The second-largest difficulty cut of 2026 reflects what miner earnings have been telegraphing all year: Marathon, Riot and their peers are pulling power and capital out of bitcoin mining and into AI data centres as fast as the contracts can be signed.
Bitcoin's mining difficulty dropped 7.76 per cent at the most recent adjustment, the second-biggest single cut of 2026, and the explanation has nothing to do with seasonal heat in Texas or geopolitics in the Strait of Hormuz. It is the same thing public miner earnings have been telegraphing all year: the largest names in the sector are moving capital and computing capacity out of bitcoin mining and into AI infrastructure as fast as they can sign the contracts.
Public bitcoin miners sold a record 32,000 BTC in Q1 alone — more than the combined total of all four quarters of 2025. Marathon Digital led the volume, dumping 15,133 coins in March for roughly $1.1 billion. Riot Platforms sold another 3,778 BTC at an average price of $76,626 in the same quarter, generating $289.5 million. Both companies cited capital needs for data centre buildouts. Neither has framed it as a retreat from bitcoin, but the cash flow tells a different story than the earnings-call narrative.
Marathon's Q1 numbers read like a company managing through a structural shift it isn't sure it can survive. Net loss of $1.3 billion. Revenue down to $174.6 million from $213.9 million in Q1 2025. Hashrate up sharply — from 54.3 EH/s to 72.2 EH/s, a 33 per cent jump — because the rigs are still running even as the financial logic of running them has weakened. The combination is the story of the entire sector: more machines, less revenue per machine, and a treasury that needs to be liquidated to pay for the next thing.
Riot's quarter was worse on the optics. Its core bitcoin mining business shrank 21.7 per cent, from $142.9 million in revenue to $111.9 million, against power costs the company has kept impressively low at three cents per kilowatt-hour. The compensating line was the new data centre operations division, which contributed $33.2 million — small in absolute terms, but the only segment growing. Riot's Q1 filing made the pivot explicit; the company is no longer pretending the mining business and the data centre business are separate strategies.
MARA's filing went further. The company filed Q1 as a compute company, restructured its convertible debt, and signed a $1.5 billion deal for a 505 MW Ohio gas plant most of which was financed through debt the seller already owed. The Ohio plant is not earmarked for bitcoin mining. It is for AI training compute and inference workloads, which carry margins bitcoin mining hasn't seen since the 2021 cycle.
The difficulty drop is the network's reaction to the rig disconnects that follow this kind of capital reallocation. When miners turn off machines to redeploy power to GPU clusters, the network's effective hashrate falls; when the difficulty adjustment happens, the target drops to match. The 7.76 per cent cut at the most recent epoch is the second-largest of 2026, after the smaller 2.3 per cent cut on May 1 that took hashrate back below one zettahash. The pattern is consistent: difficulty rises in months when the bitcoin price is climbing, then drops in months when miners are pivoting capacity to other workloads.
That cycle has security implications. Bitcoin's security model rests on the cost of mounting a 51 per cent attack, which scales with the network's total hashrate. A sector-wide migration of compute toward AI workloads means the marginal economics of attacking the network keep getting cheaper, even if the absolute hashrate remains historically high. The 945 EH/s number on the chart looks reassuring; what it does not show is how much of that hashrate is being run by operators who are actively winding down the underlying business.
Power costs are the other constraint. Riot at three cents per kilowatt-hour sits at the low end of the public miner cohort, and that floor is part of why the company is still profitable on a per-coin basis even as revenue shrinks. Marathon's all-in costs are higher, which is part of why the volume of coin sales in March made sense as a balance-sheet decision. The miners with cheaper power can keep mining through this cycle; the ones with more expensive power are converting the rigs into something else entirely.
For the next quarter, the operational question is whether public miners will continue selling treasury coins to fund data centre capex or whether they will rely on debt issuance instead. Marathon has already restructured. Riot has kept its debt position lower but is moving in the same direction. The 32,000 BTC sold in Q1 was a one-time event in one sense and a leading indicator in another. The Q2 numbers will say which.
What the difficulty drop confirms is that the sector pivot is not just an earnings-call slide. It is showing up in the network itself, one machine at a time.
