Public bitcoin miners have signed more than $70 billion in cumulative AI and high-performance computing (HPC) contracts since late 2025, triggering the first quarterly hashrate decline in six years and a record liquidation of 32,000 BTC in Q1 2026 alone — more than the sector sold in all of 2025....
"The ability for miners to convert to AI is one of the largest infrastructure shifts of this decade." — Adam Sullivan, CEO of Core Scientific
Public bitcoin miners have signed more than $70 billion in cumulative AI and high-performance computing (HPC) contracts since late 2025, triggering the first quarterly hashrate decline in six years and a record liquidation of 32,000 BTC in Q1 2026 alone — more than the sector sold in all of 2025. The network's total computational power has retreated from a peak of 1.16 zettahash per second (ZH/s) in October 2025 to approximately 0.96 ZH/s in mid-July 2026, a decline of roughly 17%.
The economic logic is straightforward: AI infrastructure generates $8 million to $15 million per megawatt of revenue, compared with $700,000 to $1 million per megawatt for bitcoin mining, according to CoinShares. With the weighted average cash cost to produce one bitcoin at approximately $79,995 and spot prices hovering near $64,000, publicly listed miners are losing roughly $16,000 per coin produced. The pivot is not a strategic choice. It is a survival mechanism.
Mining equities have responded accordingly, decoupling from bitcoin's price. Former mining stocks have outperformed BTC by 70% year-to-date in 2026. Hut 8 is up 131%, TeraWulf has gained 73.6%, and Core Scientific sits above 40% — all while bitcoin itself trades 49% below its October 2025 all-time high of $124,500.
The April 2024 halving cut bitcoin's block subsidy from 6.25 BTC to 3.125 BTC. The network's difficulty adjustment algorithm continued to increase mining difficulty through late 2025 as hashrate climbed. These two forces — halved revenue, rising costs — have created an extended margin squeeze that CoinShares describes as the most challenging period for miners since the halving.
Key figures from the CoinShares Q1 2026 Bitcoin Mining Report:
| Metric | Value | |---|---| | Weighted avg. cash cost per BTC (Q4 2025) | ~$79,995 | | BTC spot price (July 2026) | ~$64,000 | | Implied loss per BTC produced | ~$16,000 | | Hashprice (July 2026) | ~$29/PH/s/day | | Hashprice at halving (April 2024) | ~$70/PH/s/day | | Fleet unprofitable at current economics | 15–20% |
The hashprice — daily revenue per petahash of deployed mining power — has fallen to approximately $29 per PH/s per day in July 2026, according to Hashrate Index data. This represents a 59% decline from the $70/PH/s level seen at the time of the April 2024 halving. At current levels, only operators running next-generation ASICs (Antminer S21 XP, S23 Hydro, and equivalents rated under 15 J/TH) remain profitable, and only at electricity rates below $0.088/kWh, according to mining profitability analysis from D-Central and CoinWarz.
JPMorgan estimated that bitcoin has traded below its average production cost for five consecutive months as of July 2026. The bank's analysts noted that mining economics "worsened" materially, with BTC trading 19% below the estimated average cost of production.
The $70 billion figure represents cumulative contracted AI and HPC revenue across the public mining sector. The largest individual agreements:
| Company | Counterparty | Contract Value | Term | Capacity | |---|---|---|---|---| | TeraWulf (WULF) | Anthropic | $19B | 20 years | 401 MW | | Core Scientific (CORZ) | CoreWeave | $10.2B | 12 years | 590 MW | | Hut 8 (HUT) | Undisclosed | $9.8B+ | 15 years | River Bend campus | | Core Scientific (CORZ) | CoreWeave | $1.2B expansion | — | Denton, TX site |
TeraWulf-Anthropic stands as the single largest transaction. Announced July 6, 2026, the 20-year triple-net lease covers a purpose-built AI infrastructure campus at TeraWulf's Justified Data site in Hawesville, Kentucky. The facility will accommodate 401 MW of critical IT load, with initial capacity expected online in H2 2027 and full ramp by early 2028. WULF shares jumped 14% on the announcement day, according to data from Motley Fool.
Core Scientific rejected a $9 billion all-stock acquisition offer from CoreWeave in October 2025 — stockholders voted against the merger after proxy advisors ISS and Glass Lewis raised concerns, and investors Two Seas Capital and VanEck publicly opposed the deal. Core Scientific subsequently maintained its independence and expanded the existing CoreWeave hosting relationship. As of Q1 2026, Core Scientific was providing CoreWeave with 243 MW of compute, with 347 MW scheduled for delivery in early 2027. The company also sought a $3.3 billion bond sale in April 2026 to fund further AI data center conversion.
Hut 8 announced that its Beacon Point location secured a 15-year triple-net lease valued at $9.8 billion, which could exceed $25 billion with escalators, per Q1 2026 disclosures.
