Bitcoin miners cut reserves by more than a third to 1.19 million BTC

A fall from 1.9 million coins at the end of 2025 to about 1.19 million ^[1] is roughly 710,000 BTC, or 37% of the stockpile ^[14]. Cryptopolitan puts output at 450 BTC a day above cost ^[2]. At that rate the decline equals about 1,578 days of production, a little over four years, sold in about one ^[15]. At the $80,000 level the report treats as the cost line, those coins come to about $56.8 billion ^[17].

The pool balances in the same report account for very little of that. Binance Pool slipped from 42,000 BTC in March to 41,897, and AntPool added coins ^[5]. Foundry USA, which mines 25% of blocks ^[4], sold its last 236 BTC in September ^[6]. F2Pool went from about 6,000 to about 4,000 over two years ^[7]. Those itemised reductions total about 2,339 BTC, roughly 0.3% of the aggregate decline ^[12]. Binance Pool, F2Pool and Foundry now hold about 45,900 BTC between them, under 4% of the 1.19 million ^[13]. Foundry also pays block rewards out to the third-party miners who use it ^[8], so its empty balance says little about what its members kept.

Above $80,000, the hash ribbon shows proceeds exceeding expenses, and miners have been out of the distress zone for two months ^[2]. Hashrate sat around 987 EH/s in October, near its one-year average, and difficulty has trended down for most of 2026 ^[3]. Lower difficulty helps whoever keeps machines running. The 450 BTC is a production count, though, so the profit on it rises and falls with how far spot sits above $80,000.

Part of the answer on where the cash went is in the report. Cryptopolitan links some of the selling to the pivot into AI and new data centers, with quantum-risk worries also cited ^[18]. Some miners moved to Zcash for riskier, higher returns ^[19]. Miners used to be holders of last resort, carrying more idle BTC than treasury companies or ETFs ^[22]. According to the report, without the AI narrative the companies no longer rely on their BTC reserves as a value proposition ^[20].

The shares ran well ahead of bitcoin before turning down from a summer peak ^[11]. IREN, at $38.69 after a 6.27% one-day drop, is 43% below its June high of $67.84 ^[9][16]. TeraWulf is about 50% off its 2026 peak at $14.40 ^[10]. The report does not track money leaving miners for other AI stocks. In my view the slide is the AI premium coming out of companies priced on the data-center build, or rather on the expectation of it, while the bitcoin side still earns above cost ^[2].

Three readings fit these numbers. If the coin sales are financing data centers ^[18], the reserve was a funding line and the shares are a construction bet. If the selling was mostly fear of quantum risk or other loss ^[18], it should taper and the reserve should flatten near 1.19 million ^[1]. If bitcoin drops below $80,000 during a winter the report expects to cut hydro-powered mining ^[21], miners go back into distress with 37% less inventory to sell ^[14]. I think the first reading fits best, because selling four years of output in about one year is a decision about where capital goes ^[15]. A reserve figure that rises while spot holds above $80,000 would prove that wrong.