Bitcoin miners didn't plan to become infrastructure kingpins, but 151 data center moratoriums just handed them a monopoly on the one thing AI can't code its way around: permitted power capacity.
The Summary
- At least 225 data center moratoriums or restrictions are active across the US, blocking AI developers from plugging in new facilities while Bitcoin mining sites hold scarce, already-permitted power capacity
- ERCOT disclosed over 5 GW of conditional capacity allocated for AI deployments as miners pivot from Bitcoin to compute, highlighting the grid strain both industries create
- Publicly listed Bitcoin miners produced BTC at an average cash cost of $75,500 in Q2 2026, and even a price recovery won't bring them back from AI hosting deals
- Miners own the permits. AI companies need the power. The math is obvious.
The Signal
The irony is thick. Bitcoin miners spent years fighting for the right to exist, lobbying against power restrictions, weathering ESG criticism, and navigating hostile local governments. Now those same battle-tested sites, with permits already in hand, are exactly what AI developers desperately need. CoinShares documented 225 data center moratoriums or restrictions still active in Q2, and the number of outright bans sits at 151. That's not a permitting slowdown. That's a wall.
The permitting process for new data centers in the US now runs 18 to 36 months in most jurisdictions, sometimes longer if local opposition materializes. Miners already cleared that gauntlet. They hold operational permits, grid interconnection agreements, and relationships with utilities that took years to build. AI companies building frontier models can't wait three years to plug in a new cluster. They need power now, and miners are the only sellers in town.
"Miners own the permits. AI companies need the power. The math is obvious."
ERCOT's disclosure of over 5 GW of conditional capacity for AI deployments makes the pivot explicit. Texas, home to some of the largest mining operations in North America, is watching its grid absorb a new kind of load. The energy profile of Bitcoin mining and AI inference looks similar on paper: both want cheap, abundant power and neither cares much about location. But the economics diverged fast. When Bitcoin miners' average ex-tax cash cost hit $75,500 per BTC in Q2, and AI companies started offering multi-year hosting contracts with guaranteed revenue, the decision tree got simple.
Here's what's happening on the ground:
- Miners are repurposing existing sites for AI inference and training workloads, keeping the power infrastructure but swapping ASICs for GPUs
- AI developers are partnering with or acquiring mining operations outright to bypass permitting queues
- CoinShares projects that even a Bitcoin price recovery won't pull miners back from AI deals because the contracts offer more predictable cash flow
The grid can't tell the difference between a SHA-256 hash and a transformer inference run. Both pull watts. But regulators care a lot about the narrative. Bitcoin mining got framed as wasteful. AI compute gets framed as progress. That perception gap, combined with the moratorium wave, created a market dislocation. Miners hold an asset, permitted power capacity, that they didn't realize would become this scarce this fast.
The Implication
Watch for more mining companies to rebrand as "energy infrastructure providers" and cut deals with hyperscalers and AI labs. The ones that held onto their permits through the bear market now have pricing power. If you're building agents or training models, your deployment timeline just got tied to whoever owns the substation and the county permit. The compute bottleneck isn't chips anymore. It's the permission to turn them on.
For policymakers, the moratorium strategy backfired. It didn't stop energy-intensive computing. It just handed market power to the companies that got in early and made it harder for new entrants to compete. The unintended consequence: Bitcoin miners are now infrastructure landlords, and AI is their highest-paying tenant.