Bitcoin's network hashrate has spent 316 consecutive days below its October 2025 peak of 1,151.6 EH/s — the longest sustained decline in a decade. The seven-day average sat near 914 EH/s on August 31, 2026, roughly 20.6% below the record. Mining difficulty has dropped 14% from this year's high. H...
"We're living, I think, through the first bear market in hash rate that we've ever seen in Bitcoin history." — Rapha Zagury, CEO, Twenty One Capital
Bitcoin's network hashrate has spent 316 consecutive days below its October 2025 peak of 1,151.6 EH/s — the longest sustained decline in a decade. The seven-day average sat near 914 EH/s on August 31, 2026, roughly 20.6% below the record. Mining difficulty has dropped 14% from this year's high. Hashprice, the dollar revenue per petahash per day, recorded its three lowest monthly averages ever in 2026: $30.37 in June, $31.21 in July, and $31.27 in March.
The cause is not regulatory crackdown or geopolitical disruption, as it was during China's 2021 mining ban. This time, publicly listed Bitcoin miners are voluntarily redirecting power, capital, and facilities toward artificial intelligence infrastructure. Over the course of 2025 and early 2026, listed miners signed more than $70 billion in aggregate AI and HPC co-location contracts with hyperscalers, according to CoinShares. The firm projects AI and HPC work could supply 70% of listed miners' revenue by year-end 2026, up from approximately 30% at the start of the year.
Twenty One Capital CEO Rapha Zagury coined the term "hashrate bear market" during a keynote at Bitcoin Asia 2026 in Hong Kong on August 28, characterizing a structural condition he described as unprecedented in Bitcoin's 17-year history. The company subsequently filed the prepared transcript with the U.S. Securities and Exchange Commission.
The data supports the framing. Bitcoin's hashrate peaked at 1,151.6 EH/s in October 2025. As of August 31, 2026, the seven-day moving average registered approximately 914 EH/s. A single reading on September 4 briefly touched 1,001 EH/s, but the sustained average remains well below the record. Mining difficulty stood at approximately 124.6 trillion as of the most recent adjustment on September 5, down from highs above 145 trillion earlier in the cycle.
Previous hashrate declines — China's 2021 ban wiped out roughly 50% of capacity — were followed by rapid recoveries as displaced miners relocated and plugged back in. CryptoSlate and CoinShares analysts note this cycle is structurally different: miners are not relocating mining equipment. They are decommissioning it and replacing it with GPU racks.
The arithmetic driving the transition is straightforward. Post-halving Bitcoin block rewards stand at 3.125 BTC. Hashprice — the standard measure of mining revenue per unit of computational power — fell to $27.66 per petahash per day in late June 2026, within one cent of its February floor. The most efficient air-cooled ASICs now operate at approximately 13.5 joules per terahash, rendering older hardware uneconomic at current revenue levels.
AI infrastructure offers a fundamentally different revenue model. Instead of volatile, subsidy-dependent returns, miners sign fixed-rate leases spanning 12 to 20 years. A megawatt dedicated to AI colocation generates predictable, contracted cash flow with margins that CoinShares estimates at 80–85% for colocation services, compared to the compressed and fluctuating margins of Bitcoin mining.
Miners possess what the AI economy currently lacks: energized, grid-connected, power-dense facilities in the United States. New data center construction faces 3–5 year permitting and grid interconnection timelines. Miners already have the permits, the power purchase agreements, and the physical infrastructure. The conversion requires replacing ASIC rigs with GPU clusters and upgrading cooling systems — a capital expenditure measured in months, not years.
According to a CoinDesk report from March 2026, miners are also selling Bitcoin reserves to finance the conversion. Core Scientific sold approximately $175 million in Bitcoin to accelerate its AI infrastructure buildout.
The scale of individual company transitions illustrates the depth of the shift:
Core Scientific (CORZ): Having emerged from bankruptcy in January 2024, Core Scientific signed an initial $10.2 billion, 12-year agreement with CoreWeave covering approximately 590 MW. On July 28, 2026, the company added a 15-year deal with AMD for up to 2.5 GW of energy capacity, valued at over $14 billion in potential revenue. Q2 2026 total revenue reached $164.2 million, with AI colocation contributing $136.7 million — 83% of the total — at a 59% gross margin. Total leased customer power capacity reached 1.1 GW, representing over $24 billion in aggregate contracted revenue.
Riot Platforms (RIOT): On August 11, 2026, Riot signed a $9.1 billion, 20-year computing capacity lease with Anthropic covering 191 MW at its Rockdale, Texas campus. The deal could reach $16.1 billion if extended for two additional five-year periods. Riot was originally a biotech diagnostics company (Bioptix), pivoted to Bitcoin mining, and is now pivoting again to cloud computing infrastructure.
