Bitcoin miners spent the previous cycle acquiring interconnected sites with large, firm power contracts in regions with cheap electricity. Those positions, assembled for an entirely different purpose, turn out to be the scarcest asset in the AI buildout. A miner with 300 MW already flowing can deliver compute capacity years before a greenfield developer can energise the same load.
The conversion is not trivial. Mining halls are built for tolerance, not uptime: ambient-air cooling, minimal redundancy, and a design philosophy that accepts hardware failure as a cost of doing business. AI training and inference workloads demand the opposite. Retrofitting for Tier-equivalent redundancy, liquid cooling, and carrier-grade connectivity consumes much of the apparent cost advantage.
Why the Financing Is the Signal
What makes the raise notable is that lenders underwrote it. A $7.5 billion project financing against converted mining assets implies that credit committees now accept the conversion thesis, including the retrofit risk and the counterparty risk of leasing to AI tenants whose own revenue models are young.
That acceptance is itself a market indicator. Capital is pricing interconnection position as more valuable than facility quality, because facility quality can be bought and interconnection cannot.
The Risk Nobody Is Pricing
Parselion's assessment: converted mining sites will carry a structurally different reliability profile from purpose-built facilities for at least a decade, and tenants may not be diligencing that difference carefully. We recommend that any operator or investor evaluating converted capacity commission an independent structural and thermal assessment rather than relying on the nameplate megawatt figure, which says nothing about whether the hall can sustain a 100 kW rack.