Bitcoin's mining sector is undergoing a structural transformation. Nine publicly listed mining companies have announced partial or full pivots to artificial intelligence and high-performance computing infrastructure since the start of 2026, collectively signing more than $70 billion in AI/HPC con...
"We are no longer a Bitcoin company. We are an infrastructure-first owner and developer for HPC/AI data centers across North America." — Ben Gagnon, CEO, Keel Infrastructure (formerly Bitfarms)
Bitcoin's mining sector is undergoing a structural transformation. Nine publicly listed mining companies have announced partial or full pivots to artificial intelligence and high-performance computing infrastructure since the start of 2026, collectively signing more than $70 billion in AI/HPC contracts, according to CoinShares' Q1 2026 Bitcoin Mining Report. The shift has drained roughly 100 exahashes per second (EH/s) from the network, triggered the 11th-largest downward difficulty adjustment in Bitcoin's history on June 20, and raised questions about the long-term security budget of the world's largest proof-of-work blockchain.
The economics are straightforward. JPMorgan estimates the weighted average all-in production cost of one bitcoin at approximately $78,000. Bitcoin has traded below that figure for five consecutive months, hovering near $63,000–$67,000 through June 2026. Public miners responded by selling a record 32,000 BTC in Q1 2026 — more than they sold in all four quarters of 2025 combined — and redirecting capital toward AI compute, where revenue per megawatt is higher and margins are more predictable.
CoinShares projects that as much as 70% of listed miner revenue could come from AI infrastructure by the end of 2026, up from roughly 30% at the start of the year. The result is a sector that increasingly resembles a data center business with a bitcoin mining side operation — not the other way around.
On June 20, 2026, Bitcoin's mining difficulty fell 10.09% at block height 953,568, dropping from 138.96 trillion to 124.93 trillion. The adjustment ranks as the 11th-largest downward move in the network's 17-year history and the second-largest of 2026, following an 11.16% cut on February 7 and a 7.76% reduction in March.
Three downward adjustments exceeding 5% within a six-month window is unusual. The pattern indicates sustained hashrate departure rather than a temporary fluctuation. The adjustment cycle lasted approximately 15.6 days — significantly longer than the theoretical 14-day target — confirming that substantial mining power went offline during the period.
Bitcoin's network hashrate peaked at approximately 1,160 EH/s in October 2025. By February 2026 it had fallen to 850 EH/s, a decline of roughly 27% from peak. More recent intraday readings have shown drops as low as 663 EH/s before partial recovery. The network currently hovers around 1 ZH/s (1,000 EH/s), but the trend line since Q4 2025 points downward — the first sustained quarterly decline in hashrate since 2020.
The April 2024 halving cut block rewards from 6.25 BTC to 3.125 BTC. Network difficulty, meanwhile, has increased roughly tenfold since 2021 levels. The combined effect: mining profitability has compressed by an order of magnitude.
JPMorgan analysts led by managing director Nikolaos Panigirtzoglou stated in a June 2026 client note that bitcoin mining economics have "worsened." The bank places the weighted average cash cost to produce one bitcoin among publicly listed miners at approximately $79,995 as of Q4 2025, with all-in production costs near $78,000 when accounting for hardware depreciation and overhead. Bitcoin has traded 19% below that figure for five straight months.
Hashprice — the daily revenue per petahash per second — sits near $33/PH/s/day, below the approximately $35 breakeven threshold estimated for the median miner. Roughly 20% of active miners are operating at a loss, according to JPMorgan data.
The result is a rational economic response: shut down unprofitable machines, sell bitcoin reserves, and redirect infrastructure toward higher-margin workloads.
The aggregate value of AI and high-performance computing contracts announced by public Bitcoin miners now exceeds $70 billion, according to CoinShares. The scale of individual deals underscores the economics driving the transition:
| Company | AI/HPC Contract Value | Term | Details | |---|---|---|---| | Core Scientific | $10.2B | 12 years | Expanded CoreWeave colocation deal | | TeraWulf | $12.8B | Multi-year | Contracted HPC revenue | | Hut 8 | $7.0B | 15 years | River Bend campus AI lease | | IREN | $9.7B | Multi-year | Microsoft deal: 76,000 NVIDIA GB300 GPUs, 200 MW at Childress, TX |
These are not speculative memoranda of understanding. Core Scientific's AI colocation revenue already accounts for 39% of its total revenue. IREN has 200 MW of liquid-cooled GPU capacity under construction. The contracts involve named counterparties — Microsoft, CoreWeave — with defined capacity and duration.