TeraWulf (WULF): HPC revenue surpassed Bitcoin mining revenue for the first time in Q1 2026, with 60 MW of operating AI capacity generating $21 million in lease revenue versus $13 million from Bitcoin mining. In July 2026, TeraWulf signed a 20-year lease with Anthropic to provide approximately 401 MW at its Justified Data campus in Hawesville, Kentucky, starting in 2027. Expected revenue: approximately $19 billion over the initial term. The company disclosed nearly 600 MW of HPC capacity under contract and announced plans for a 1 GW facility in Kentucky.
IREN (formerly Iris Energy): Cut installed self-mining capacity from 50 EH/s in June 2025 to 23.2 EH/s by June 2026. Secured a $9.7 billion deal with Microsoft for 76,000 NVIDIA GB300 GPUs across 200 MW at its Childress, Texas campus. For full fiscal year 2026, total revenue hit $707 million. Quarterly AI cloud services revenue of $70.5 million exceeded Bitcoin mining revenue of $66.7 million. IREN completed a $3 billion convertible notes deal to fund the transition.
Hut 8 (HUT): Fully commercialized its 1 GW Beacon Point campus in Texas through a $9.8 billion, 15-year lease, pushing total contracted portfolio value to $26.6 billion.
According to CoinShares' Q1 2026 Bitcoin Mining Report, listed miners have signed over $70 billion in aggregate GPU co-location and cloud service agreements. The contracts span 12–20 year terms and involve counterparties including CoreWeave, Anthropic, Microsoft, AMD, Core42, and Fluidstack.
The concentration is notable. Five companies — Core Scientific, Hut 8, TeraWulf, Riot Platforms, and IREN — account for the vast majority of contracted value. Hut 8 leads at $26.6 billion, followed by Core Scientific at $24 billion-plus, TeraWulf at $19 billion, Riot at $9.1–16.1 billion, and IREN at $9.7 billion.
The demand side is driven by a structural shortage of U.S. data center capacity. AI model training and inference require power-dense facilities that the market cannot build fast enough. According to CoinShares, U.S. data center grid bottlenecks could further accelerate the reallocation of mining capacity.
A significant gap exists between announced contract values and realized revenue. According to Noah Intelligence, listed miners generated $341.2 million in AI-related revenue during the first half of 2026, against $5.11 billion in cumulative AI infrastructure spending. The $70 billion in announced contracts represents lifetime revenue potential spanning 12–20 years, not current run-rate.
Core Scientific's case is illustrative. While the company reports $24 billion in total leased contracted value, Q2 2026 colocation revenue was $136.7 million — an annualized rate of approximately $547 million. The company logged a $1.2 billion loss as infrastructure buildout costs front-load against back-loaded revenue streams.
The pattern repeats across the sector: capital expenditure today, contracted revenue over decades. Whether these contracts will generate the projected returns depends on continued AI compute demand growth, counterparty credit risk over multi-decade terms, and grid capacity remaining constrained enough to sustain pricing power.
The hashrate decline raises questions about Bitcoin's security model, though context is required. At 914 EH/s, Bitcoin's network remains the most computationally secured blockchain by orders of magnitude. The decline is relative — from a record high — not absolute. Network hashrate is still multiples above 2021 levels.
However, the structural nature of the decline is novel. Previous drawdowns were cyclical (price-driven) or event-driven (regulatory bans). The current decline reflects a permanent reallocation of capital and infrastructure away from Bitcoin mining. If AI compute demand continues to grow and mining economics remain compressed, some fraction of the departed hashrate may never return.
The difficulty adjustment mechanism ensures the network continues to function: lower hashrate leads to lower difficulty, which eventually restores profitability for remaining miners. But the adjustment does not address the broader question of whether the mining industry — once the primary occupant of power-dense data centers — is ceding that role permanently.
The Bitcoin mining industry is undergoing a structural transformation that has no precedent in the network's history. Listed miners are not diversifying into AI as a hedge — they are converting core operations. The economic logic is clear: fixed-rate, multi-decade contracts with hyperscalers offer margin profiles that post-halving Bitcoin mining cannot match at current hashprice levels.
The $70 billion in announced contracts signals market confidence in the model, but the gap between contracted value and realized revenue warrants scrutiny. Front-loaded capital expenditure, multi-decade counterparty risk, and dependence on sustained AI compute demand growth all represent material uncertainties.
For Bitcoin's network, the security implications are manageable at current levels but structurally different from prior cycles. The hashrate may stabilize at a lower equilibrium rather than recover to prior peaks — a scenario that would represent a permanent shift in the relationship between Bitcoin mining economics and data center infrastructure markets.