Bitcoin’s seven-day average hashrate has fallen to roughly 915.8 EH/s while miner holdings dropped by 1,530 $BTC over one week, according to mining data covering Sept. 20 through Sept. 26.
Digital Asset reported that the seven-day moving average stood at 915,844,520 TH/s on Sept. 26, down approximately 34.86 million TH/s from a week earlier. The reading was the lowest since around Sept. 3.
During the same period, CryptoQuant data cited by the report placed miner Bitcoin reserves at 1,192,766 $BTC. The balance was 1,530 $BTC lower than seven days earlier.
Bitcoin hashrate slips after September recovery
The hashrate reading indicates that less computing power was securing the Bitcoin network on a seven-day average basis than one week earlier.
Hashrate measures the combined computational power miners use to process Bitcoin transactions and compete to add new blocks. A higher network hashrate generally means more computing resources are participating in mining.
Public Mempool mining data supports the weaker late-September trend. Individual daily and pool-level readings moved below one zettahash during several sessions after climbing above that threshold earlier in September.
CoinWarz recorded Bitcoin network hashrate at roughly 954 EH/s on Sept. 25 after readings near 984 EH/s a day earlier. Its daily estimates showed substantial swings throughout September, including levels above 1 ZH/s on Sept. 15 and Sept. 9.
The movement follows a longer decline from Bitcoin’s late-2025 peak. Twenty One Capital CEO Raphael Zagury described the period as Bitcoin’s first “hashrate bear market,” based on the distance from the network’s previous record.
Zagury’s presentation put the decline from the late-2025 peak at roughly 22% to 24%. He linked part of the reduction to mining companies directing more power and capital toward artificial intelligence computing.
The Sept. 26 seven-day average represents a shorter-term reading and should not be treated as proof that all miners are reducing capacity. Bitcoin hashrate can fluctuate because of block timing, power conditions, mining difficulty and changes in active equipment.
Miner reserves fall by 1,530 $BTC in one week
Miner-held Bitcoin declined during the same week. CryptoQuant data cited by Digital Asset showed miner reserves at 1,192,766 $BTC on Sept. 26. The figure was down 1,530 $BTC from the previous week.
CryptoQuant defines miner reserve as Bitcoin held in wallets associated with miners and mining pools. A falling balance indicates that coins have moved out of tracked miner wallets, although the destination and purpose can vary.
Outflows can involve exchange sales, transfers to custodians, collateral arrangements, lending or other treasury activity. A decline in reserves therefore does not establish that every transferred Bitcoin was sold in the open market.
Earlier in September, miner reserves briefly moved in the opposite direction. Reporting based on CryptoQuant data placed miner holdings near 1.1919 million $BTC on Sept. 5, an increase of 261 $BTC from the previous week.
Public mining companies continue to show different treasury policies. CleanSpark, for example, mined 593 $BTC in August but sold 821 $BTC during the month.
As crypto.news reported on CleanSpark’s August production, the company finished August holding 13,703 $BTC. Its average operating hashrate reached 38.3 EH/s. Other miners have reduced or abandoned mining activity as computing infrastructure is redirected toward AI customers. Hyperscale Data stopped Bitcoin mining at its Michigan facility on Sept. 1 while preparing the site for an AI computing contract.
Crypto.news reported on the Michigan shutdown that the company’s Bitcoin holdings had fallen roughly 79% from 1,006 $BTC in July to around 215 $BTC.
Bitcoin’s Puell Multiple increased during the week even as miner reserves and hashrate declined. Digital Asset reported a reading of 1.13 on Sept. 26, up 0.24 from seven days earlier. The metric compares the daily U.S. dollar value of newly issued Bitcoin with its 365-day moving average.
A reading above one means daily miner issuance revenue is higher than its one-year average. The indicator does not directly measure miner profit because it does not account for each operator’s electricity, financing or equipment costs.
Public mining dashboards show similar readings around the neutral range. HalvingLens placed the Puell Multiple near 1.01 on Sept. 26 and classified the level as normal. The site estimated daily miner issuance near 450 $BTC and daily miner revenue around $37.8 million at the time.
The difference between individual Puell readings can result from data timing, price sources and methodology. The Digital Asset figure of 1.13 refers specifically to the source it cited for Sept. 26.
Bitcoin’s current block subsidy remains 3.125 $BTC after the April 2024 halving. Crypto.news explained in its updated halving guide that the next scheduled reduction is expected around 2028, when the subsidy will fall to 1.5625 $BTC per block.
Mining conditions remain uneven across regions
Mining conditions have differed sharply between operators during September. Ethiopia reduced electricity supplied to Bitcoin miners to 23% of contracted levels after reservoir inflows fell 20%, according to Ethiopian Electric Power.
Crypto.news reported on the electricity restrictions that Bitcoin mining consumed nearly one-third of Ethiopia’s electricity and generated 35% of the utility’s revenue during the previous fiscal year.
The utility plans to reassess conditions in October. Further restrictions could depend on reservoir levels and electricity demand. At the network level, Bitcoin continued producing blocks normally despite the lower seven-day hashrate reading. Mempool showed blocks continuing to clear on Sept. 26 while pool shares and short-term hashrate estimates shifted throughout the day.
Bitcoin’s difficulty mechanism adjusts approximately every 2,016 blocks to account for changes in network computing power. The next adjustment will respond to actual block production over the full difficulty period rather than a single seven-day hashrate reading.