The revenue differential explains the shift. Bitcoin mining generates approximately $33/PH/s/day at current hashprice levels, with margins negative for a significant portion of operators. AI compute, by contrast, offers higher and more predictable per-megawatt revenue with multi-year contract visibility.
As of June 2026, nine publicly listed mining companies have announced pivots of varying degrees:
Full exit from mining: Bitfarms completed its transformation into Keel Infrastructure on April 1, 2026, redomiciling from Canada to the United States and trading on Nasdaq and TSX under the symbol KEEL. CEO Ben Gagnon declared the company is "no longer a Bitcoin company." The company held approximately 2,500 BTC that it planned to sell gradually to fund the transition. Its portfolio spans 2.2 GW of energized, secured, and expansion-stage capacity across Pennsylvania, Quebec, and Washington. Keel reported a $145 million net loss in Q1 2026 during the transition.
Hybrid operations (mining + AI): Core Scientific, IREN, Hut 8, TeraWulf, CleanSpark, HIVE Digital, Cipher Mining, and Riot Platforms have all announced AI/HPC pivots while maintaining some mining operations. The degree of pivot varies — Core Scientific derives 39% of revenue from AI already, while others are in earlier stages of infrastructure conversion.
Revenue projection: CoinShares estimates that 70% of listed miner revenue could derive from AI by year-end 2026, up from approximately 30% at the start of the year. This represents a structural shift in what these companies are, regardless of their legacy identities as "Bitcoin miners."
The migration of hashrate away from Bitcoin raises a question that has been debated since the network's early years: is the security budget sufficient?
Bitcoin's security model rests on the economic cost of attacking the network. A 51% attack becomes feasible when the cost of acquiring or deploying majority hashpower falls below the potential profit from double-spending or censorship. Lower hashrate directly reduces that cost.
During the June trough, when hashrate dipped below 700 EH/s intraday, the theoretical cost of a 51% attack fell 30–40% from October 2025 peak levels, according to analysis from KuCoin Research. The difficulty adjustment mechanism — which recalibrates every 2,016 blocks — eventually restores equilibrium, but the window between hashrate departure and adjustment creates a temporary vulnerability.
The structural concern extends beyond short-term fluctuations. Block rewards shrink by 50% every four halving cycles by design. Transaction fees have not grown proportionally to offset the decline. If AI generates more stable, higher-margin revenue than Bitcoin mining, rational miners will continue reallocating capital, resulting in a structurally lower hashrate and a diminished economic security budget.
A concentration risk also emerges. If most miners exit while a few move forward aggressively, the remaining operators could hold a disproportionate share of hashrate within 12–18 months. Hashrate distribution data already shows the top five mining pools controlling a significant majority of block production.
The AI pivot is being financed through two primary channels.
Bitcoin sales: Public miners sold a record 32,000 BTC in Q1 2026, according to data compiled by Blockonomi and The Block. Major sellers included MARA, CleanSpark, Riot Platforms, Cango, Core Scientific, and Bitdeer. The Q1 figure alone exceeds the approximately 20,000 BTC sold during Q2 2022 — a period marked by the Terra-Luna collapse — and surpasses these companies' combined sales for all of 2025.
Convertible debt: IREN now carries $3.7 billion in convertible notes across five series. Publicly listed miners have collectively reduced their BTC treasuries by over 15,000 BTC from peak levels to fund infrastructure conversion.
The selling pressure from miners adds supply to a market already contending with $1.67 billion in weekly Bitcoin ETF outflows as of mid-June 2026 and sustained "extreme fear" readings on sentiment indices for 30 consecutive days.
The data describes a sector undergoing identity change. Companies that built their businesses on SHA-256 computation are repurposing the same power infrastructure, cooling systems, and grid connections for GPU-based AI workloads. The economic logic is clear: when Bitcoin trades 19% below production cost for five months while AI contracts offer multi-year revenue visibility at higher margins, capital follows the return.
For Bitcoin, the implications are measurable. Hashrate is declining. Difficulty is adjusting downward at historically significant magnitudes. Miner selling is at record levels. None of these individually constitute a crisis — the difficulty adjustment mechanism is functioning as designed, and the network continues to produce blocks. But the trend raises a structural question that market participants and protocol developers will need to address: as block rewards continue to halve on schedule, and as the infrastructure operators who once secured the network find more profitable uses for their capacity, what sustains Bitcoin's security budget?
The mining sector's answer, for now, is that Bitcoin mining is a feature of their business — not the business itself. Whether the network can function securely under that premise remains an open question.