How the September 2026 Bitcoin difficulty adjustment works

The network adjusts to price pressure

Bitcoin difficulty reached 127.45 trillion at block 965,664 following a 1.31 percent upward retarget on September 5, 2026. This adjustment follows an August 23 decline of 1.31 percent that set difficulty at 125.81 trillion. I see the network stabilizing after the 10.09 percent drop in June which reduced difficulty from 138.96 trillion to 124.93 trillion at block 953,568. This June drop was the second largest of 2026, following an 11.16 percent fall in February that hit difficulty levels of 144.39 trillion after a US winter storm. The hashprice sits between $32 and $33 per PH/s per day. Mining one bitcoin costs about $84,300 while Bitcoin trades near $65,600.

The network is in a stress zone.

The June price decline squeezed miner margins and forced some operators to shut off unprofitable hardware, which caused the epoch length to stretch to 15.6 days and triggered the automatic difficulty correction for the entire network.

The mechanics of the 2,016 block epoch

The Bitcoin protocol recalculates difficulty every 2,016 blocks to preserve a 10 minute block time. When miners find blocks faster than 10 minutes, the target decreases. If the actual epoch time exceeds 20,160 minutes, the difficulty decreases. The algorithm uses the ratio between the expected 20,160 minutes and the actual time taken to mine the last 2,016 blocks. For example, if the actual time to mine the last 2,016 blocks was 13 days, the difficulty would increase by a factor of 14/13. Difficulty remains the ratio of the maximum possible target to the current target.

The September 5 adjustment happened because miners found blocks slightly quicker than the 10 minute target in the previous period. A single adjustment cannot reduce difficulty by more than 75 percent. I find the technical simplicity of the 2,016 block boundary effective. The calculation uses the formula new target = old target * clamp(actual timespan, 3.5 days, 56 days) / 14 days.

Can mining hardware return to the network?

AI shifts and miner revenues

Miners shift capacity to artificial intelligence and high-performance computing. Cango already shut down a third of its equipment to expand into AI. This movement explains why the network hashrate dropped to 740-886 EH/s from October peaks. I observe that the remaining miners earn about 11 percent more bitcoin per unit of active computing power following the June difficulty drop. The Puell Multiple fell from 0.83 to 0.74 over ten days.

The mining market changed.

You know the basics of hashrate, so I will focus on the revenue. While revenue rises for those staying online, the price drawdown from the last difficulty low reached 21 percent. A meaningful deterioration would require bitcoin to drop below $55,000 without another difficulty adjustment. Hashrate fell. The target changed.

The network hashashrate settled around 900 to 912 EH/s following the recent retargets. This reflects a recovery from mid-year lows near 886 EH/s. Foundry USA holds 26 percent of the hashrate, followed by AntPool at 17 percent and F2Pool at 15 percent.