Strategy raises capital and spends it on Bitcoin. But not every raised dollar ends up as a sat. Here's the rest.
The cash flow
reconciliation
For each fiscal year through 2025, the 10‑K's Consolidated Statement of Cash Flows tells you exactly how much was raised, how much bought BTC, how much paid coupons, debt, and operating costs, and how much settled into the treasury as cash. The page below tabulates that. The annual identity is beginning cash + operating + financing + investing + FX = ending cash — the SEC enforces it; we just cite it.
The line every dollar has to fit on
=
beginning cash
+ operating cash (software + interest + taxes − non-cash)
+ financing cash (ATMs, converts, preferreds − dividends − debt paydown)
+ investing cash (mostly − BTC purchases)
+ FX effect
It's an identity, not an opinion. Every number on every line of every 10‑K has to fit. Our job here is tabulation and citation — read the filings, transcribe the rows, point you at the source. The interesting part isn't the arithmetic; it's the decomposition: where the cash that didn't become BTC actually went.
Three years on one row each
All figures from the FY2025 10‑K Consolidated Statement of Cash Flows, in millions of USD. View the source filing →
| Period | Cash begin | Raises | BTC spent | Operating | Pref divs | Debt paydown | Cash end | Burn ratio |
|---|---|---|---|---|---|---|---|---|
| FY2023 | $50.9M | $1.89B | $1.90B | +$12.7M | — | — | $48.7M | operating + |
| FY2024 | $48.7M | $22.13B | $22.07B | −$53.0M | — | $160.0M | $39.9M | 0.24% |
| FY2025 | $39.9M | $24.84B | $22.47B | −$67.2M | $381.4M | $515.4M | $2.30B | 0.27% |
Burn ratio = |operating cash burn| ÷ raises. Answers "out of every dollar raised, how much got consumed by the non-BTC parts of the business?" The FY2025 ratio of 0.27% means 27 basis points — roughly a quarter of one percent. Operating cost in this business is a rounding error.
From closed loop to treasury buffer
FY2023
Pre-ATM-mania era: ATM common raised $2.03B; BTC spend $1.90B; ~$127M not used for BTC (covered interest, secured debt, and mild operating gain).
FY2024
Near-perfect closed-loop year: raised $22.13B (ATM $16.36B + converts $6.21B + other) and spent $22.07B on BTC; net cash drift −$8.8M for the whole year. Operating burn $53M covered by the $60M residual after BTC spend.
FY2025
Preferred issuance debut: $7.04B of STRF/STRC/STRE/STRK/STRD proceeds. First $381M of preferred dividends paid. Cash balance jumped from $40M to $2.30B — management now holds a real treasury buffer rather than matching inflows to BTC purchases.
The pattern through FY2024 was a near-perfect closed loop: raise dollars, spend dollars on BTC, sleep at night with a few tens of millions in cash. FY2025 broke the pattern — Strategy retained $2.38B of net cash and ended the year with a real treasury buffer for the first time. The next section is why.
Three years of outflows, kept in cash
Starting in 2026, management formally adopted a target of holding roughly 3× current annual non-BTC cash outflows in the treasury — a buffer that covers operating costs, interest, taxes, dividends, and debt principal for three years even if every capital-raising channel closed at once.
The buffer is a strategic shift, not an accounting accident. Through FY2024 Strategy ran with cash balances in the tens of millions — every raised dollar that didn't become BTC stayed in cash for days, not years. Starting in 2025, preferred-stock proceeds began arriving alongside ATM common, and the obligation profile got more complex (monthly STRC dividends, secured-debt amortization). Management chose to hold a real buffer rather than match every raise to the nearest BTC purchase. At 2.3 years of coverage today, the company could pay every coupon, every interest payment, every salary, and every debt principal for that long without raising a dollar — long enough to ride out a multi-year ATM shutdown if one ever happened.
Atomic tranches vs. actual cash flow
The 10‑K's Consolidated Statement of Cash Flows shows the physical flow of dollars. In FY2025, $7.04B came in as preferred stock proceeds and $381.4M went out as preferred dividends. The 10‑K does not show a "proceeds from sales of digital assets" line — Strategy did not sell any BTC in FY2025. So physically, preferred dividends were paid out of newly-raised cash, not out of BTC sold from the tranches that back those preferreds.
The site models it differently. On the tranche pages, every dollar of new raise proceeds is treated as buying BTC (the tranche stays whole), and each preferred tranche pays its own coupon by nominally selling BTC from itself at the period's market price. The resulting numbers are economically equivalent to the physical flow — the ending balance sheet matches — but the per-tranche story is legible: each tranche is a closed system with its own BTC balance, its own coupon stream, and its own forward asymptote. The BTC-cost page and preferreds page are built on that atomic framing.
The cash buffer is the seam where the two models meet. When aggregate post-model BTC basis and the actual 10‑K balance sheet diverge — because Strategy is carrying a real cash buffer — the difference is modeled as "unallocated cash," swept into BTC on the next purchase disclosure that exceeds the preceding quarter's raise total. This page is where the divergence gets disclosed; the per-tranche pages are where the atomic model lives.
What this page doesn't (yet) show
- Quarterly granularity. The table above is annual because the 10‑K is annual. To split into quarters we need each year's three 10‑Qs (Q1, Q2, Q3) plus the 10‑K residual for Q4. That ingestion is on the roadmap; it's not in the current build.
- Per-raise attribution to specific BTC tranches. The 10‑K aggregates ATM proceeds into a single line ($16.27B in FY2025). Strategy never tells you which week's ATM proceeds funded which week's BTC. The site models this as fungible within the quarter and attributes costs proportionally — which is a methodological choice, not a shortcoming, but worth flagging.
- Per-tranche raise-vs-BTC delta on every row. The tranche ledger links each row to its 8‑K and breaks out per-leg shares + USD where the filing discloses them. The full per-row reconciliation against the 10‑K cash flow statement is partial — about 30 of 107 rows have an explicit raise-vs-BTC delta computed. The rest tie out at the annual level (this page) but not at the row level. Closing that gap is incremental data-entry work.
Bottom line. The annual identity above is exact and tied to the 10‑K. The per-tranche atomic model is internally consistent and ties to the same balance sheet, by construction, even though it's a different lens on the same dollar flows. What the site currently doesn't surface week-by-week is the path the cash buffer takes between filings — and that requires the 10‑Q ingestion that's still on the roadmap.