Bitcoin mining difficulty has fallen 19.9% from its November 2025 peak, marking only the second year-over-year decline in the network's history. The first occurred in mid-2021 when China banned mining. This time, the cause is not a government ban but an economic one: AI infrastructure now pays 3x...
"As access to power becomes the defining constraint on AI infrastructure development, we believe our ability to combine energy expertise, infrastructure control, and execution at scale will become increasingly valuable." — Paul Prager, CEO, TeraWulf
Bitcoin mining difficulty has fallen 19.9% from its November 2025 peak, marking only the second year-over-year decline in the network's history. The first occurred in mid-2021 when China banned mining. This time, the cause is not a government ban but an economic one: AI infrastructure now pays 3x to 25x more per kilowatt than Bitcoin mining, and capital is following the margin.
The network hashrate has dropped from a peak of 1,066 EH/s in December 2025 to approximately 940 EH/s as of August 3, 2026 — a 12% contraction. Hashprice stands at $32 per PH/s per day, below breakeven for 15–20% of the global mining fleet according to CoinShares. Public miners have collectively signed over $70 billion in AI and high-performance computing contracts, and industry analysts project that 70% of public miner revenue will come from AI workloads by December 2026, up from roughly 30% at the start of the year.
The aggregate miner Bitcoin reserve has declined from 1.86 million BTC at the end of 2023 to roughly 1.8 million BTC by mid-2026, as operators liquidate holdings to fund data center conversions. Bitdeer reduced its treasury to zero in February 2026. This is not a temporary disruption. It is a structural repricing of what Bitcoin mining infrastructure is worth.
Bitcoin's mining difficulty currently sits at 126.23 trillion, approximately 14% below this year's high and 19.9% below the all-time peak set in November 2025. The next adjustment, estimated at +0.67%, is expected on August 8, 2026.
This is the second time in Bitcoin's history that difficulty has registered a year-over-year decline. According to Hashrate Index, the first instance occurred in mid-2021 following China's mining ban, which removed roughly half the network's computing power overnight. Difficulty recovered to new highs within months as miners relocated.
The 2026 contraction differs in character. There is no single policy shock. Instead, multiple structural pressures have converged:
| Metric | Peak (Late 2025) | Current (Aug 3, 2026) | Change | |---|---|---|---| | Difficulty | 155.6T (est.) | 126.23T | -19.9% | | Hashrate (30-day avg) | 1,066 EH/s | 940 EH/s | -11.8% | | Hashprice | ~$42/PH/s/day | $32.10/PH/s/day | -23.6% | | BTC Price | ~$86,400 | ~$63,160 | -26.9% YTD |
The difficulty has experienced three major downward adjustments in 2026: -11.16% on February 7, -7.76% on March 21, and -10.09% in June. Each adjustment reflected sustained periods of above-target block times, with the February adjustment following average block intervals of 11.4 minutes.
Hashrate Index attributed the 2026 contraction to five overlapping forces:
1. Post-Halving Economics. The April 2024 halving cut the block subsidy from 6.25 BTC to 3.125 BTC. At current prices (~$63,160), each block yields approximately $197,375 in subsidy revenue. Transaction fees contributed 0.77% of total block rewards in the week ending August 3 — roughly $1.57 million across all blocks, or 25 BTC.
2. AI/HPC Capital Competition. AI data center demand has created a structural competitor for the power, land, and capital that previously flowed to SHA-256 mining. Public miners have retrofitted sites or reserved power capacity for GPU workloads, reducing potential Bitcoin hashrate even when the underlying facilities remain operational.
3. Texas Power Curtailments. Severe winter storms in late January 2026 forced Texas-based miners offline to protect the grid. The network hashrate fell 30–40% during that period, hitting a seven-month low of 663 EH/s, with some US pools losing up to 60% of capacity.
4. Iran Disruption. Internet restrictions and enforcement against unauthorized mining farms removed an estimated 7 EH/s from the network. Approximately 1,300 megawatts of baseload electricity demand disappeared from Iran's grid in January 2026.
5. Bitcoin Price Decline. BTC has fallen 26.9% year-to-date, from approximately $86,400 at the start of 2026 to $63,160 as of August 5. The price decline compounds the post-halving subsidy cut, creating a double hit to miner revenue denominated in fiat terms.
CoinShares' Q1 2026 Bitcoin Mining Report documented the severity of the margin compression:
Profitability is now a function of hardware vintage and electricity cost. At current hashprice levels:
| Hardware | Efficiency (J/TH) | Breakeven Electricity Rate | |---|---|---| | Antminer S23 Hydro | 9.5 | ~$0.124/kWh | | Antminer S21 XP | 13.5 | ~$0.088/kWh | | Antminer S19 XP | 21.5 | ~$0.055/kWh | | S19-class (older) | 30+ | Unprofitable at most rates |
Operations running S19-class machines at electricity costs above $0.055/kWh are generating negative margins. These units are being shut down, sold, or scrapped — contributing to the ongoing hashrate decline.
