The Quiet Exodus
Bitcoin's network hashrate peaked above 1,000 exahashes per second in early 2026. It has not been back since.
The protocol has responded the way it always does when miners leave: it made the puzzle easier. Six difficulty reductions in 2026 have brought mining difficulty down 10.7% from its January peak. On May 1, the latest cut of 2.3% pushed difficulty to 132.47 trillion, and the hashrate now sits around 959 EH/s — well below the symbolic 1 zettahash threshold the network briefly celebrated crossing.
What makes this drop different from every previous dip is the reason behind it. Miners aren't shutting down because they can't pay their power bills. They're shutting down because they found something that pays better.
Follow the Money
At least eight publicly traded mining companies have announced plans to pivot partly or entirely into AI and high-performance computing (HPC). The numbers explain why.
One megawatt allocated to Bitcoin mining generates volatile, shrinking revenue at roughly $1 per kilowatt-hour. The same megawatt leased to a hyperscaler under a long-term AI infrastructure contract delivers predictable income at approximately $25 per kilowatt-hour. The gap is not close.
Core Scientific now draws 39% of its revenue from AI colocation. TeraWulf is at 27%. IREN is at 9% and scaling fast, with up to 200 megawatts of liquid-cooled GPU capacity under construction. Across the public mining sector, over $70 billion in cumulative AI and HPC contracts have been signed.
Industry projections suggest AI and HPC could account for 70% of revenue for these transformed miners by end of 2026, up from roughly 30% today. These are no longer Bitcoin mining companies. They are data center operators who happen to still mine some Bitcoin.
The Security Question
Every Bitcoin holder should understand what this means for the network.
When a miner signs a five-year contract to host AI compute, that power capacity is effectively removed from Bitcoin's security budget for half a decade. It cannot dynamically return to mining if the network needs it. This is not a temporary dip that self-corrects when price rises — it is a structural reallocation of energy infrastructure.
The difficulty adjustment mechanism is doing its job, keeping blocks flowing at roughly ten-minute intervals. But the total computational work securing each block is declining. A lower hashrate means a lower cost to attack the network. At 959 EH/s, Bitcoin remains extraordinarily secure by any practical measure. But the trend line matters more than the snapshot.
CoinShares projects hashrate could reach 1.8 zettahashes by the end of 2026 — but only if Bitcoin returns above $100,000, which would make mining profitable again for older, less efficient rigs. At current prices around $73,000, that recovery looks distant.
The Debt Problem
The pivot to AI is not risk-free for the companies making it.
IREN now carries $3.7 billion in convertible notes across five series. TeraWulf holds $5.7 billion in total debt split between convertible notes and senior secured loans. These are massive leveraged bets on sustained AI compute demand from hyperscalers like Google, Microsoft, and others.
If AI compute demand slows — or if hyperscalers build out their own infrastructure faster than expected — the most leveraged former-miners face serious exposure. They would have abandoned Bitcoin revenue without securing a stable replacement.
A Historical Parallel
This is not the first time external economics pulled compute power away from Bitcoin. In 2010, the arrival of GPU mining made CPU mining obsolete overnight, causing a temporary hashrate disruption before the network stabilized at a higher level.
The difference in 2026 is that the competing use case — AI training and inference — is not a more efficient way to mine Bitcoin. It is an entirely separate industry competing for the same scarce resource: cheap electricity and data center space.
The weighted average cash cost to produce one bitcoin among publicly listed miners rose to approximately $79,995 in Q4 2025. With Bitcoin trading around $73,000, most public miners are operating below their cost basis. The AI pivot is not opportunism — it is survival.
Bitcoin Gate Take
The difficulty adjustment is Bitcoin's most elegant feature. It will keep working. Blocks will keep coming every ten minutes whether hashrate is 500 EH/s or 5 ZH/s. But the long-term security budget — the economic incentive that keeps miners honest — depends on Bitcoin being worth enough to compete with alternative uses of energy. The AI economy is the first serious competitor for that energy at scale. This does not break Bitcoin. But it does mean the next bull market is not just about demand — it is about whether the price rises enough to bring miners back.