The exception became the engine: Strategy's dividend carve-out is now its funding model
A rule written for the defensive case now runs the capital structure — and the reserve it feeds is, on same-date arithmetic, very nearly the company's entire liquid balance sheet.
In July 2025, Strategy published a rule about when it would not sell its own stock: below 2.5x mNAV it would not issue common equity "except to (1) pay interest on debt obligations and (2) fund preferred equity dividends". Twelve months later, the company's weekly Form 8-K telemetry shows that exception operating as the principal engine of the capital structure. In the single disclosed week of 2026-07-20 to 2026-07-26, Strategy sold 5,429,160 MSTR common shares for $544.5M net, moved $525M of those proceeds into its USD Reserve, and spent $25.0M buying back its own 12%-rate STRC preferred below par — while reporting zero repurchases of every other security and zero preferred ATM sales.
The headline the issuer offers — a record $3.75B USD Reserve covering roughly 25 months of preferred dividends and debt interest — is real cash. But same-date arithmetic on the Q2 balance sheet shows the policy-defined reserve is not a buffer on top of the company's liquidity; it substantially is the company's liquidity: the $2.4B quarter-end reserve compares to $1.71B of GAAP cash plus $736.1M of short-term investments — $2.45B in total (derived from disclosed figures; not a forecast). The reserve is built, and continuously rebuilt, from the same market-access channel — common-equity issuance, backstopped by bitcoin sales — whose interruption it exists to insure against. That circularity, not the coverage number, is what a credit desk should be modelling.
- The 2025 policy's exception lane is now the operating mode. The ANNOUNCED 2025-07-31 framework barred common issuance below 2.5x mNAV except to pay debt interest and preferred dividends. The ANNOUNCED 2026-06-29 statement instead described disciplined issuance "at or near 1x mNAV per Share" and did not restate the bands; no disclosure reconciles the two, and their formal relationship is unresolved. Observed weekly behavior — common ATM sales of $544.5M with $525M routed to the dividend reserve — matches the exception lane, not the accumulation lanes.
- Bitcoin has become a disclosed funding source for dividends, in reconcilable size. Strategy expressly attributed approximately $218.4M of 2026 year-to-date BTC sales to funding preferred dividends. The three disclosed sale tranches — 32 BTC for $2.5M, then 1,363 BTC for $80.8M and 2,225 BTC for $135.2M — sum to 3,620 BTC and $218.5M (derived), reconciling with the issuer's attribution to within $0.1M. That equals roughly 12% of one year's $1.762B dividend-and-interest run-rate (derived).
- The holdings bridge closes exactly, and it closes flat-to-down. 847,363 BTC after the 2026-06-22 purchase, minus 1,363 sold by 06-30 equals the 846,000 BTC the issuer reported at quarter-end, minus 2,225 more equals the 843,775 held on 07-05 and still on 07-19 (derivations shown in Analysis). The last disclosed purchase is $35M on 2026-06-22; the July windows show none.
- The reserve grew ~$2.9B in nine weeks — almost entirely from equity issuance. $871M (05-25) → $900M → $1.0B → $2.55B (06-28) → $3.225B (07-19, a figure that included expected proceeds from unsettled ATM sales) → $3.75B (07-26). Part of the insurance policy is a receivable from the channel being insured.
- The two 2026 liability buybacks solved different problems, in very different sizes. In May, Strategy EXECUTED a $1.38B cash repurchase of $1.50B principal of its 0% converts due 2029 — an 8% discount that removed a maturity wall but zero annual cash coupon. In July, it EXECUTED a $25.0M repurchase of STRC — the layer costing 12% per annum — retiring $28.9M of stated amount and about $3.5M of annual dividends, a 13.9% cash-on-cash reduction in fixed charges (all derived from; not a forecast). The ratio of cash deployed was 55:1 in favour of the zero-coupon instrument.
- One instrument's monthly rate decision governs ~71% of the entire fixed-charge base. STRC's $10.4895B notional at its July 12.00% rate implies $1.26B of the $1.762B annual burden (derived); each percentage point of rate is ~$105M per year (derived). Management's CURRENT POLICY INTENT of 2026-07-27: retain the existing rate until STRC trades near $100, repurchase below par, and issue none below $100 — which, with STRC in the $70s–$80s, leaves the $17.5B STRC ATM shelf (36% of all remaining authorized capacity, derived from) inert by policy and concentrates live funding on the MSTR common ATM.
