AI’s power crunch turns Bitcoin miners’ grid access into an asset
AI data-center power capacity reached about 29.6 GW by end-2025, showing that electricity access, not compute chips, is the main bottleneck. GPU costs have fallen sharply and chips are far more efficient, but total AI demand keeps rising as savings are reinvested into larger models. Stanford estimates the biggest training runs now need over 100 MW, and AI capacity has grown about 200-fold since 2022. Bitcoin miners are well positioned to benefit because they already own energized sites, grid hookups, cooling shells, and power contracts. Their ASICs cannot train AI models, but the surrounding infrastructure can be repurposed or leased to AI firms, especially in cheap-power states like Texas and Louisiana. As mining economics weaken, many miners are signing large AI/HPC deals with Microsoft, Google-backed partners, CoreWeave, and others. Investors have rewarded this pivot, though converting sites is expensive and time-consuming, and the business case depends on sustained AI demand.
