Divide $432.5 million by 6,948 coins and the average realized price lands near $62,250 [12], a price rather than a valuation, and that fact alone is the entire content of the change. The 2020 opening purchase was $250 million [3]; the sales since May have raised about 1.73 times that [13].
The buying machine ran on a premium, and the premium was the mechanism that made the whole thing work, not a cosmetic feature of it. At 3.89 times net asset value in November 2024 [9], every share sold bought more bitcoin than it diluted, so issuance was accretive per share and the convertible notes stacked on top of it looked like Saylor's Manhattan developer borrowing against a rising appraisal [4]. Below one times NAV, and Decrypt reports mNAV fell below 1 as bitcoin dropped in 2026 [9], the same trade runs backwards: selling stock to buy coin leaves each existing share with less bitcoin behind it. The premium compressed by at least 74% along the way [14]. Sherwood Media had flagged the arithmetic in the same month the multiple peaked, with MSTR worth 3x the bitcoin it held and forced liquidation as the tail [10].
The proximate cause, in Decrypt's account, was STRC falling below its $100 par in June 2026, which analysts it cites called a structural crack, on the logic that the longer the preferred traded below par the likelier Strategy would have to choose between issuing new shares and selling bitcoin to keep the dividend current; broadly, that is what happened, and the dive closed a route the company had been using to buy [8].
Which makes the two-account split the interesting term rather than the sale itself. A USD Reserve ring-fenced for dividends and interest converts an obligation that used to rest on continuous market access into a claim on a stated dollar balance, while an unrestricted cash account that may be spent on anything, bitcoin included, preserves the accumulation story for common holders [2]. Those are two audiences being answered with one pot of money, and the reserve dollars are earmarked for covering that near-term obligation now, leaving the $13 million target for compounding coins over 21 years to whatever sits outside the ring-fence [11].
This is probably wrong in one specific way, so name it: "never be a net seller" [1] is literally satisfiable if purchases over the same window exceeded 6,948 coins, in which case this is treasury plumbing and not a funding retreat. A second reading has the sale as a prudent liquidity buffer ahead of dividends across STRK, STRC, STRD, STRF and STRE [6], costing nothing structural. A third has the copycat balance sheets at MARA, Metaplanet and Riot Platforms [7] walking the same sequence with thinner capital markets access and worse execution.
The desk's view, held loosely: once mNAV sits below 1, the reserve is the cheapest capital the company owns, and a ring-fence is what you build when you expect to draw on it again. The thesis breaks if the next disclosure shows net purchases over the same period, or shows the USD Reserve funded out of issuance proceeds rather than coin sales. The source also leaves two things unspecified: the year the May sales began [1], and the dividend and interest run rate the reserve is meant to cover. The disclosure that settles it is the holdings count.
