Bitcoin Miners Sell BTC and Load Up on Debt as AI Pivot Reshapes the Sector
Bitcoin mining is undergoing a structural reorientation that markets have been slow to price correctly. A new CoinShares report, released Wednesday, documents the confluence of compressed hash prices, rising weighted-average cash costs, and an accelerating pivot toward AI and high-performance computing infrastructure, all while Bhutan’s sovereign treasury has quietly offloaded more than $152 million in Bitcoin since January, reducing its reserves by 66% from peak levels.
The arithmetic of mining profitability has deteriorated materially over the past two quarters. CoinShares calculates that public miners’ weighted-average cash cost per Bitcoin reached roughly $79,995 in the fourth quarter of 2025. Against that cost baseline, the hash price fell from approximately $36 to $38 per petahash per second per day in Q4 2025, then dropped further to around $29 in Q1 2026. The live hash price, as of the report’s release, sits near $32.36 per PH per day, with transaction fees contributing just 0.40% of block rewards. The six-month forward market average hash price is priced even lower, at approximately $30.42. The network has also logged three consecutive negative difficulty adjustments, the first such streak since July 2022, which signals that a meaningful portion of capacity is being switched off. CoinShares estimates that between 15% and 20% of the Bitcoin mining fleet is currently operating at a loss.
What miners do under those conditions is structurally important and frequently misread by retail participants. The persistent assumption that large miner treasury balances represent structurally bullish supply overhang is not supported by how miners actually behave when margins deteriorate. MARA revised its strategy in 2025 to permit sales of Bitcoin from operations, then extended that policy in 2026 to include balance sheet holdings. Riot Platforms sold 1,818 BTC in December 2025 for $161.6 million. Core Scientific sold just over 1,900 BTC in January 2026 for approximately $175 million and now holds fewer than 1,000 BTC on its balance sheet. Riot separately liquidated roughly 1,080 BTC to fund a 200-acre land acquisition at Rockdale. These are not discretionary portfolio decisions; they are the behaviour of commodity producers managing liquidity under cost pressure, and the selling is concentrated precisely when Bitcoin is already weak, making treasury policy pro-cyclical rather than stabilising. This pattern aligns with the broader wave of large BTC holders resuming distribution that has characterised the first quarter of 2026.
The AI Pivot Is Not Speculative; It Is Already Generating Revenue
Public miners collectively hold 121,516 BTC, worth approximately $8.63 billion at current prices. That inventory makes them meaningful marginal sellers, but the more consequential structural shift is the recomposition of their revenue base. CoinShares estimates that listed miners could derive up to 70% of revenues from AI by end-2026, compared with roughly 30% today. That is not a forecast premised on future contracts; it is grounded in capacity that is already being energised or contractually committed.
Core Scientific has energised approximately 350 MW for CoreWeave and is targeting around 590 MW by early 2027. Its Q4 2025 revenue already showed $42.2 million from self-mining against $31.3 million from colocation, a split that will almost certainly invert further as new capacity comes online. Hut 8 has signed a 15-year, 245 MW AI data centre lease with a $7 billion base-term value. IREN reported $17.3 million in AI Cloud Services revenue and secured $3.6 billion in GPU financing tied to a Microsoft contract, with management guiding toward a $3.4 billion annual recurring revenue target by end-2026. TeraWulf has reported more than $12.8 billion in long-term customer contracts and $6.5 billion in long-term financings completed in 2025. Riot has signed its first AMD data-centre lease. These are not pivot announcements; they are executed or contracted positions with identifiable counterparties and financing structures.
The mechanism funding much of this transition is debt. Miners are borrowing to retrofit data centres, acquire land, and build out power infrastructure, simultaneously selling Bitcoin from operations and balance sheets to manage near-term liquidity. For equity investors, that combination means a listed miner stock now bundles exposure to Bitcoin price, hyperscaler demand cycles, lease execution risk, retrofit capital expenditure, financing costs, and counterparty quality within a single ticker. CoinShares describes this explicitly as a bifurcation, with AI and HPC-linked names attracting valuation premiums relative to pure-play miners whose economics remain directly tethered to hash price.
Bhutan’s Liquidation Accelerates as the Gelephu Pledge Becomes Untenable
Against the miner-side selling, sovereign disposals from Bhutan add a separate, independent source of supply pressure. The Royal Government transferred 519.7 BTC valued at approximately $36.75 million on Wednesday, the latest in a series of transactions tracked by Arkham Intelligence. Cumulative 2026 outflows have now surpassed $152 million. At their peak in late 2024, Bhutan’s reserves reached approximately 13,000 BTC, a portfolio that approached $1.88 billion in value at Bitcoin’s near-$119,000 highs. Current holdings stand at 4,453 BTC, worth roughly $315 million, representing a 66% reduction from peak. Bitcoin has traded between $65,000 and $75,000 throughout March, well below those prior highs.
The liquidation pattern has escalated in scale as the year has progressed. Individual transfers during January and February ranged from $5 million to $15 million. By March, transaction sizes had risen to the $35 million to $45 million bracket. The prior week saw approximately $72 million transferred within a seven-day period, with the single largest movement involving 595.8 BTC worth $44.44 million. Singapore-based QCP Capital has been the recipient of at least three distinct transfers totalling approximately $16.6 million year-to-date, a recurring pattern consistent with a formalised over-the-counter distribution arrangement. OTC structures allow a large sovereign holder to unwind positions without directly hitting exchange order books, minimising immediate price impact, though the supply is absorbed into the broader market nonetheless.
Bhutan originally accumulated its Bitcoin through government-operated hydroelectric mining facilities. With surplus hydropower keeping marginal production costs near zero, every unit sold represents near-pure revenue for the state. Blockchain monitoring reveals virtually no fresh Bitcoin entering Bhutan’s wallets from mining activity in recent months, suggesting that operations were scaled back or suspended following the April 2024 halving. The fiscal logic is straightforward: costs are sunk, the asset has appreciated substantially from cost basis regardless of the price decline from 2024 highs, and the proceeds serve immediate sovereign spending needs. Last December, Bhutan announced a Bitcoin Development Pledge earmarking up to 10,000 BTC toward its Gelephu Mindfulness City project, a commitment valued at approximately $860 million at the time of announcement. With current holdings below 4,500 BTC and the sell-off continuing, fulfilling that pledge in its original form is arithmetically impossible without a complete reversal of current policy and substantial open-market purchases.
Taken together, the CoinShares mining data and Bhutan’s liquidation trajectory describe a market in which two structurally distinct categories of seller, industrial miners managing cost-price compression and a sovereign government monetising a near-zero-cost asset, are simultaneously active. Neither is irrational given their respective circumstances, and neither is likely to reverse quickly. The conditions that produced the current hash price, post-halving supply reduction combined with BTC prices well below the 2024 peak, are not transitory dislocations; they reflect a recalibration of the economics that supported the prior mining expansion cycle. Investors treating miner equity as a simple Bitcoin proxy, or interpreting sovereign selling as a temporary overhang, are working with an incomplete model of the supply side as it now functions.
