Former crypto miners are now worth $48 billion because they control something Big Tech desperately needs: power contracts.

The Summary

The Signal

The AI infrastructure race has a chokepoint, and it isn't compute or talent. It's power. While everyone watches NVIDIA's margins and OpenAI's roadmap, a cluster of crypto refugees turned themselves into power brokers. They hold utility contracts that can deliver megawatts to AI data centers without the 3-5 year lead times that normally come with industrial-scale power procurement.

This is infrastructure arbitrage at scale. When crypto collapsed, these firms were left holding contracts for cheap electricity in places like Ohio and Texas. Turns out training frontier models requires the same thing that mining Bitcoin did: absurd amounts of stable power. Big Tech needs compute capacity yesterday. These firms can flip the switch in months instead of years.

The market is pricing this access accordingly. A 23x valuation jump in three years isn't hype, it's recognition that these companies own a scarce resource in an environment where Microsoft, Google, and Amazon are committing to unprecedented AI infrastructure spending. B. Riley analyst Nick Giles calls it a "winning scenario" for investors. That's restrained language for what amounts to being the only seller in a seller's market.

But the real test starts now. Power contracts got them in the door. Execution keeps them there. These firms have to deliver complex, capital-intensive data center projects on tight timelines for customers who won't tolerate downtime. They're competing against established infrastructure players who may lack the power contracts but don't lack experience building at scale.

The Implication

Watch how many of these eleven firms are still standing in 18 months. The ones that can't execute will get crushed. The ones that can will become permanent fixtures in the AI stack, potentially acquisition targets for hyperscalers who'd rather own the power contracts than rent them. If you're building agent infrastructure or training models, map who controls power in the regions you're scaling into. It matters more than rack space.


Source: Business Insider Tech