Hook
Strategy bought 1,665 bitcoin last week.
That was the headline. The number most desks actually pulled was 847,666 — total holdings, $63.9 billion cost basis, $75,437 average entry. Twenty-four consecutive quarters of net accumulation, by the company's own counting convention.
Then the number almost nobody quoted. BTC Yield, year to date: -3.7%. In Strategy's own framework, that is a BTC Gain of -25,085 coins. Roughly $2.1 billion of per-share bitcoin content, deleted.
Both figures are correct. They are not in conflict. One measures the size of the vault. The other measures the size of each share's claim on that vault. When you fund coin purchases with newly issued equity, the vault grows and the claim thins — and if issuance runs faster than purchases, the claim thins faster than the vault grows. Silence is the most expensive asset in a bubble. This particular silence was sitting inside a press release that most readers skimmed in nine seconds.
Context
Strategy is not a protocol. No consensus layer, no audited contract, nothing to re-run on a testnet. What can be examined is a capital structure — and it has become one of the more elaborate ones in public markets.
The machinery is a stack. Common equity is sold through an at-the-market program with $18.84 billion of unused capacity; last week that channel moved 1.47 million shares for $246.2 million. Beneath it sit four preferred classes — STRC, STRF, STRK, STRD — with STRC the load-bearing one: 12% annualized dividend, designed to trade around a $100 par, marketed to income buyers rather than crypto natives. Behind those obligations sits a $5 billion USD reserve plus roughly $1 billion of general cash.
On October 28 a special shareholder meeting votes on compressing preferred dividends from twice-monthly to daily, with a November 1 record date and a November 2 first payment. It reads like a shareholder-friendly upgrade. It is more usefully read as an engineering change to the duration of a liability.
I spent 2017 parsing Geth node logs during the Parity freeze, hunting a 0.04% gas-fee discrepancy for high-volume traders. The lesson that stayed with me was structural: the interesting anomaly is never in the marketing layer. It is three layers down, in the arithmetic nobody is incentivized to surface.
Core
Start with the arithmetic that did get reported, because it is clean.
Strategy's latest purchase: 1,665 bitcoin at $85,681, or $142.7 million. Holdings: 847,666 bitcoin. Cost basis $63.9 billion, implying $75,437 average. The implicit mark on the treasury, $70.7 billion, implies a current bitcoin price near $83,400. That places the newest tranche roughly 2.7% underwater at the time of writing, against a 10.6% unrealized gain on the whole position.
Now the arithmetic that did not.
Analysts tracking the company modeled $250-300 million of weekly purchases. Strategy deployed $142.7 million. Roughly half. Year to date the company has still absorbed 175,166 bitcoin, so the machine is running. It is just running slower than the street expected.
Here is where the allocation flipped. Over the same window, Strategy repurchased $151.7 million of STRC.
Spend on defending the preferred: $151.7 million. Spend on acquiring bitcoin: $142.7 million.
That is the first time in this cycle the two lines crossed in that direction, at least inside the disclosure set I can reconcile. A treasury company that buys more of its own funding instrument than it buys of its treasury asset has changed its objective function. The stated strategy is still accumulate bitcoin. The revealed strategy is keep the financing channel open.
The reveal is legible in one more figure: Strategy's repurchases accounted for 18.9% of STRC's traded volume, down from 22.8% the prior week. Nearly a fifth of the float's daily turnover was the issuer bidding for its own paper.
I have seen this signature before. In 2021 I clustered wallets behind a profile-picture project and found 60% of the community routed through three addresses running wash trades. The prices looked organic. They were not. I wrote the report, kept it, and said nothing publicly — a decision I would make differently now.
Then the flywheel math underneath all of it. YTD BTC Yield -3.7%. Quarterly BTC Gain of -100,275 coins, about -$8.37 billion. Twenty-four consecutive quarters of net accumulation is an absolute-holdings statement. Negative BTC Yield is a per-share statement. Both can be true while the second one quietly eats the first.
Contrarian
The popular read is that STRC recovered. It traded up from a $71 low toward $99, within a percentage point of par. Stabilization achieved, narrative intact.
Correlation is not causation, and here the causation is documented rather than inferred. An issuer that is 18.9% of the volume in its own instrument has not been validated by the market. It has been underwritten by itself. The honest test is not whether STRC holds $99 while the bid is running. It is whether STRC holds $99 in the week after the bid slows. Yield is often the interest paid on risk you didn't underwrite, and 12% annualized on a perpetual preferred issued by a company whose only asset is a volatile one is not a generous coupon. It is a price.
The daily-dividend proposal deserves the same skepticism. It is framed as an upgrade for holders. Mechanically it shortens the accrual interval, compresses duration, and reduces the discount pressure that forces the issuer to defend par with cash. That is cost optimization wearing a growth costume.
One correction to the bear case, in fairness: Strategy is not a Ponzi. It holds 847,666 real coins. What it is, precisely, is a reflexivity vehicle. Accretion depends on selling paper above net asset value and converting proceeds into an asset that keeps appreciating. Remove the premium and the loop stops being accretive. It does not stop being solvent.
Worth flagging as well: one date in the circulating data set reads 2026, and the 847,666 figure runs high against numbers I can independently reconcile. I trust the code, not the community. When I cannot match a figure to an 8-K, I carry it with a confidence flag, not a conclusion.
Takeaway
Three signals to watch, and none of them is the bitcoin count. First, the implied mNAV premium: if the implied $167.5 per share is accurate, the premium that makes ATM issuance accretive has compressed hard. Second, whether the streak extends to a third consecutive week — half-sized purchases plus a defensive rotation is a momentum question, not a conviction question. Third, STRC volume net of the issuer: if the repurchase share keeps falling and par holds, the market is genuinely there. If par only holds while the bid runs, the stabilization is a purchase, not a recovery.