Multiple miners have indicated they could derive up to 70% of total revenue from AI by end of 2026, effectively completing a transformation from mining companies to data center operators that mine bitcoin on the side.
Bitcoin's network hashrate trajectory in 2026:
| Date | Hashrate | Event | |---|---|---| | October 2025 | 1.16 ZH/s | All-time peak | | January 25–26, 2026 | 663 EH/s | Winter Storm Fern (temporary) | | February 2026 | ~850 EH/s | Post-storm recovery; structural decline begins | | Q1 2026 avg. | ~950 EH/s | First Q1 drop in 6 years | | July 2026 | ~960 EH/s (0.96 ZH/s) | Current level |
The Q1 2026 decline was the first first-quarter hashrate drop since 2020, according to CoinDesk reporting. While the January plunge to 663 EH/s was driven by Winter Storm Fern — an arctic weather system that caused widespread power outages — the sustained decline through H1 2026 reflects structural capital reallocation.
The average block time over the most recent 2,016-block difficulty epoch was 11 minutes 37 seconds in mid-July, running 1 minute 37 seconds slower than the 10-minute target. A difficulty retarget is estimated for July 27, 2026, which is expected to adjust downward to reflect reduced hashrate.
Mining difficulty had previously hit all-time highs as hashrate approached 1 ZH/s in late 2025 but has since undergone multiple downward adjustments.
Public miners collectively sold approximately 32,000 BTC in Q1 2026, according to CoinTelegraph and CoinMarketCap data. This figure exceeds total net miner sales for the entirety of 2025 and surpasses the previous quarterly record of 20,000 BTC set during Q2 2022 — a period defined by the Terra-Luna collapse.
Major sellers included MARA (Marathon Digital), CleanSpark, Riot Platforms, Cango, Core Scientific, and Bitdeer. Core Scientific confirmed plans to sell the majority of its bitcoin holdings by end of 2026 to fund AI conversion of its 1.2 GW facility capacity.
The contrast with corporate bitcoin treasury strategies is notable. Strategy (formerly MicroStrategy) continued accumulating bitcoin during the same period, with co-founder Michael Saylor signaling continued purchases. Miners are selling to fund an infrastructure pivot; treasury companies are buying on the assumption of long-term price appreciation. The two strategies represent divergent views on whether bitcoin's value accrues to holders or operators.
Nakamoto Inc., a smaller public miner, sold 600 BTC for $48 million in June 2026 specifically to repay debt owed to Kraken. Shares rose 18.1% on the announcement, suggesting the market rewards deleveraging in the current environment.
Year-to-date 2026 performance of former mining stocks, compiled from Yahoo Finance and MarketBeat data:
| Stock | YTD Gain (2026) | Recent Price | Primary AI Revenue Driver | |---|---|---|---| | Hut 8 (HUT) | +131% | ~$118 | $9.8B+ Beacon Point lease | | TeraWulf (WULF) | +73.6% | ~$24.05 | $19B Anthropic lease | | Cipher Digital (CIFR) | +58% | ~$25 | AI compute conversion | | Core Scientific (CORZ) | +40%+ | — | CoreWeave hosting expansion | | Bitcoin (BTC) | -49% from ATH | ~$64,000 | — |
Morgan Stanley issued coverage noting "big upside" in WULF, CIFR, and MARA after the sector's AI pivot, according to Stocktwits reporting. The decoupling is structural: these companies are now valued on contracted AI revenue and megawatt capacity, not bitcoin production or treasury holdings.
Mining stocks have collectively outperformed bitcoin by approximately 70% in 2026, according to Bitcoin.com analysis. The market is pricing these equities as AI infrastructure plays, not cryptocurrency proxies.
The migration raises a question about bitcoin's long-term security model. The network's security budget — the total economic cost required to attack the chain — is a function of hashrate and hardware costs.
Key considerations, drawn from analysis by KuCoin Research, CoinDesk, and Techi:
Approximately 20% of miners are currently operating below production cost, according to CoinShares estimates. The Miner Cycle Stress Composite and Puell Multiple have both entered ranges historically associated with miner capitulation.
The bitcoin mining industry is undergoing a structural transformation that is unlikely to reverse. The economics are unambiguous: AI infrastructure generates 8x to 15x more revenue per megawatt than bitcoin mining. With production costs exceeding spot prices by $16,000 per coin, the rational response is capital reallocation, and that is precisely what is occurring at scale.
The implications extend beyond mining equities. Bitcoin's security model depends on sufficient economic incentive for miners to secure the network. As the block subsidy continues halving and miners redirect capacity toward AI, the network's long-term security budget becomes increasingly dependent on transaction fee revenue. Whether bitcoin's fee market can develop to fill this gap remains an open question — and one the market has not yet priced.
For now, the data is clear: what were once bitcoin mining companies are becoming AI infrastructure companies. The hashrate they leave behind will be absorbed by the difficulty adjustment. But the capital they take with them is not coming back.