Bitcoin difficulty dropped from 138.96T to 124.93T at block 953,568 on June 14, the second-largest negative adjustment of the year. Hashrate is down 12% in a month to 886 EH/s, and the marginal kilowatt is increasingly being sold to AI hyperscalers instead.
Bitcoin's mining difficulty dropped 10.09% at block 953,568 on June 14, falling from 138.96 trillion to 124.93 trillion in the second-largest negative adjustment of the year and the eleventh-largest in the protocol's history.
The adjustment was mechanical. Difficulty resets every 2,016 blocks, and the network had been mining slower than the 14-day target (the epoch took 15.6 days to complete) because hashrate had been coming offline for two weeks. The cause of that drop was equally mechanical. Bitcoin slid roughly 15% over the first half of June, from the high $70,000s into the low $60,000s, and the price now sits around $64,000 with ETF flows in their sixth consecutive week of net outflows. At that price, a meaningful chunk of the global fleet is producing blocks at a loss on an all-in basis. The marginal rigs went dark.
The numbers tell a consistent story. Total network hashrate has fallen roughly 12% over the past month, sitting at 886 EH/s according to Blockchain.com, a 23% drop from the all-time high reached in October. Hashprice, the daily revenue per petahash per second tracked by Hashrate Index, jumped 13% on the adjustment to about $33 per PH/s per day, pushing the median operator back above the rough $30 gross breakeven. That is the network doing what it is designed to do: when revenue collapses, difficulty drops and the survivors collect a bigger share until the next epoch.
The all-in picture is uglier. Spot revenue per PH/s is back above breakeven on operating cost, but the all-in cost (debt service, depreciation, hosting, ASIC capex) still sits closer to $80,000 of BTC for the average public miner. Bitcoin at $64,000 is roughly a quarter below that line. The difficulty drop bought the network's miners some oxygen; it did not put them in the money.
This is the second time in two months the same story has run. Difficulty fell 7.76% in mid-May on the back of an earlier price slide, and again 2.3% on May 1 before that. Sunday's cut takes the year-to-date count of negative adjustments to eight. The network has not seen a sustained run of negative adjustments like this since the China mining ban exodus in mid-2021.
What separates 2026 from 2021 is what the surviving capacity is doing with itself. The Chinese fleet that went offline four years ago migrated to Kazakhstan, Texas and Paraguay; the capacity that has gone dark this month, by contrast, is increasingly being repurposed rather than relocated. MARA sold 20,880 bitcoin in Q1 for $1.5 billion and refiled its 10-Q describing itself as a compute company; CoreWeave, Galaxy and Iris are all running revenue lines that did not exist in 2024. The AI training and inference demand that public miners have been hedging toward for two years is now the more profitable use of their substations, their leased land and their cooling water on most days when bitcoin trades below $70,000.
The result is a network whose downside has structurally changed shape. In a 2018-style bear market, miners switched off and waited. In a 2026-style bear market, they switch their generators over to a different customer and stop competing for blocks. The hashrate that comes back when bitcoin recovers is not necessarily the same hashrate that left. Some of it is now contracted to hyperscalers on terms that make it uneconomic to flip back to SHA-256 mining at any plausible bitcoin price.
That has second-order implications for difficulty. If a meaningful share of the lost hashrate does not return, the next positive adjustment will be smaller than the recent negative ones, and the long-run trend of ever-rising difficulty (the line every miner business model assumes) flattens. For the marginal operator, that is good news. For the public majors that have built their fleets around the assumption that they can outscale their cost of capital, it is not.
Marathon Digital, CleanSpark, Riot and Core Scientific remain the four largest public US miners by hashrate going into the second half of the year. Each has spent the last 18 months telling investors that the AI pivot is additive rather than substitutive — that the bitcoin business stays intact while the high-performance computing business grows on top. The June difficulty cuts make that line harder to defend. Capacity is fungible. Capital allocators have to choose. The marginal kilowatt at a Texas substation is currently better deployed selling H100 hours to Anthropic than chasing a block subsidy that is six months past its fourth halving.
The next difficulty adjustment is due around June 28. Hashprice is still well below where it sat at the start of the year, and ETF outflows have not turned. If price stays here, the cut will be smaller — most of the marginal hashrate has already left — but the direction will not.
