Bitcoin's network hashrate has spent 316 consecutive days below its late-October 2025 peak of approximately 1,150 EH/s, settling near 934 EH/s as of September 7, 2026 — a 19% decline. Mining difficulty has fallen 19.3% from its November 2025 high of 155.97 trillion to 127.45 trillion. Ten of 17 d...
"We're living, I think, through the first bear market in hash rate that we've ever seen in Bitcoin history." — Raphael Zagury, CEO, Twenty One Capital
Bitcoin's network hashrate has spent 316 consecutive days below its late-October 2025 peak of approximately 1,150 EH/s, settling near 934 EH/s as of September 7, 2026 — a 19% decline. Mining difficulty has fallen 19.3% from its November 2025 high of 155.97 trillion to 127.45 trillion. Ten of 17 difficulty adjustments in 2026 moved lower.
The cause is structural, not cyclical. Public Bitcoin miners cut realized hashrate by 15% in H1 2026 and redirected capacity toward artificial intelligence and high-performance computing infrastructure. With all-in production costs near $90,000 per BTC and spot prices around $79,700 in early September, pure-play mining margins remain negative for most operators. AI hosting contracts, by contrast, offer predictable multi-year revenue at higher margins per megawatt-hour. The result is the longest hashrate slump in Bitcoin's 17-year history, one that Twenty One Capital CEO Raphael Zagury described at Bitcoin Asia 2026 on August 28 as "the first bear market in hash rate."
Bitcoin's 7-day average hashrate peaked near 1,150 EH/s in late October 2025. By late August 2026 it sat at approximately 914–934 EH/s, a decline of roughly 19–22%, depending on the measurement window. On September 2, a snapshot showed 829 EH/s; the 7-day smoothed figure recovered to 934 EH/s by September 7.
Mining difficulty tells the same story. From the November 2025 record of 155.97 trillion, difficulty declined through ten of 17 retargets in 2026 — net cumulative moves of +33.34% upward versus -45.27% downward. The most notable single drop came in June: a 10.09% reduction at block 953,568, triggered by a price-driven miner shutdown cycle. The most recent adjustment on September 6, 2026 at block 965,664 moved difficulty up 1.31% to 127.45 trillion, the first sign of stabilization after months of declines.
Year-to-date, hashrate is down approximately 4%. This marks the first first-quarter decline since 2020 and breaks five consecutive years of double-digit growth.
Key difficulty events in 2026:
| Date | Block | Change | Notes | |------|-------|--------|-------| | Feb 7 | — | -11.16% | Largest single drop | | Feb 19 | — | +14.7% | Sharp recovery 12 days later | | Jun 14 | 953,568 | -10.09% | Second-largest 2026 decline | | Sep 6 | 965,664 | +1.31% | First increase after sustained drops |
The April 2024 halving cut the block subsidy from 6.25 BTC to 3.125 BTC. At September 2026 spot prices near $79,700, the block subsidy generates approximately $249,000 per block. Fees contribute just 0.43% of total miner revenue, per Cryptolexicon data from September 7, making Bitcoin mining almost entirely subsidy-dependent.
Hashprice — revenue per petahash per second per day — stood at $39.63 on September 7, up 22.24% over the prior 30 days from $32.42. The gain was driven almost entirely by BTC's price recovery from sub-$70,000 levels in August, not by fee growth or hashrate contraction alone.
The break-even equation is straightforward: a 200 TH/s machine drawing 3.5 kW at $0.06/kWh generates roughly $2.89 per day after electricity, before pool fees and depreciation. At $0.10/kWh, margins compress to near zero. At the U.S. residential average of $0.16–$0.20/kWh, the operation runs at a loss.
Industrial miners operating modern hardware (sub-15 J/TH) at power rates below $0.08/kWh maintain 20–50% margins. Everyone else is losing money. The efficiency gap between the Antminer S19 (34.2 J/TH) and the S21 XP (13.5 J/TH) represents a 60% improvement — older-generation hardware is economically obsolete at current prices.
The shift from mining to AI hosting is not theoretical. It is occurring across the majority of publicly listed Bitcoin miners. According to insights4vc research updated through July 2026, the sector's capital expenditure increased 400% between March 2025 and February 2026.
IREN (Iris Energy): Reduced mining capacity from approximately 50 EH/s to 23.2 EH/s as of June 30, 2026. Deployed 11,000 of 23,000 contracted GPUs. Operational capacity: 810 MW, with 2,100 MW under construction. Signed a $1.9 billion Microsoft prepayment and a $3.4 billion cloud contract with Nvidia. AI clients include Cohere, Perplexity, and Figure AI. Booked $638.8 million in non-cash impairments on decommissioned mining hardware. Market cap: $14 billion as of February 2026. AI annualized revenue run-rate: $4 billion for 2026.