- The "never missed" record was mostly earned at a much lighter load. $1.06B of cumulative preferred dividends paid over 18 consecutive months averages ~$59M/month; the current run-rate is ~$147M/month (derived) — roughly 2.5x the historical average pace. The track record is genuine; it is also short and mostly pre-dates today's burden.
Analytical practice for credit, treasury and risk desks; not investment advice:
- Read the weekly 8-K as a three-line funding-flow statement, not as holdings news: (1) common ATM net proceeds, (2) reserve delta, (3) repurchases by series. Flag any week where stated uses exceed disclosed common proceeds — in the 07-20 week, $525M + $25.0M = $550M against $544.5M of net proceeds, implying a small draw on other cash (derived).
- Decompose every reserve print into settled cash versus expected/unsettled ATM proceeds, which the 07-19 figure explicitly included, and reconcile the policy reserve quarterly against GAAP cash plus short-term investments. Coverage-months figures are stock measures conditional on continued market access; model them that way.
- Do not roll the encumbrance anchor forward. The last same-date pledged/unencumbered split is 762,099 BTC / 0 pledged as of 2026-03-31. Later holdings updates are gross-only; the Q2 exhibit's senior-claims language, the "not collateralized by bitcoin" caveat, and the non-ring-fenced reserve mean seniority, not security, is the operative protection.
- Track the STRC monthly rate decision as the single largest fixed-charge variable (~$105M/yr per point, derived) and the below-par repurchase pace against the $975.0M remaining authorization.
- Maintain a dual claims ledger. Strategy's updated mNAV (effective 2026-07-23, prior series not comparable) deducts preferred at notional excluding accrued and unpaid dividends; the third-party CEBE framework deducts at accrued liquidation preference. Today the wedge is small because dividends are current; under any deferral the issuer-defined metric would mechanically understate senior claims relative to the accrued basis. Compute both, every period.
- Map cross-holdings before treating treasury companies as independent credits. Strive's own 8-K showed $50.1M of Strategy STRC held as an asset; one issuer's liability layer is another's balance sheet. Watch the cost-relief arithmetic too: the purchases-page aggregate cost fell $413M on 3,588 coins sold — ~$115,100/BTC against a $75,651 portfolio average (derived) — implying lot selection worth confirming in the Q2 10-Q.
- If the weekly filing cadence continues in its present form, it remains the primary telemetry for this credit, and the decisive variables it carries are funding flows — equity proceeds, reserve movements, repurchases — rather than coin counts. A change in filing frequency or content is the trigger to re-plan.
- If the preferred dividend rate remains at its stated elevated intent while the instrument continues to trade below its stated amount, the fixed-charge base stays elevated and the preferred shelf stays dormant as a funding source. A rate reduction, or the instrument reaching its stated amount, is the trigger that reopens that lane.
- If bitcoin sales continue to be attributed by the issuer to dividend service, the accumulation-era framing does not describe the marginal flow. Successive filings showing dividend service without bitcoin sales or common issuance are the trigger to revert to that framing.
- If issuer-defined metrics continue to be redefined without restatement of prior values, internally consistent time series must be maintained on the desk's own definitions rather than the issuer's. A restatement, or a published reconciliation between old and new definitions, is the trigger to re-adopt the issuer's series.
- If cross-holdings between treasury issuers continue to deepen, single-name analysis understates correlated stress. Disclosure of one issuer holding another's preferred is the trigger to re-map sector exposure.
The complete 13-page note: all seven analysis sections, the 23-item dated evidence register, and the full source notes with links to every filing cited.
- Analysis — From accumulation rulebook to servicing rulebook
- Analysis — The bridge closes: coins, cash and the reserve
- Analysis — The reserve is the liquid balance sheet
- Analysis — Two buybacks, two different problems
- Analysis — Coverage arithmetic at its strongest — both ways
- Analysis — Definitional drift: the denominator moved mid-series
- Analysis — The preferred layer is a correlated one
- Evidence — 23 dated, source-traced facts
- Acronyms and full sourcing