Public Bitcoin miners sold more than 32,000 BTC in Q1 2026 — exceeding total sales for all of 2025 — as the industry's largest operators redirect capital toward artificial intelligence infrastructure. The liquidation coincides with hashprice falling to $31.73 per PH/s/day as of August 10, a level...
"Bitcoin mining is now essentially in runoff." — Adam Sullivan, CEO, Core Scientific
Public Bitcoin miners sold more than 32,000 BTC in Q1 2026 — exceeding total sales for all of 2025 — as the industry's largest operators redirect capital toward artificial intelligence infrastructure. The liquidation coincides with hashprice falling to $31.73 per PH/s/day as of August 10, a level that renders 15–20% of the global mining fleet unprofitable, according to CoinShares. Publicly listed miners have collectively signed over $70 billion in AI and high-performance computing (HPC) contracts since mid-2024.
The divergence is structural, not cyclical. Network hashrate sits near 855–903 EH/s depending on measurement methodology, while the Bitcoin price hovers around $64,000 — well below the weighted average cash production cost of $79,995 per BTC reported by CoinShares for listed miners in Q4 2025. The result: an industry that built itself on proof-of-work validation is methodically converting its power infrastructure into GPU hosting for frontier AI labs. By year-end 2026, up to 70% of revenue at transformed miners may come from AI, up from roughly 30% at the start of the year.
Hashprice — the daily revenue earned per petahash of mining capacity — stood at $31.73/PH/s/day as of August 10, 2026, according to Hashrate Index. This represents a decline of over 60% from pre-halving levels in early 2024, when miners received 6.25 BTC per block rather than the current 3.125 BTC.
The mathematics are unforgiving. At current difficulty (127.48T as of August 8) and a Bitcoin price near $64,000, the electricity-only breakeven sits around $74,000 per BTC for efficient operations. Full cost models — including depreciation, SG&A, and debt service — exceed $100,000 per coin at many facilities. CoinShares' Q1 2026 mining report pegged the weighted average cash cost among public miners at approximately $79,995.
Riot Platforms reported a cost to mine one bitcoin (excluding depreciation) of $49,912 in Q2 2026. Even at this relatively efficient level, the company posted a net loss of $237.2 million on $174.2 million in revenue for the quarter. Mining revenue specifically fell to $113.7 million, down from $142.9 million a year earlier, despite producing more bitcoin (1,587 vs. 1,426 BTC).
The week of August 3–10 saw hashrate decline while difficulty increased — a divergence that signals marginal miners are shutting down even as the difficulty adjustment lags behind. The next adjustment, estimated at -3.07% on August 23, would mark a modest reprieve. Over the past week, miners collected approximately 3,158 BTC in block rewards, equivalent to roughly $202 million split across the entire global network.
The scale of miner selling in 2026 has no precedent in the industry's corporate history. Publicly traded miners liquidated more than 32,000 BTC during Q1 2026 alone — a figure that exceeds combined sales for the entirety of 2025. Marathon Digital, Core Scientific, Riot Platforms, and Cango were among the largest sellers.
Bitdeer Technologies provides the most extreme case study. The company sold its entire self-mined Bitcoin treasury, reducing its BTC balance to zero. Core Scientific disclosed plans to monetize "substantially all" of its bitcoin holdings during 2026. The company held 2,537 BTC (worth approximately $222 million) at year-end 2025 and sold roughly 1,900 BTC for approximately $175 million in January alone.
The selling is not panic liquidation. It is deliberate capital reallocation. Proceeds are being directed toward AI data center construction, GPU procurement, and debt reduction — investments that management teams view as offering structurally higher and more stable returns than proof-of-work mining.
The aggregate value of AI and HPC contracts announced across the public mining sector exceeds $70 billion, according to CoinShares' Q1 2026 report. The deals are long-term (12–20 years), involve credit-worthy counterparties, and carry operating margins between 80% and 90% — far above the compressed single-digit margins available in Bitcoin mining at current hashprice levels.
The contract pipeline dwarfs the entire Bitcoin mining revenue pool. For context, the global mining network generated approximately $202 million in a single week in early August. The AI contracts represent decades of committed revenue at multiples of that run rate.
Three deals illustrate the scale:
TeraWulf–Anthropic: A 20-year lease worth $19 billion for 401 MW of capacity at a former aluminum smelting facility in Hawesville, Kentucky. TeraWulf expects to invest $3–4 billion — less than one-fifth the lease's value — with initial capacity online in H2 2027 and full operations the following year.
Core Scientific–AMD: A 15-year lease for 529 MW worth more than $14 billion, announced alongside Q2 2026 earnings. This sits on top of Core Scientific's existing 590 MW contract with CoreWeave, originally projected at $10.2 billion over 12 years.
Riot Platforms–Unnamed Frontier AI Lab: A 20-year lease for 191 MW at Riot's Rockdale, Texas campus, expected to generate approximately $9.1 billion over the initial term, with two five-year extension options that could raise total value to $16.1 billion.
Riot Platforms (RIOT): Q2 2026 revenue of $174.2 million, net loss of $237.2 million. Bitcoin mining revenue declined to $113.7 million. Data center revenue reached $33.2 million in Q1 — its first quarter generating AI hosting income. CEO Jason Les called the AI lease "a defining moment in our evolution into a leading developer of large-scale data centers." Together with an existing AMD lease, Riot has contracted 241 MW of critical IT capacity.
