The Solvency Map Is Not the Credit

Strategy published an interactive chart of the bitcoin prices at which each of its credit instruments goes undercollateralized. We reproduced its floors to the dollar, reverse-engineered its formulas — and found the market charging up to sixty-three times the model's spreads. Both are measuring the same balance sheet. Only one of them is pricing the risk that matters.

TON618 Capital Research · MSTR $96.09 · BTC ~$63,467 · closes 11 Aug 2026 · Download PDF · June 2026 valuation · Q2 addendum

A flywheel at 1.0× needs a new story

The context is the story. MSTR trades at roughly 1.0× our estimate of common net asset value — $37.8B of market capitalization against ~$36B of bitcoin and cash net of senior claims, before a modest software value. The accretive-issuance premium that powered the 2024–25 flywheel is gone, and with common-at-a-premium unavailable as the marginal funding source, the capital machine has rotated to credit: five preferred series totaling $15.2B of notional, led by the $10.3B variable-rate STRC. Strategy has sold bitcoin five times in 2026 — most recently 1,690 BTC for ~$108.6M on 10 August — and directed the latest proceeds, for the first time, to repurchasing its own discounted preferreds rather than adding coins.

Credit buyers need a story about why these instruments are money-good. The BTC Credit model, unveiled by Michael Saylor on X last week and live at strategy.com/credit, is that story, quantified: an interactive table with three user-adjustable inputs — bitcoin price ($63,467 at our read), volatility (40% default), and a bitcoin expected-return assumption ("BTC ARR," 10% default) — that assigns each instrument a coverage multiple, a probability of undercollateralization, a model credit spread color-coded investment grade / high yield / distressed, and the headline number: the BTC Floor, the bitcoin price at which that instrument's claim stack stops being covered.

The soundings: what the model says

Dropped onto one gauge, the model reads as a mariner's chart. The waterline is spot. The preferred floors cluster at $16,000–$20,600 — 68–75% below it. The convertible floors are abyssal: $369–$2,456, requiring a 96–99% drawdown held to maturity before the debt is uncovered. The headline stats above the table: a $53.34B bitcoin reserve, a $4.65B USD reserve, a "BTC duration" of 30.7 years, and a "breakeven ARR" of 3.25% — the annual bitcoin return at which everything is covered at terminal.

Coverage declines monotonically down the waterfall because each instrument is tested against the cumulative claims senior to and including it. Market spreads: current yield on stated rates (STRF 10.00%, STRK 8.00%, STRD 10.00% per $100) minus the 10-year Treasury at 4.70%. STRC's rate resets monthly; we quote its price against its $100 par instead.

The math checks out — we ran it

Strategy publishes no methodology page, but the mechanics can be recovered from the numbers, and three of them reconcile against the live dashboard:

Two loose ends matter for a product marketed as transparency. The near-dated convertible rows carry an undocumented adjustment — Convert 2028 shows $310M of cumulative notional against a $1,010M note, which does not reconcile under the formula that fits every other row and appears to phase the USD-reserve credit in over time. And the "BTC Floor ARR" column for the preferreds implies horizons of roughly 10–15 years rather than the displayed 7–10 year durations. Neither changes the picture; both mean the methodology should be published.

Credit where due. The floors are honest, independently checkable arithmetic; the seniority ordering matches both the documented capital structure and the yield ordering the market charges; and publishing any of this, with adjustable assumptions, is a transparency step almost no issuer takes. The 40% volatility default is not even a cherry-pick against current conditions — we measure realized bitcoin volatility at 28% (30-day) and 34% (90-day). The problem is not the defaults' honesty. It is what the machine built on them leaves out.

What the market says back

Put the model's spreads next to the spreads the market actually charges — same instruments, same day — and the model's tier legend grades its own output:

By the model's legend, STRF, STRC, STRE and STRK print investment grade and STRD barely clears into high yield. The market prices STRF at ~569 bps, STRK at ~700, STRD at ~997 — and will not hold STRC, engineered to sit at $100, above $95.32. A gap this wide is not a liquidity premium. It is the market pricing what the model omits: path risk, dividend sustainability, subordination beneath a growing senior stack, and cross-default — which the site's own disclaimer concedes it ignores, along with liquidation preferences that can exceed notional.

Where the model breaks

It prices terminal solvency, not the way credits die. The breach test is European — does bitcoin sit below the floor on the horizon date. There is no path dependence, and first-passage probability is always higher than terminal probability. The instruments that matter near-term, the 2028–2030 converts, face refinancing walls, not collateral shortfalls: a 2027 drawdown that never touches any floor can still close the capital markets, stall the ATMs, and force coin sales. That is not a hypothetical — it is a description of 2026 to date.

The flow problem sits entirely outside the model. Management's own figure puts fixed obligations — preferred dividends plus interest — at roughly $1.76B a year against a $4.65B USD reserve: about 2.6 years of runway with no new issuance, and the software business does not bridge it. The model nets the reserve as a static stock against terminal claims while the same dollars are consumed quarterly. Worse, the funding is reflexive: dividends are paid substantially by issuing more STRC — senior to STRE, STRK and STRD — and every $1B of STRC issued raises STRD's floor by roughly $790 even after crediting the bitcoin it buys. The coverage ratios are endogenous to the market's willingness to keep believing the model.

The spread outputs are two honest inputs away from distressed. Hold the model's own math and move only the dials:

The same exercise moves STRC — the $10.3B keystone — from 116 bps to roughly 375–910 bps. Bitcoin realized 60–80% volatility in the 2022 drawdown; a credit model whose rating tier flips from investment grade to distressed on a volatility regime the asset has entered repeatedly — 2018, March 2020, 2022 is a scenario tool, not a rating. The 10% drift does quiet work too: pricing credit off a favorable real-world drift rather than anything risk-neutral flatters every spread, and the headline "3.25% breakeven ARR" is a certainty-equivalent statistic that ignores sequence risk entirely.

And the denominator is no longer sacred. The floors assume the coin count. Strategy has sold bitcoin five times this year. Directed at buying back preferreds at 68–96 cents on the dollar, sales currently improve junior coverage — retiring $1.00 of claim for well under a dollar. The same lever, run to pay dividends in a deeper drawdown, shrinks the collateral the entire model rests on. Never sell your bitcoin is no longer an assumption anyone can grant — our Q2 addendum traced its formal retirement.

Verdict

As disclosure, the dashboard is genuinely good — best-in-class for the asset class, and the floors are numbers a reader can verify from the page itself. As credit analysis, it is not decision-grade. It answers will the bitcoin cover the stack at maturity under friendly assumptions — almost certainly yes, for the debt — while the live question the preferred market is pricing is can the dividend machine keep running through a bear market without eating the collateral. Converts: read them off this map   Preferreds: do not

Three tripwires

Sources & Method