The conventional industry narrative is that Bitcoin mining and AI hyperscale data-center workloads are complementary — both can leverage stranded power, both benefit from grid-stabilization payments, both are flexible large-load profiles. There is some truth to this story for the largest, most operationally sophisticated miners. For everyone else, AI is going to be a more aggressive competitor for the same finite power and transmission slots than the industry wants to admit.
The structural reality is that AI training workloads cannot tolerate the same level of curtailment that Bitcoin mining can absorb. AI tenants will pay more for predictable, always-on power than miners typically can, and they will negotiate longer-term contracts with utilities and grid operators. ERCOT's 226-gigawatt queue, dominated by AI projects targeting 2027 to 2030 commissioning, is the most visible example of the dynamic — Bitcoin miners' share of new interconnection requests has fallen below ten percent for the first time in five years, even as the overall queue has more than doubled.
The economics make the competition asymmetric. A modern Bitcoin miner pays roughly $50 to $80 per megawatt-hour to be operationally viable; AI hyperscalers can comfortably support $80 to $120 per megawatt-hour and increasingly negotiate above that for premium reliability tiers. Utility planners running competitive interconnection allocations naturally route the limited transmission capacity toward the higher-value tenant, especially when that tenant is also offering longer contract durations and reliability premiums. Miners who fail to bring something the AI customer wants — flexible curtailment commitments, behind-the-meter dispatchability, locational substation expansion — get progressively crowded out of the most attractive jurisdictions.
The largest miners have already pivoted in response. Iris Energy, TeraWulf, and Core Scientific have all signed multi-year HPC hosting contracts with named AI counterparties and are explicitly repurposing parts of their fleets toward GPU compute. Riot has stayed pure-play Bitcoin but is doing so on the back of long-term power contracts that pre-date the AI surge and won't renew on the same terms. The smaller, weaker miners — the kind that listed in the 2021-2022 mining-IPO wave — have neither the capital nor the technical sophistication to make the transition. They are the most exposed to what comes next. Bitfarms and CleanSpark have begun similar transitions at smaller scale. The pure-play miners that resisted the AI-pivot through 2024 are now the ones losing ground in interconnection negotiations, and the operational lesson is becoming hard to ignore: a Bitcoin-only thesis at the median public-miner level is no longer competitive with a mixed-workload one, and the gap is widening every quarter.
The defense of the synergy thesis deserves a hearing. The largest miners' AI-pivot success stories are real, the demand-response payments earned by flexible mining loads are growing not shrinking, and Bitcoin mining's geographic flexibility means it can chase stranded gas and renewable curtailment in places AI cannot easily follow. All true. But that argument is strongest at the top of the miner cohort, where operational sophistication and capital depth are highest. It does not extend to the median public miner, much less the long tail of smaller private operators who lack the scale to negotiate.
The miners who survive the next several years will be the ones who either pivot meaningfully into HPC hosting or accept being progressively crowded out of the most attractive jurisdictions. The remainder will likely have a harder time than the optimistic case suggests. The 2026-2027 milestones to watch are interconnection-queue dispositions in Texas and Wyoming, the rate at which miner-to-AI repurposing announcements continue, and whether any of the smaller miners produce a credible pure-play strategy that doesn't depend on Bitcoin price doing a meaningful share of the work. If they don't, the bear-market cohort of mining bankruptcies will be larger and faster than 2022's, and AI will be the proximate cause as much as Bitcoin price will be.