Data insight · Sep 29, 2026 · 4 min read
Bitcoin has recovered most of this year's losses since its June low. Mining difficulty, which measures how much machinery is actually competing for each block, has not: it is still 15% below the record it set in October, even after three straight increases. Here is what the gap between the two means for anyone running a machine today.
Bitcoin's low for 2026 came on 30 June, when it closed at $59,014. By 24 September it had closed at $84,385, a gain of 43% in twelve weeks. Most of that ground was made up in five weeks across August and September, when short covering and a run of ETF inflows drove the price from the low $60,000s to the mid-$80,000s.
The machinery competing to mine those coins has not kept pace. Difficulty, which resets every two weeks to whatever hashrate the network can actually support, closed its most recent adjustment on 19 September at 132.76 trillion. That is 15% below the all-time high the network reached on 29 October 2025, 155.97 trillion, and only a small part of the way back from where difficulty itself bottomed in June.
The fall was not smooth. Difficulty cut 11.16% on 7 February 2026, gave some of it back, then cut a further 7.76% on 20 March and 10.09% on 14 June — the day it reached its low for the year, 124.93 trillion, 19.9% below October's peak. Each of those cuts is machines switching off in numbers large enough for a two-week average to notice.
Single days are noisier than that. Our daily hashrate reading went from 826 exahash a second on 20 September to 1,016 the very next day — a swing bigger than the entire gap between June's low and today. It is estimated from how quickly blocks are found, and any one day's figure means little on its own. The seven- and thirty-day averages are what to read, and they tell a calmer story: a peak of 1,110 EH/s on 12 November 2025, a low of 897 EH/s on 29 July 2026, and about 932 EH/s now.
Since its June low, difficulty has added back 6.3%: +0.99% on 8 August, +1.31% on 5 September and +4.16% on 19 September, the largest single increase since the network started recovering. Three straight gains is a real turn. It has not closed the gap — 132.76 trillion is still well under 155.97 trillion, and difficulty's climb from its own low sits far behind the bitcoin price's.
In March, before the June low, CoinShares' head of research James Butterfill argued the network's winter decline was nothing like 2021's: "While the recent hashrate declines may seem alarming, looking on a log scale one can see they were far less severe than the China mining ban in 2021, representing a combination of cyclical and weather factors rather than something more ominous for the industry." The decline he was describing ran through February. Difficulty then fell further into June, and only in the last six weeks has it turned back up.
Bitcoin is up 43% from its low. Difficulty, from its own low, is up 6%.
Part of the shortfall is temporary: machines that went dark when hashprice fell below what they cost to run, waiting for it to clear again. Part of it is not. CoinShares' Q2 2026 mining report, published 15 September by researcher Luke Nolan, counted at least 35 EH/s scheduled to leave the group of publicly listed miners for AI and high-performance computing infrastructure — Keel already at zero, IREN due to exit by 31 December 2026 with 23.2 EH/s installed, Cipher likely to follow by the end of 2027 with 11.6 EH/s at its Odessa site. Core Scientific paid $41.9 million just to cancel 15 EH/s of hardware it had on order, on a bitcoin mining business Nolan put at a −56% gross margin even before the cancellation. Add the two together and roughly 50 EH/s that used to compete for blocks is being redirected on purpose.
"Bitcoin mining is not dying," Nolan wrote. "We are witnessing a large cohort of public miners shift their business models to capture what is a lucrative and seemingly secular tailwind." That is a reasonable read of the AI trade, and it is not a reason to worry about the network's security — hundreds of exahash of competition remain. It is a reason difficulty may keep lagging a rising price for longer than a typical cycle: some of the hashrate that used to come back when margins improved now has somewhere else to go instead.
Fewer machines competing for the same roughly 450 bitcoin paid out each day means more of it for the ones still running. Hashprice, what a petahash of mining earns in a day, was about $40 on 24 September, close to its recent high of $41 two days earlier and well above the low $30s it sat at for most of the summer. Part of that is the price rally. Part of it is a network that has not yet added back the hashrate the same rally would normally draw in.
None of that changes what a coin costs to produce — that is set by a site's electricity rate and the difficulty on the day, not by the price. But every retarget that lags a rising price widens the margin between hashprice and the electricity behind it, and for a miner who keeps every coin mined rather than selling it, that margin is what compounds. Difficulty has never stayed this far under a peak for long in the data we hold; the difficulty-adjustments chart shows every previous gap closing within a few quarters. Until this one does, it belongs to the machines already running.
Firsthand Bitcoin sells and hosts mining hardware, including this machine. No manufacturer, distributor or affiliate programme paid for or reviewed this page and we take no commission on the links above. Historical figures are computed from daily bitcoin price and network hashprice, each day valued at its own prices. Nothing here is investment advice.