Public bitcoin miners sold 32,000 BTC in Q1 2026 — a quarterly record — and redirected the proceeds into artificial intelligence data center contracts now totaling over $70 billion in aggregate committed revenue. The weighted average cash cost to mine one bitcoin has risen to approximately $80,00...
"We are no longer a Bitcoin company." — Ben Gagnon, CEO, Bitfarms (now Keel Infrastructure)
Public bitcoin miners sold 32,000 BTC in Q1 2026 — a quarterly record — and redirected the proceeds into artificial intelligence data center contracts now totaling over $70 billion in aggregate committed revenue. The weighted average cash cost to mine one bitcoin has risen to approximately $80,000, according to CoinShares, while bitcoin has traded in the $68,000–$70,000 range, producing estimated losses of $19,000 per coin. The economic math has forced the sector's hand: miners are dismantling ASIC racks, zeroing out bitcoin treasuries, and converting power capacity into GPU-dense AI infrastructure.
The transformation is not incremental. Bitfarms rebranded as Keel Infrastructure and announced a full exit from mining. Bitdeer reduced its treasury to zero. TeraWulf has stated its intention to exit mining entirely. Core Scientific now derives two-thirds of its revenue from AI colocation. CoinShares projects that listed miners could generate 70% of total revenue from AI by December 2026, up from roughly 30% in Q1. The first quarterly decline in bitcoin hashrate since 2020 — down approximately 4% year-to-date — suggests the structural reallocation is already visible on-chain.
The April 2024 halving reduced block rewards from 6.25 BTC to 3.125 BTC. Eighteen months later, the consequences are fully priced into mining operations. CoinShares' Q1 2026 mining report puts the weighted average cash cost to produce one bitcoin among public miners at approximately $79,995 — effectively at parity with, or above, market price.
Bitcoin has traded between $68,000 and $70,000 through much of early 2026. At those levels, publicly listed miners lose an estimated $19,000 per coin mined. The post-halving squeeze, compounded by rising energy costs across North American grids, has eliminated the margin buffer that sustained the industry through prior cycles.
The result is a sector-wide conclusion: mining bitcoin at scale, with public-company cost structures and compliance overhead, is no longer economically viable at current prices. The capital, land, and power assets that miners accumulated over a decade are worth more when repurposed.
The aggregate value of AI and high-performance computing (HPC) contracts signed by former bitcoin miners now exceeds $70 billion. The largest deals:
| Company | Contract Partner | Deal Value | Capacity | Term | |---------|-----------------|------------|----------|------| | Hut 8 | Undisclosed (investment-grade) | $9.8B | 352 MW IT | 15 years | | IREN | Microsoft | $9.7B | 200 MW (76,000 GB300 GPUs) | Multi-year | | Cipher Digital | Multiple hyperscalers | $9.0B pipeline | 2.4 GW development | Multi-year | | Hut 8 | Fluidstack | $7.0B | 245 MW IT | 15 years | | Core Scientific | CoreWeave | $10B+ | ~590 MW | 12 years | | TeraWulf | Undisclosed | $12.8B | 522 MW critical | Multi-year |
These are not letters of intent. They are binding, take-or-pay contracts with hyperscalers and enterprise AI firms, structured on triple-net leases that transfer operating risk to the tenant. The contract durations — 12 to 15 years — exceed the typical bitcoin cycle by a factor of three or four.
On May 26, IREN signed a $1.6 billion purchase agreement with Dell for Blackwell GPU systems to service its Microsoft contract, with commissioning targeted for early 2027 at its Childress, Texas campus.
Core Scientific reported Q1 2026 total revenue of $115.2 million, of which $77.5 million — 67% — came from AI colocation. Its CoreWeave contract alone represents over $10 billion in potential revenue across approximately 590 MW. The company has announced plans to liquidate substantially all remaining BTC holdings to fund the transition. Core Scientific sold roughly 1,900 BTC ($175 million) in January alone.
Hut 8 signed two separate 15-year leases totaling $16.8 billion in base-term contract value. The 352 MW Beacon Point deal and 245 MW River Bend campus lease are both structured as triple-net, take-or-pay agreements. Hut 8 shares jumped over 30% on the announcement of the Beacon Point contract on May 6.
IREN secured the $9.7 billion Microsoft deal for 76,000 NVIDIA GB300 GPUs across 200 MW at Childress, Texas. In Q1 FY26 (ending December 2025), total revenue reached $240.3 million, with AI cloud revenues accelerating as deployments ramped. IREN also closed a $3.65 billion GPU-backed financing facility.
TeraWulf generated $21 million in HPC revenue out of $34 million total in Q1 2026 — 62% from AI. Management has explicitly stated its intention to exit bitcoin mining entirely, positioning as a pure-play AI data center developer with 522 critical MW under contract.
Bitfarms/Keel Infrastructure announced a complete exit from bitcoin mining, rebranding as Keel Infrastructure effective around April 1, 2026. The company is selling all bitcoin holdings and pivoting to a 2.2 GW AI data center pipeline across North America. It completed the sale of its 70 MW Paraguay site, exiting Latin America entirely to concentrate capital on North American HPC/AI infrastructure.
