Bitcoin's network hashrate has fallen 21% from its 2026 peak, dropping from 1.14 zettahash per second to 900 exahash per second, as public mining companies redirect gigawatts of contracted power toward AI and high-performance computing instead of block production.
The Difficulty Retreat
!Bitcoin Hashrate & Hash Ribbon Source: CoinMetrics · Hashrate + 30D/60D Hash Ribbon · Trailing 12 months
Bitcoin's mining difficulty fell below its year-ago level on August 1, 2026, only the second time that has happened in the network's history. Difficulty currently sits at 127.48 trillion, roughly 14% below this year's high of 148 trillion. The next retarget is estimated for August 22, 2026, with 593 blocks remaining and a projected increase of just 0.85%, a sign the bleeding has slowed but not reversed.
The first year-over-year difficulty decline came during the 2022 bear market, when miner bankruptcies forced hardware offline. This time the driver is different. Miners are not going bankrupt. They are choosing to walk away from block production because something else pays better right now.
Why Miners Are Walking Away From Blocks
Combined realized hashrate across the tracked public miner cohort fell from 368.3 EH/s in Q4 2025 to 344.4 EH/s in Q1 2026 and 319.0 EH/s in Q2 2026, a 13.4% decline over two quarters. These are not idle machines. Operators including MARA and Riot have retrofitted sites and reserved electrical capacity for AI and HPC clients, work that generates steadier revenue per megawatt than mining at current network economics.
The timing lines up with price. Bitcoin trades near $64,400, down 48.9% from its $126,000 all-time high set in October 2025, and has been range-bound between $62,000 and $66,800 for more than a week. Falling coin-denominated block rewards, combined with a price that has not recovered, squeezed margins enough that AI contracts became the better trade for capital already sitting on power infrastructure.
The Risk Nobody Is Pricing
A structural hashrate exodus is not free of risk to the network. Lower hashrate means lower absolute security spend, even if difficulty adjusts to compensate. If AI demand for compute cools before Bitcoin's price recovers enough to make mining competitive again, some of this capacity may not return quickly. Miners who repurposed cooling and substation infrastructure for GPU racks face real switching costs to reverse course.
There is a bullish read too. Puell multiple compression during this stretch, driven by falling miner revenue relative to its 365-day average, has historically marked value zones rather than distress. Combined with 43,000 to 54,000 BTC in net accumulation by wallets holding 100+ BTC over the trailing 30 to 60 days, the hashrate exodus looks more like capital reallocation during a quiet accumulation phase than a miner capitulation event.
Bitcoin's security budget has weathered larger shocks before. What is different in 2026 is that the exit ramp, AI infrastructure, did not exist in any previous cycle. The machines did not go dark. They found a new job.
Follow @UTXOMacro for daily Bitcoin, macro, and AI breakdowns.