Public Bitcoin miners cut realized hashrate 13.4% between Q4 2025 and Q2 2026, according to BlocksBridge Consulting. Excluding Bitdeer, the decline was 21.2%. The network's 30-day average hashrate sits at approximately 934 EH/s as of September 7, 2026 — roughly 20% below the October 2025 peak of ...
"Every major US-listed Bitcoin miner has moved toward AI data centers in some form." — Bernstein Research
Public Bitcoin miners cut realized hashrate 13.4% between Q4 2025 and Q2 2026, according to BlocksBridge Consulting. Excluding Bitdeer, the decline was 21.2%. The network's 30-day average hashrate sits at approximately 934 EH/s as of September 7, 2026 — roughly 20% below the October 2025 peak of 1,151 EH/s. The gap has persisted for 316 consecutive days, the longest sub-peak streak in a decade.
The cause is not a mining crisis in the traditional sense. It is a capital reallocation. Former Bitcoin miners have signed over $50 billion in aggregate AI and HPC data center lease contracts with hyperscalers and frontier AI labs in 2026 alone. Riot Platforms, Hut 8, Core Scientific, Cipher Mining, IREN, and TeraWulf have each executed multibillion-dollar, decade-plus lease agreements. CoinShares projects AI and HPC could supply 70% of listed miners' revenue by year-end 2026. The Bitcoin mining sector is becoming the AI data center sector — with mining as a side business.
Bitcoin's smoothed hashrate peaked near 1,151 EH/s in late October 2025. As of early September 2026, the 30-day average sits at approximately 934 EH/s, according to CoinWarz data at block 966,350. That is a 19% deficit sustained over 316 days — the longest period the network has operated below a prior peak since the 2014-2015 mining downturn.
The decline is structural, not cyclical. BlocksBridge Consulting's cohort of public miners reported realized hashrate of 319 EH/s in Q2 2026, down from 368.3 EH/s in Q4 2025. Excluding Bitdeer — which increased its hashrate 44% to 63 EH/s — the remaining cohort dropped 21.2%, from 324.6 EH/s to 255.9 EH/s.
Year-to-date, network difficulty has declined from 146.47 trillion to 127.45 trillion as of September 6, 2026 — a 13% reduction across 18 distinct difficulty adjustments. The September 6 adjustment at block 965,664 added 1.31% to difficulty, the first upward move in several periods.
Hashprice — the dollar-denominated revenue per petahash per day — climbed from $32.42 to $39.63 per PH/s/day in early September 2026, a 22% increase. That jump was driven primarily by a Bitcoin price recovery rather than hashrate expansion.
Bitcoin traded at approximately $78,137 on September 9, 2026. CoinShares' Q1 2026 mining report estimated all-in production costs near $90,000 per bitcoin for the average listed miner. At current spot, that implies negative operating margins for all but the most efficient operators.
The April 2024 halving cut block rewards from 6.25 BTC to 3.125 BTC. With network fees failing to compensate, per-block revenue fell sharply. Hashprice at roughly $39.63/PH/s/day remains well below pre-halving levels and is comparable in real terms to the post-COVID trough of mid-2020, when hashprice sat around $70/PH/s/day before the network was a fraction of its current size.
The economics are binary: miners either operate next-generation ASICs below 15 J/TH (joules per terahash) and have access to sub-$0.04/kWh power, or they lose money on every bitcoin produced. The Cambridge Centre for Alternative Finance estimates Bitcoin mining now consumes approximately 138 TWh annually, a 17% year-over-year increase that accounts for roughly 0.54% of global electricity consumption. The network's average efficiency of 28 J/TH confirms that a large share of deployed hardware remains unprofitable at current prices.
The scale of AI infrastructure contracts signed by former mining companies in 2026 has no precedent in the crypto sector:
| Company | Counterparty | Capacity | Term | Contract Value | |---------|-------------|----------|------|---------------| | Core Scientific | AMD | 530 MW (options to 2.5 GW) | 15 years | $14B base | | Riot Platforms | Anthropic | 191 MW | 20 years | $9.1B | | Hut 8 | Undisclosed | Phase 2, Beacon Point TX | 15 years | $9.8B | | Cipher Mining | AWS, Google, Fluidstack | 300 MW+ | 10-15 years | $8.5B combined | | IREN | Various | Transitioning by Dec 2026 | Various | $4B ARR target |
Riot's Anthropic deal at its Rockdale, Texas campus will repurpose 191 MW of critical IT capacity, generating an estimated $9.1 billion over the initial 20-year term. Core Scientific's AMD partnership covers approximately 530 MW across five sites, with AMD holding options on up to 1.9 GW of additional capacity through 2028 — lifting the total potential to 2.5 GW. Cipher Mining's combined leases with AWS, Google, and Fluidstack total $8.5 billion, with the AWS component alone covering a 15-year, 300-MW lease valued at $5.5 billion. Hut 8's $9.8 billion deal covers the second phase of its Beacon Point AI data center campus in Texas.