Core Scientific (CORZ): Revenue doubled year-over-year in Q2 2026. CEO Adam Sullivan stated bitcoin mining is "essentially in runoff," with certain operations maintained primarily to satisfy minimum power commitments as legacy sites convert to AI colocation. Core Scientific secured $1 billion in financing from Morgan Stanley to accelerate its AI pivot and acquired bitcoin miner Polaris for $421 million to expand its Oklahoma AI data center campus.
IREN Limited (IREN): Disclosed $2.8 billion in new multi-year AI cloud contracts in July 2026 with leading AI developers, raising its year-end 2026 annualized run-rate revenue target from $3.7 billion to over $4 billion. IREN's anchor contract is a $3.4 billion, five-year AI cloud deal with NVIDIA to deploy Blackwell GPUs. The company operates at an 85% project-level EBITDA margin on AI workloads.
TeraWulf (WULF): The Anthropic lease positions TeraWulf as a brownfield conversion specialist — repurposing industrial sites with existing power transmission and fiber-optic infrastructure. The 790-acre Hawesville campus offers cost and timeline advantages over greenfield construction.
The profitability crisis has triggered a parallel dislocation in the ASIC hardware market. Bitmain has cut prices aggressively across multiple generations, offering S19 and S21 machines at $3–4 per terahash. S21 immersion-cooled units sell at roughly $7/TH, while S21+ Hydro models are priced near $8/TH before promotional discounts.
The efficiency threshold for profitability has narrowed sharply. At the current $31.73/PH/day hashprice, any rig performing worse than an Antminer S19 XP facing power costs of $0.06/kWh or higher operates at a loss. Older models — including the once-flagship S19 XP+ Hydro, Whatsminer M60S, and Avalon A1466I — are now unprofitable under most electricity rate scenarios.
Only recent-generation ASICs (S21 series, S23, sub-15 J/TH efficiency) remain economically viable. Fleet-wide efficiency has improved to approximately 18.3 J/TH as of late July, reflecting the ongoing replacement of older hardware, but this improvement is driven partly by unprofitable rigs simply being switched off.
The Cambridge Bitcoin Electricity Consumption Index estimated network power demand at 16.09 GW as of August 1, 2026, with annualized consumption of 141.02 TWh. The modeled range spans 8.61 GW (lower bound) to 28.19 GW (upper bound).
Despite a roughly 35% year-over-year increase in hashrate, energy consumption grew by an estimated 10–15%. The gap reflects rapid deployment of next-generation ASICs — the Antminer S21 Pro and WhatsMiner M60S deliver substantially more hashes per watt than predecessors. In absolute terms, Bitcoin mining accounts for less than 0.5% of global electricity consumption.
Network hashrate touched an all-time high near 1,050 EH/s in December 2025, though it has since moderated to the 855–903 EH/s range in August 2026. The decline aligns with the hashprice squeeze: marginal operators are exiting, and major public miners are redirecting power capacity from mining to AI hosting.
The industry transformation raises a question that few mining executives address publicly: what happens to Bitcoin network security as its largest infrastructure operators convert megawatts from proof-of-work to AI?
In the near term, the network remains secure. Difficulty adjustments ensure that block production continues at roughly 10-minute intervals regardless of how much hashrate exits. The upcoming -3.07% adjustment on August 23 demonstrates this self-correcting mechanism.
The longer-term concern is concentration. As large-scale operators exit mining, hashrate may consolidate among private operators in jurisdictions with the lowest electricity costs — a dynamic that reduces the geographic and institutional diversity of the validator set. Public miners provided a degree of transparency into hash distribution that private operators do not.
The counterargument: AI contracts are 12–20 year commitments. If Bitcoin's price rises substantially, these facilities cannot easily revert to mining. The conversion is, for practical purposes, irreversible at current contract terms. The Bitcoin network's security budget will increasingly depend on transaction fees and smaller-scale miners rather than the publicly traded industrial operators that drove hashrate growth from 2020 to 2025.
The Bitcoin mining industry is undergoing a structural transformation that its architects likely never anticipated. Companies that spent years and billions of dollars acquiring power contracts, constructing facilities, and deploying ASICs are now methodically converting those same assets to serve a different computational workload — one that offers higher margins, longer contract durations, and more predictable cash flows.
The shift is rational. At $31.73/PH/s/day hashprice and $79,995 weighted average production cost, mining Bitcoin is a money-losing proposition for most public operators. AI hosting at 80–90% operating margins is not. The capital markets have endorsed this logic: stocks of miners with significant AI contract backlogs have substantially outperformed pure-play Bitcoin miners in 2026.
What remains unclear is the second-order effect on the Bitcoin network itself. The hashrate has already declined from its December 2025 peak above 1,000 EH/s. If the price does not recover meaningfully, more marginal capacity will exit. The network will continue functioning — difficulty adjustments guarantee that — but the composition of its validator set is changing in ways that merit monitoring.
The companies that built themselves on Bitcoin are leaving Bitcoin behind. The infrastructure stays. The purpose changes.