Publicly traded Bitcoin miners reduced realized hashrate by 13.4% between Q4 2025 and Q2 2026, redirecting power and data center capacity toward artificial intelligence and high-performance computing. Excluding Bitdeer — the one major operator that increased mining output — the cohort's hashrate ...
"We don't underwrite applications. We underwrite scarce power. Applications change, customer demand changes, technology changes." — Asher Genoot, CEO, Hut 8 Corp.
Publicly traded Bitcoin miners reduced realized hashrate by 13.4% between Q4 2025 and Q2 2026, redirecting power and data center capacity toward artificial intelligence and high-performance computing. Excluding Bitdeer — the one major operator that increased mining output — the cohort's hashrate fell 21.2%, from 324.6 EH/s to 255.9 EH/s, according to BlocksBridge Consulting data published August 16, 2026. The broader Bitcoin network recorded a 10.6% decline over the same period, from 1,071 EH/s to 957 EH/s, marking its first sustained contraction since 2020.
The catalyst is economic. With a weighted average cash cost near $80,000 per BTC against a spot price hovering around $64,000, industrial-scale mining is producing negative unit economics for a substantial portion of the listed cohort. Simultaneously, AI infrastructure demand has generated over $70 billion in cumulative contract value across the public mining sector, with CoinShares projecting AI and HPC services could account for 70% of listed miners' revenue by year-end 2026, up from approximately 30% at the start of the year.
The result is a structural transformation of an industry. Companies that entered 2026 as Bitcoin miners are exiting the year as power infrastructure platforms. The economic value is migrating from proof-of-work block rewards toward long-duration colocation leases with hyperscalers, frontier AI labs, and semiconductor manufacturers.
Bitcoin's network hashrate dropped from 1,071 EH/s in Q4 2025 to 957 EH/s in Q2 2026. As of August 11, 2026, network difficulty stands at 127.48T, with the next adjustment on August 23 projected to decrease by approximately 3.07%, according to CoinWarz data. Hashprice, the revenue metric per petahash per day, sits at $32 — at or below breakeven for operators running older-generation hardware.
The publicly traded cohort's hashrate decline from 368.3 EH/s to 319.0 EH/s accounts for a disproportionate share of the network contraction. These companies historically represented roughly 30-35% of global hashrate. Their accelerated withdrawal has forced two difficulty reductions in 2026, including a 10% drop that marked the second-largest single adjustment of the year.
The decline is not uniform. Bitdeer increased its hashrate 44% to 63.0 EH/s, producing 990 BTC in June 2026 — a 388% year-over-year increase — driven by the deployment of its proprietary SEAL03 miners. In contrast, Cango's hashrate collapsed 63%, from 44.8 EH/s to an estimated 16.5 EH/s. Keel Infrastructure completed a full shutdown of all U.S. mining operations during Q2 2026.
The April 2024 halving reduced the block subsidy to 3.125 BTC, cutting miner revenue by half while operating costs continued to rise. CoinShares' Q1 2026 mining report placed the weighted average cash cost to produce one bitcoin at approximately $80,000. With BTC trading around $64,000 as of mid-August, that implies a loss of roughly $16,000 per coin for the average listed miner.
The profitability picture is hardware-dependent. At a standard industrial electricity rate of $0.07/kWh:
| Hardware | Efficiency (J/TH) | Approx. Cost/BTC | Margin at $64,000 | |---|---|---|---| | Bitmain S23 (hydro) | 9.5 | ~$32,000 | ~50% | | Bitmain S21 XP | 13.5 | ~$45,000 | ~30% | | Bitmain S21 (air) | 17.5 | ~$59,000 | ~8% | | Older fleet (>20 J/TH) | 20+ | ~$70,000+ | Negative |
The disparity explains the divergence in strategy. Operators with sub-15 J/TH fleets — primarily Bitdeer with its proprietary chips — remain profitable miners. Operators running mixed or older fleets face a binary choice: invest hundreds of millions in hardware upgrades or repurpose the power for higher-margin workloads.
Transaction fees offer no relief. Over the week ending August 10, miners collected approximately 3,158 BTC in block rewards totaling $202 million, but transaction fees constituted just 0.74% of that total — 23 BTC worth $1.49 million.
The aggregate value of AI and HPC contracts signed by listed miners exceeds $70 billion in base-term contracted revenue. The largest deals:
| Company | Counterparty | Capacity (MW) | Contract Value | Term | |---|---|---|---|---| | Hut 8 | Undisclosed (Phase 1 & 2) | 949 | $26.6B | 15 years | | TeraWulf | Anthropic | 401 | $19-33B | 15-25 years | | Core Scientific | AMD / Neocloud | 529 | $14B+ | 15 years | | Riot Platforms | AMD + Frontier AI lab | 636+ | $9.1B+ | 10 years |
These are not speculative commitments. Core Scientific is already billing 437 MW as of mid-July 2026, generating $136.7 million in colocation revenue in Q2 alone. TeraWulf's HPC lease revenue hit $31.9 million in Q2 2026, up 52% sequentially. The contracts carry 80-90% operating margins, according to CoinShares, compared to the compressed and often negative margins in mining.
The counterparty list reveals the demand source: AMD, Anthropic, CoreWeave, and undisclosed hyperscalers. These are not speculative crypto-native customers. They are well-capitalized technology companies with multi-year compute roadmaps and the ability to underwrite 10-15 year leases.
