumulation machine. It has become, simultaneously, a liability-servicing vehicle — and the second function now constrains the first.

What is structurally attractive

The absence of triggers is a genuine and underappreciated strength. No margin call, no borrowing-base test, no collateral on either the notes or the preferred, and no maturity on $15.43 billion of the capital structure. Strategy cannot be forced to liquidate by a price decline. That is a materially better structure than the leveraged bitcoin vehicles it is often compared to.

The convertible stack carries an unusually low blended cash coupon of roughly 0.5% on $6.71 billion. Management has demonstrated willingness to retire liabilities opportunistically, repurchasing the 2029s at an 8% discount for a $113.9 million gain. And the capital-markets franchise is real: $8.41 billion raised in a single quarter is not a capability most issuers possess.

What is structurally dangerous

The preferred capital is perpetual, while the funding access required to service the structure is not. An annualised capital-structure cash cost of $1.759 billion sits against $9.85 million of half-year operating cash flow, with the funding gap met from issuance proceeds that in the most recent week flowed directly from common stock sales to preferred dividends. That loop works while STRC clears near par; if market demand weakens, maintaining the same market level may require a higher issuer-set dividend rate, increasing the funding burden precisely when access to capital is under pressure.

The put concentration compounds it. 88.1% of convertible principal becomes exercisable in a twelve-month window that opens in September 2027, inside a reserve runway of 2.64 years, with the Company's own reported fair values already below par.

And the asset is underwater. A 14.7% deficit against cost basis means there are no embedded gains to monetise without crystallising a loss.

What the market appears to be pricing

A 29% discount to the bitcoin, and approximate parity to the residual after senior claims. That is not a market pricing a bitcoin proxy. It is a market pricing a leveraged residual claim, and pricing it at roughly what the arithmetic says it is worth.

The distinction that is the subject of this article

Method, and What Could Not Be Verified

Every figure in this article traces to a primary filing, identified in the footnotes. Financial data is drawn from the Form 10-Q for the quarter ended June 30, 2026 and from Current Reports on Form 8-K filed subsequently. Instrument terms are drawn from certificates of designations, offering documents and the FY2025 Form 10-K. Secondary commentary has not been used as a source for any figure.

Model outputs were constructed in a spreadsheet and independently recomputed in Python before being written into the text. Where the Company publishes its own version of a metric, the calculation here was reconciled against it and the variance explained.

Subsequent events

This article presents a consistent snapshot: financial data as of June 30, 2026, bitcoin holdings and the USD Reserve as of August 9, 2026, and market prices as of August 13, 2026. A Current Report on Form 8-K filed August 17, 2026 discloses more recent figures, which are noted here rather than blended into the analysis, because coverage ratios struck on mixed dates are not comparable.

As of August 16, 2026: bitcoin holdings were unchanged at 840,447, with no purchases or sales in the week; the aggregate purchase price and average cost were unchanged at $63.36 billion and $75,385. The USD Reserve had risen from $4.65 billion to $4.80 billion. Strategy repurchased 1,388,720 STRC shares for $132.2 million, leaving $653.0 million of the Digital Credit Securities Repurchase Program available, alongside an untouched $1.0 billion MSTR repurchase authorisation.

The composition of that week is worth noting on its own. Of $333.7 million of net MSTR common ATM proceeds, $52.4 million funded STRC dividends, $132.2 million funded STRC repurchases and $149.1 million increased the USD Reserve. Common equity was issued to pay preferred dividends, buy back preferred stock, and build the reserve that pays preferred dividends — the circularity described in Section XII, visible in a single week of filings.

On the updated reserve, coverage would be 2.73 years rather than 2.64, and the funding gap through September 2028 would be $4.78 billion rather than $4.93 billion. Neither changes any conclusion in this article.

Stated limitations

No rating is issued and no probability is assigned to any bitcoin price path. The stress test in Section XIII is a threshold model, not a probability-weighted forecast.

Claims in the coverage table are stated at notional. Because the liquidation preference ratchets above the stated amount when a series trades above it, those thresholds understate claims in any scenario where a preferred series trades above par.

The tax position is only partly analysed. The basis of the return-of-capital treatment is established in Section VI, but the $1.93 billion deferred tax liability that reversed during the first half of 2026 is not examined and no view is offered on valuation allowances.

No market-impact assumption is applied to the bitcoin sale scenarios in Section XIII. Selling tens of thousands of coins annually is modelled arithmetically, not with an execution cost.

Fundamental-change provisions in both the convertibles and STRD are noted but not modelled. A change of control would accelerate obligations in ways this analysis does not capture.

Market prices for the listed preferred series are not stated, because no dated quote source was verified. Section XV is framed as mechanism rather than as observed levels.

All six convertible put dates have been corroborated against the completion announcement for each offering, but have not been read against the underlying indentures themselves.

The euro-dollar rate applied to STRE is approximate.

Statements attributed to management from the July 30, 2026 earnings call rest on secondary reporting; the transcript was not obtained.

The August 2025 issuance guidance is described as published in August 2025; whether it remains operative policy was not independently confirmed.

Market data throughout is as of August 13, 2026 and financial data as of June 30, 2026. This capital structure changes weekly: STRC is repurchased weekly, bitcoin is bought and sold weekly, and the STRC dividend rate is reset at the issuer's discretion. Readers should refresh against current filings before acting on any figure here.

This article is analysis, not investment advice. The author is not a registered investment adviser. Readers should conduct their own research and consult appropriate professionals before making investment decisions.