The economic logic is straightforward. AI infrastructure generates 3x to 25x more revenue per kilowatt than Bitcoin mining, with profit margins between 80% and 90%, compared to 60% gross margins for Bitcoin mining at current prices. Bitcoin miners already own the three assets AI companies need: large-scale power infrastructure, physical data center space, and cooling systems.
The scale of the pivot is now quantifiable. Public miners have announced more than $70 billion in cumulative AI and HPC contracts:
| Company | AI/HPC Partner | Capacity | Contract Value | Term | |---|---|---|---|---| | Cipher Mining | Amazon Web Services | 300 MW | $5.5B | 15 years | | TeraWulf | Anthropic / Core42 | 401 MW | ~$19B | 20 years | | IREN | Microsoft | Undisclosed | ~$3.7B (est. annualized) | Multi-year | | Core Scientific | Multiple | Expanding | $3.3B capital raise | Ongoing |
TeraWulf reported Q2 2026 revenue of $44.8 million, with 71% ($31.9 million) generated from HPC workloads. The company is on track to exit Bitcoin mining entirely by year-end.
Analyst projections from S&P Global and CoinShares indicate that AI/HPC will account for approximately 70% of public miner revenue by December 2026. IREN's HPC revenue share is expected to surge from 3% (2024) to 71%. Core Scientific's is projected to move from 5% to 71%.
The transformation is structural, not cyclical. These are 10- to 25-year lease agreements backed by counterparties including AWS, Microsoft, Google, and Anthropic. Once a mining facility converts to AI hosting, it does not convert back.
The AI pivot requires capital. Miners are funding conversions in part by liquidating their Bitcoin reserves.
Bitdeer provides the most complete case study. The company began 2026 with roughly 2,000 BTC. By the end of January, holdings had fallen to 1,530 BTC. By February 13, the figure stood at 943.1 BTC. On February 22, Bitdeer announced its corporate Bitcoin holdings had reached zero — it had mined 184 BTC that day and sold the entire batch along with the remaining reserves.
The trend extends across the sector. The aggregate miner reserve — the total Bitcoin held by mining companies — has declined from more than 1.86 million BTC at the end of 2023 to roughly 1.8 million BTC by mid-2026. According to crypto.news, this sustained drawdown reflects not opportunistic selling but a structural shift, driven by the debt burden many miners accumulated during the 2024–2025 expansion cycle.
Core Scientific plans to sell most of its Bitcoin holdings to fund AI expansion. Cango, Riot Platforms, and Bitfarms are also reallocating capital toward AI infrastructure.
The hashrate decline raises questions about Bitcoin's security model, though the situation requires context.
At 940 EH/s, the Bitcoin network remains, by any standard, the most computationally secured blockchain. Even after the 12% decline from peak, the cost of a 51% attack remains prohibitively expensive — estimated in the billions of dollars for sustained hardware and electricity costs.
However, the economic composition of hashrate is shifting. As public miners redirect capacity to AI, the remaining hashrate is increasingly provided by private, geographically concentrated operators in jurisdictions with low electricity costs. This makes the network's hashrate distribution less transparent and potentially less decentralized.
The foundational concern, consistent with webthreepedia's economic value framework, is that Bitcoin mining remains a subsidy-dependent operation. The network generates approximately $115 million annually in transaction fee revenue while requiring an estimated $54–72 billion annually in miner subsidies (via block rewards and electricity costs) to secure it. The AI pivot does not change this equation — it simply demonstrates that the market has found a more productive use for the same infrastructure.
The 2026 Bitcoin mining contraction is qualitatively different from previous episodes. The 2021 China ban was a policy shock followed by rapid recovery. The current decline is an economic verdict: AI workloads generate higher returns per megawatt, per dollar of capital, and per unit of management attention than Bitcoin mining.
The data does not support a narrative of imminent collapse. Bitcoin's network remains secured by more hashpower than it had at any point before late 2025. Difficulty adjustments are functioning as designed, recalibrating the network to lower hashrate. Miners with modern hardware and cheap power remain profitable.
What the data does show is that Bitcoin mining as a standalone business model is becoming a minority use case for the infrastructure it created. The companies that built mining capacity at scale are now deriving the majority of their revenue from hosting AI workloads. The implications for Bitcoin's long-term security budget — still overwhelmingly dependent on block subsidies rather than transaction fees — remain unresolved.