Core Scientific (CORZ): Power portfolio: 1.2 GW. Added 400 MW of new AI data center capacity. HPC revenue projected to reach 71% of total, up from 5% in 2024. LTM revenue: $510.67 million.
TeraWulf (WULF): Contracted 510 MW of HPC capacity. Signed a $9.5 billion joint venture with Fluidstack/Google for a 168 MW Texas project. Google holds a 14% equity stake with a $3.2 billion backstop. Q1 2026 AI/HPC revenue of $21 million exceeded Bitcoin mining revenue for the first time.
Riot Platforms (RIOT): Power capacity: 1.7 GW. Signed a 10-year AMD lease for 25 MW (expandable to 200 MW), projecting $311 million in initial revenue and $1 billion if fully extended. Retrofit budget: $89.8 million. Sold 1,080 BTC to fund land acquisition ($96 million).
Cipher Mining (CIFR): Signed a 15-year AWS lease for 300 MW, projecting $5.5 billion in revenue. Issued $1.3 billion in convertible notes at 0% interest. Phase I delivery target: 168 MW by September 2026.
CleanSpark (CLSK): Holds 1.5 GW of power access but stated in Q1 2026 that "Bitcoin mining investment doesn't make a lot of sense" at current hashprices compared to AI lease returns. Targeting 285 MW for Texas AI site. FY2025 revenue: $766 million. BTC holdings: 13,099 units.
Hut 8 (HUT): Signed a $7 billion, 15-year Fluidstack lease for 245 MW in Louisiana. Manages 1,020 MW, with 1,230 MW under development across 19 sites.
The aggregate figures are large. Public mining firms sold more than 32,000 BTC in Q1 2026 to fund the transition. Industry-wide AI and HPC contracts are estimated at $70–$90 billion. Build costs for AI-ready facility conversions run $8–$11 million per MW, though Applied Digital achieved $3.6 million per MW on retrofit projects.
AI revenue among transitioning miners rose 52% in H1 2026. Mining revenue, by contrast, is projected to fall from 85% of total revenue in early 2025 to below 20% by end-2026 for companies with secured AI contracts.
The economic logic is clear: AI hosting generates predictable, contractually locked revenue for 10–20 year terms, denominated in dollars, at higher margins per megawatt-hour than Bitcoin mining. Once power and cooling infrastructure is committed to an AI tenant, it does not return to Bitcoin mining when BTC prices rise — creating a structural floor under the hashrate decline.
The security question is straightforward. Bitcoin's proof-of-work security is a function of total hashrate. A 19–22% decline from peak reduces the cost of a theoretical 51% attack proportionally.
However, context matters. At 934 EH/s, the network still commands extraordinary computational commitment — roughly 934 quintillion SHA-256 hashes per second. This is higher than any level recorded before October 2024. The protocol's difficulty adjustment mechanism, which recalibrates every 2,016 blocks (approximately every two weeks), ensures that block times remain near 10 minutes regardless of hashrate fluctuations.
The more material risk is geographic concentration. The January 2026 Winter Storm Fern caused hashrate to plunge from 1.1 ZH/s to 663 EH/s over a single weekend — a 30–40% drop — as U.S.-based mining operations lost power. This exposed dependence on a single geography for a disproportionate share of global hashrate, a vulnerability that the AI pivot may exacerbate as U.S. miners redirect infrastructure.
As of September 7, 2026:
| Metric | Value | |--------|-------| | Network hashrate (7-day avg) | ~934 EH/s | | Mining difficulty | 127.45 T | | Hashprice | $39.63/PH/s/day | | BTC spot price | ~$79,700 | | Block subsidy | 3.125 BTC (~$249K) | | Fee share of block reward | 0.43% | | Break-even electricity rate (modern hardware) | ~$0.08/kWh | | Break-even electricity rate (legacy S19) | ~$0.03/kWh | | YTD difficulty adjustments down | 10 of 17 |
The Bitcoin mining industry is undergoing its most significant structural transformation since the 2021 China ban. The difference: in 2021, hashrate migrated to new geographies and recovered within six months. In 2026, hashrate is not migrating — it is being replaced by a higher-value use case.
The AI pivot is rational at current economics. A megawatt committed to AI hosting earns more than a megawatt committed to SHA-256 computation at $79,700 BTC. The question is what happens if Bitcoin prices rise substantially. Long-duration AI contracts, typically 10–20 years, lock in infrastructure commitments that cannot easily revert to mining. The traditional hashrate recovery cycle — price rises, miners return, difficulty increases — may be permanently altered.
For Bitcoin, 934 EH/s remains a formidable level of computational security. The protocol does not require hashrate growth to function. But the market structure that produced fifteen years of near-continuous hashrate expansion has changed. Computing infrastructure now has a competing buyer with deeper pockets and more predictable demand than a volatility-prone digital asset.