Bitdeer reduced its bitcoin treasury to zero on February 22, 2026, selling its remaining 943.1 BTC plus 189.8 BTC produced that week. Despite the liquidation, Bitdeer maintained its position as the largest public miner by self-managed hashrate at 63.2 EH/s. The company priced a $325 million convertible notes offering and $43.5 million equity raise to fund datacenter and AI cloud expansion.
MARA Holdings sold 15,133 BTC for approximately $1.1 billion in March, pivoting capital toward AI infrastructure in partnership with Starwood. CEO Fred Thiel stated: "Bitcoin miners have energy available today. It's an easy pivot."
DMG Blockchain Solutions signed a letter of intent on June 1, 2026, to convert its British Columbia bitcoin mining facility into a 50 MW AI data center — the latest in a string of smaller operators following the large-cap miners' lead.
The scale of bitcoin selling by public miners is without precedent. In Q1 2026 alone, listed miners sold over 32,000 BTC — a quarterly record. Notable liquidations:
This represents a structural reversal of the "HODL" strategy that defined miner treasury management for the prior five years. The selling pressure from miners is now a measurable component of bitcoin's supply dynamics, though its market impact is partially offset by institutional ETF flows in the opposite direction.
Bitcoin's network hashrate posted its first first-quarter decline since 2020, falling approximately 4% year-to-date to hover around 1 ZH/s (zettahash per second). The decline is modest in absolute terms but directionally significant: it reflects the reallocation of power capacity from ASIC rigs to GPU clusters.
CoinShares forecasts the network hashrate will still reach 1.8 ZH/s by year-end 2026 and 2.0 ZH/s by March 2027, driven by efficiency gains in next-generation ASIC hardware (sub-10 J/TH) and private miners filling capacity gaps. However, this forecast is price-dependent.
The security implications are debated. Public miners represent a meaningful but not dominant share of total hashrate. Private miners, state-backed operations (particularly in Russia and the Middle East), and stranded-energy operations continue to add capacity. The concern is not an imminent 51% attack scenario but rather the concentration of hashrate among fewer, less-transparent operators as publicly listed miners exit.
The cost differential explains the pivot. According to CoinShares:
| Metric | Bitcoin Mining | AI Data Center | |--------|---------------|----------------| | Capex per MW | $700K–$1M | $8M–$15M | | Revenue predictability | Volatile (tied to BTC price, difficulty) | Contracted (take-or-pay, 10-15 year terms) | | Revenue per MW | Variable | Structurally higher | | Margin profile | Negative at <$80K BTC | Positive on contracted basis |
AI infrastructure costs 8–15x more per megawatt to build, but it generates structurally higher and more stable returns. The take-or-pay contract model provides revenue certainty that bitcoin mining, subject to price cycles, difficulty adjustments, and halvings, cannot match.
The key asset miners bring is not compute expertise — it is power. Permitted, grid-connected sites with 50–500+ MW of capacity are the bottleneck constraint for AI data center expansion. Miners who secured power purchase agreements and built out grid interconnections during the 2021–2024 buildout now sit on what the AI industry needs most.
A tracked basket of publicly listed mining/AI infrastructure equities is up 56% year-to-date in 2026, while bitcoin itself has fallen approximately 17% over the same period. The market is pricing these companies on their AI contract pipelines, not their mining output.
Hut 8 jumped 30% on a single contract announcement. IREN's $3.65 billion financing closed at terms reflecting data-center multiples, not mining-company valuations. The re-rating is underway: Wall Street coverage is migrating from "crypto mining" desks to data-center and infrastructure analyst teams.
AI demand durability. The contracts assume sustained AI compute demand over 10–15 year horizons. If AI spending contracts — or if hyperscalers build their own capacity faster than expected — miners-turned-landlords face stranded assets at higher capex per MW.
Execution risk. Converting a mining site to a Tier III/IV AI data center requires liquid cooling, redundant power, fiber connectivity, and physical security standards that most mining operations were not designed for. Not all conversions will succeed.
Bitcoin price sensitivity. CoinShares notes that if bitcoin returns to $100,000, mining margins recover and the AI pivot slows. The current trajectory assumes bitcoin stays below miner breakeven for an extended period.
Hashrate centralization. As public miners exit, hashrate concentrates among less-regulated, less-transparent operators. This raises long-term questions about bitcoin's security model, though no acute risk is visible at current levels.
Contract counterparty risk. CoreWeave, Fluidstack, and other AI compute firms are themselves funded by venture capital and have not demonstrated sustained profitability. A $10 billion 12-year contract is only as good as the counterparty behind it.
The publicly listed bitcoin mining sector is undergoing a structural identity change. The companies that built billion-dollar businesses on SHA-256 hashing are now landlords and operators for AI compute, leveraging the same power assets under fundamentally different economics. The transformation is being financed by liquidating the very asset these companies were built to accumulate.
Whether this represents sound capital allocation or a sector chasing the next hype cycle will depend on variables outside miners' control: the duration of AI infrastructure demand, the trajectory of bitcoin's price, and the execution capability of companies transitioning from commodity mining to enterprise data center operations. The contracts are signed. The bitcoin is sold. The pivot is no longer a thesis — it is the operating reality of the sector in 2026.