Public Bitcoin miners sold 32,000 BTC in Q1 2026 — more than in all of 2025 — to fund a mass migration toward artificial intelligence infrastructure. The weighted average cash cost to mine one bitcoin rose to approximately $79,995 by Q4 2025, while bitcoin traded in the $68,000–$70,000 band, tran...
"The first quarter of 2026 marks a definitive inflection point for Riot, as we officially transitioned into an active, revenue-generating data center operator." — Jason Les, CEO, Riot Platforms
Public Bitcoin miners sold 32,000 BTC in Q1 2026 — more than in all of 2025 — to fund a mass migration toward artificial intelligence infrastructure. The weighted average cash cost to mine one bitcoin rose to approximately $79,995 by Q4 2025, while bitcoin traded in the $68,000–$70,000 band, translating to estimated losses of $19,000 per coin mined. Over $70 billion in cumulative AI and high-performance computing (HPC) contracts have been signed across the sector.
The economics are stark. Core Scientific's colocation revenue surged 800% year-over-year to $77.5 million in Q1 2026, while its self-mining revenue fell 55% to $30.1 million. TeraWulf locked $12.8 billion in contracted HPC revenue. Hut 8 signed a 15-year, $7 billion lease with Anthropic-backed Fluidstack, with options extending to $17.7 billion. Industry projections from Bloomberg suggest listed miners could derive 70% of revenue from AI by year-end 2026, up from approximately 30% in early 2026.
This report examines the financial, operational, and structural dimensions of this transition — arguably the largest capital reallocation event in the mining sector's history.
The April 2024 halving cut Bitcoin's block subsidy from 6.25 BTC to 3.125 BTC. Two years later, the full impact is visible in the data.
Hashprice — the revenue earned per terahash per second per day — currently ranges from $0.055 to $0.065/TH/s/day, according to Hashrate Index data. This is frequently below breakeven for a significant portion of the global mining fleet. Daily network mining revenue dropped to yearly lows of $28 million in late January 2026.
Bitcoin's network hashrate, meanwhile, reached approximately 1.0 ZH/s (zetahash) in May 2026, according to CoinWarz, with mining difficulty at 136.61 T. The paradox: hashrate keeps climbing even as per-unit revenue declines, driven by already-deployed hardware continuing to operate at marginal cost.
The post-halving margin compression sorted miners into three tiers:
| Tier | Energy Cost | Hardware Efficiency | Margin at Current Hashprice | |------|-------------|--------------------|-----------------------------| | Industrial-scale (Tier 1) | < $0.04/kWh | < 15 J/TH | 30–50% | | Mid-tier operators | $0.04–$0.08/kWh | 15–25 J/TH | 0–20% | | Home/small-scale miners | > $0.08/kWh | > 25 J/TH | Negative |
The weighted average fleet efficiency improved to 34 W/T in 2025, an 8% year-over-year gain. Projections indicate sub-10 W/T hardware could arrive by mid-2026. But efficiency gains alone cannot offset a 50% subsidy cut when bitcoin's price fails to double in response.
The core thesis driving the pivot is straightforward: AI compute delivers higher, more predictable returns per megawatt than bitcoin mining.
According to S&P Global Market Intelligence research published in February 2026, the revenue per megawatt for AI/HPC colocation ranges from $1.2 million to $1.8 million annually, compared to $400,000 to $700,000 for bitcoin mining at current hashprice levels. Critically, AI contracts are typically structured as multi-year, fixed-price agreements with hyperscaler counterparties — a fundamentally different risk profile than mining's exposure to bitcoin price volatility and difficulty adjustments.
Core Scientific's Q1 2026 data illustrates this: colocation revenue of $77.5 million on 243 MW billed translates to an annualized rate of approximately $1.44 million per MW. Its remaining self-mining revenue of $30.1 million operates across a broader power footprint at lower revenue density.
The tradeoff is capital intensity. AI-grade data centers require liquid cooling, redundant power, and network connectivity that mining facilities — designed for brute-force ASIC computation — typically lack. CleanSpark estimates build-out costs of $9–$11 million per MW for AI conversion, according to a Needham analyst note from May 2026.