Combined contracted revenue across these and smaller deals exceeds $50 billion, according to aggregate filings and press releases.
The quarterly revenue crossover is already occurring at the company level:
IREN reported FY26 results (year ended June 30, 2026) showing total revenue of $707.0 million, up from $501.0 million in FY25. AI Cloud Services revenue grew nearly 8x year-over-year to $128.8 million. In Q4 FY26 specifically, AI Cloud Services generated $70.5 million versus Bitcoin mining's $66.7 million — the first quarter in which AI revenue exceeded mining revenue. Three months earlier, the figures were $33.6 million (AI) and $111.2 million (mining). IREN targets substantially completing its mining-to-AI transition by December 31, 2026.
Core Scientific generated $136.7 million in colocation revenue during Q2 2026, representing 83% of its $164.2 million in total quarterly revenue. Self-mining revenue fell 66% to $21.5 million. The company was billing customers for 437 MW as of mid-July, equivalent to approximately $635 million in annualized colocation revenue.
Riot Platforms has begun booking revenue from its AMD lease at Rockdale, its first data center revenue entry.
CoinShares projects that mining revenue across listed miners could decline from approximately 85% of total revenue in early 2025 to less than 20% by end of 2026 for companies that have secured AI contracts. The firms generating 80% to 90% of their revenue from AI deals would effectively be data center operators that also mine bitcoin.
The transition carries significant near-term capital costs. According to Cointelegraph, nine comparable miners spent $5.11 billion on capital assets during H1 2026 while generating just $341.2 million in directly reported AI and HPC revenue — a roughly 15-to-1 capex-to-revenue ratio.
A broader group of 15 Bitcoin miners and AI data-center companies spent a combined $30.7 billion on capital assets in their latest 2026 reporting periods, already 42.6% more than the $21.53 billion they spent throughout all of 2025. This acceleration reflects the infrastructure build-out required to meet hyperscaler specifications: liquid cooling, redundant power, Tier III+ uptime guarantees, and GPU-optimized rack densities that differ substantially from ASIC mining configurations.
IREN's FY26 net loss of $702.6 million — driven by $638.8 million in non-cash impairments and transition costs — illustrates the accounting impact. Revenue growth exists but is back-loaded behind capital deployment.
The capital demands of the AI pivot have accelerated BTC sales. Public mining companies sold a record 32,000 BTC in Q1 2026, more than they sold in all four quarters of 2025 combined, according to CoinDesk reporting. Core Scientific sold approximately 1,900 BTC (~$175 million) in January 2026 alone and disclosed plans to liquidate substantially all remaining holdings in Q1.
Publicly listed miners have collectively reduced their BTC treasuries by over 15,000 BTC from peak levels. The sales are not distressed liquidations in the traditional mining-capitulation sense; they are strategic funding events for AI infrastructure build-outs that require billions in upfront capital before lease revenue begins flowing.
The migration of large US-listed miners away from bitcoin mining may paradoxically improve the network's decentralization. As major operators redirect megawatts to AI, a larger share of hashrate shifts to smaller, geographically dispersed operators — particularly in regions with cheaper power where mining remains marginally profitable.
The 13% year-to-date difficulty decline has made mining accessible to operators running newer-generation hardware at favorable power rates. Bitdeer's 44% hashrate increase demonstrates that not all public miners are exiting: those with proprietary ASIC development (Bitdeer manufactures its own chips) can still compete.
However, the sustained hashrate deficit raises a longer-term question about network security budgets. If block rewards continue declining through future halvings and transaction fees remain insufficient compensation, the economic incentive to mine — even as a side business for data center operators — narrows further.
The Bitcoin mining industry's transformation into an AI data center sector is no longer a thesis — it is an accounting reality. IREN's quarterly revenue crossover, Core Scientific's 83% colocation revenue share, and Riot's $9.1 billion Anthropic lease are execution milestones, not announcements. The economic logic is straightforward: with all-in Bitcoin production costs near $90,000 and spot at $78,000, the same megawatt-hour generates higher and more predictable returns hosting GPU clusters for hyperscalers than running SHA-256 ASICs.
The 15-to-1 capex-to-revenue ratio indicates this transition has significant execution risk. Multi-year lease revenue ramps, construction timelines, and counterparty concentration create exposures that did not exist in the simpler mining-only model. Whether these companies can deliver on $50 billion in contracted revenue while managing the physical build-out remains the central question for the sector through 2027.
For the Bitcoin network itself, the migration is a test of the security budget model. The hashrate gap may close as difficulty adjusts and smaller miners fill the void, but the era of public companies treating Bitcoin mining as their primary business is ending.