Core Scientific (CORZ): Total revenue doubled to $164.2 million from $78.6 million in Q2 2025. Colocation revenue reached $136.7 million — 83% of total sales, up from 67% in Q1 2026 and negligible a year prior. Mining revenue fell to $27.5 million. The company ended June with nearly 30% fewer miners online than at Q1 close and is now self-mining at only two sites. Gross margin on colocation was 59%. The company raised $3.3 billion in senior secured notes to fund expansion.
Marathon Digital (MARA): Revenue declined 27% year-over-year to $175 million, missing consensus estimates of $208 million. Net loss was $611 million. The company is pursuing a "Digital Infrastructure Triad" strategy — AI, power generation, and critical IT — via a Starwood partnership converting approximately 90% of non-hosted mining capacity to AI and compute sites. MARA targets at least two AI lease signings before year-end and projects 2027 revenue of $946 million.
Hut 8 (HUT): Revenue rose 81% year-over-year to $74.9 million. Gross margin expanded to 64% from 47%. The company's contracted AI portfolio reached $26.6 billion across 949 MW at Beacon Point and River Bend, with $7.5 billion in project-level financing secured. Initial energization at Beacon Point Phase 1 is targeted for Q1 2027. The company spun its mining operations into American Bitcoin, a separately managed entity.
TeraWulf (WULF): Total revenue was $44.8 million. HPC lease revenue was $31.9 million (71% of total), up 52% from Q1 2026. Cryptocurrency revenue dropped to $12.8 million from $47.6 million year-over-year. The Anthropic lease at its Kentucky facility covers 401 MW and represents approximately $19 billion in base-term revenue, with extensions potentially reaching $33 billion.
Riot Platforms (RIOT): Total revenue of $174 million, up 14%. Data center revenue reached $23.2 million, with the initial 25 MW delivered to AMD on time and on budget. AMD exercised its expansion option to 50 MW. Engineering revenue tripled to $37.3 million. Mining revenue was $113.7 million, still the dominant segment, but Riot's $9.1 billion AI lease signals the transition is underway.
The migration of listed miners away from Bitcoin produces several measurable network effects.
Difficulty compression. As hashrate exits, difficulty adjusts downward, improving economics for remaining miners. The August 8 adjustment was +0.99%, but the next adjustment on August 23 is projected at -3.07%. This self-correcting mechanism prevents hashrate declines from threatening network security, but it does redistribute mining revenue toward a smaller set of operators.
Geographic concentration risk. The U.S. holds 37.4% of global mining hashrate and Russia 16.9%. As U.S.-listed miners divert capacity, the relative share of non-public, geographically distributed miners may increase. This could affect U.S. policy arguments about domestic Bitcoin infrastructure security.
Hashrate floor. CoinShares projects hashrate could recover toward 1.8 ZH/s by year-end 2026 if Bitcoin's price recovers toward $100,000, which would restore positive unit economics for most hardware tiers. The current downcycle in hashrate appears cyclical rather than structural — the machines are not being destroyed, the power is being temporarily allocated to higher-margin workloads.
Energy consumption. The Cambridge Bitcoin Electricity Consumption Index estimated network power demand at 16.09 GW and annualized consumption at 141.02 TWh as of August 1, 2026. Per-coin energy intensity stands at approximately 858,000 kWh per BTC. A sustained hashrate decline could modestly reduce these figures, though efficiency gains in remaining hardware partially offset the reduction.
The sector is bifurcating along a clear economic line.
Full pivot: Core Scientific, TeraWulf, and Keel Infrastructure have effectively exited or are rapidly exiting Bitcoin mining. Core Scientific derives 83% of revenue from colocation. TeraWulf derives 71% from HPC leases. Keel shut down entirely.
Hybrid model: Hut 8, Marathon, and Riot are maintaining mining operations while building parallel AI infrastructure businesses. Hut 8 spun mining into American Bitcoin. Marathon's Starwood partnership is converting 90% of non-hosted capacity. Riot retains mining as its majority revenue source but has committed to $9.1 billion in AI leases.
Mining-first: Bitdeer stands alone as the major listed miner increasing hashrate and investing in proprietary ASIC development. Its vertically integrated model — designing chips, manufacturing miners, and operating mines — delivers lower production costs that remain viable at current BTC prices.
The pattern suggests the industry's future structure: a small number of vertically integrated, low-cost mining specialists coexisting with large power infrastructure companies that treat Bitcoin mining as one of several workloads, allocated dynamically based on relative profitability.
The listed Bitcoin mining industry is undergoing the most significant structural transformation since the China ban of 2021. But where the China exodus was a geographic migration that preserved the industry's core function, the current shift is an economic one. These companies are not relocating their mining operations — they are replacing them.
The arithmetic is straightforward. A 15-year lease with a hyperscaler generating 80-90% operating margins on committed capacity is a fundamentally different business than mining Bitcoin at negative unit margins while hoping for a price recovery. The companies that recognized this earliest — Core Scientific emerged from bankruptcy in 2024 and immediately pivoted — now hold contracted revenue pipelines in the tens of billions.
Bitcoin's network is not threatened by this migration. Difficulty adjustments ensure block production continues at roughly 10-minute intervals regardless of how many miners participate. The network's security model is price-dependent, not miner-count-dependent: a sufficiently high Bitcoin price will attract hashrate from any source.
What has changed is the identity of the companies doing the mining. The publicly traded "miner" may be an anachronism within 18 months. In its place: power infrastructure companies that allocate megawatts across Bitcoin, AI, and general compute based on real-time margin analysis. The economic value has migrated from proof-of-work to proof-of-power.