Public miners sold 32,000 BTC in Q1 2026 alone, according to KuCoin research data. This exceeds the total BTC sold by public miners throughout 2025. Marathon Digital led with over 13,000 BTC liquidated, followed by Core Scientific at approximately 1,900 BTC ($175 million) in January alone.
The selling is structural, not opportunistic. With cash cost per BTC mined at ~$80,000 and bitcoin trading near $73,500 as of May 29, 2026 (per Yahoo Finance), miners are selling at a loss to fund AI infrastructure buildouts that promise higher returns per megawatt.
This creates a measurable supply overhang. At current prices, 32,000 BTC represents approximately $2.4 billion in sell pressure injected into spot markets during Q1. As CoinDesk reported in March, "more [selling] is coming" as miners accelerate their capital reallocation strategies.
The pivot has shifted the industry's key performance indicator from exahash to megawatts. Total contracted or under-development AI/HPC capacity across the seven miners analyzed exceeds 15 GW, a figure that rivals the power consumption of medium-sized countries.
| Company | Current AI/HPC MW | Contracted/Pipeline MW | Primary AI Customer | |---------|-------------------|----------------------|---------------------| | Core Scientific | 243 | 590+ | CoreWeave | | Hut 8 | — | 2,295+ (Anthropic pipeline) | Fluidstack/Anthropic | | TeraWulf | 60 | Scaling | Multiple | | Riot Platforms | Active (new) | 2,000 (power portfolio) | AMD | | CleanSpark | Transitioning | 300–600 | Hyperscaler TBD | | IREN | Under construction | 200+ | Microsoft | | MARA | Planning | TBD | Exaion (JV) |
The customer concentration risk is notable. CoreWeave alone anchors Core Scientific's AI business. Anthropic and Google backstop Hut 8. Microsoft underpins IREN. A credit event or contract renegotiation at any single hyperscaler could impair billions in contracted revenue.
The market is pricing miners not as bitcoin proxies but as infrastructure plays. According to CryptoNews, the miners index has outperformed bitcoin by 70% in 2026, driven almost entirely by AI-pivoting names.
The divergence is stark within the sector itself:
Institutional positioning reflects this bifurcation. According to BitcoinMiningStock.io, institutional capital is concentrating in IREN, CORZ, and APLD (Applied Digital), while pure-mining exposure is declining. The market is effectively pricing two industries within one sector: high-multiple AI infrastructure and low-multiple bitcoin mining.
Execution risk. Converting mining facilities to AI-grade data centers is capital-intensive and technically complex. Build-out costs of $9–$11 million per MW at CleanSpark's estimate mean a 500 MW conversion requires $4.5–$5.5 billion. Construction delays or cost overruns could erode projected returns.
Customer concentration. The top three AI customers (CoreWeave, Anthropic/Google, Microsoft) account for the majority of contracted revenue across the sector. Any single counterparty default would be material.
BTC treasury depletion. Selling 32,000 BTC per quarter reduces miners' leverage to any future bitcoin price appreciation. If bitcoin rallies significantly, miners who liquidated reserves will have permanently forfeited upside.
AI demand cycle. The current AI infrastructure buildout assumes sustained demand growth for compute. A slowdown in AI model training cycles, regulatory intervention, or capacity oversupply could compress the margins that make the pivot attractive.
Energy cost volatility. WTI crude at $97/barrel in March 2026 squeezed margins for energy-intensive operators. Miners-turned-AI-hosts remain exposed to energy price spikes, though fixed-price contracts partially insulate revenue.
The bitcoin mining industry is undergoing the most significant capital reallocation in its history. The post-halving margin compression, combined with AI's superior unit economics per megawatt, has made the pivot economically rational for most public operators. The data shows this clearly: a megawatt allocated to AI colocation generates 2–3x the revenue of one allocated to bitcoin mining, with lower volatility and contractual revenue visibility.
The transition is not without risk. Billions in contracted revenue depend on a handful of hyperscaler counterparties. Build-out costs are substantial. And miners who liquidate their BTC treasuries permanently forfeit optionality on future price appreciation.
What the sector is building is, functionally, a new industry wearing the shell of an old one. The mining companies that survive this transition will bear little resemblance to the firms that emerged from the 2024 halving. Whether they are valued as bitcoin miners, AI infrastructure providers, or some hybrid remains an open question — one the market is answering in real time through a widening performance gap between those who pivot and those who do not.