AI competition is shifting from a race for chips to a race for electricity as data center power demands exceed infrastructure capacity.

A single hyperscale AI data center can consume as much electricity as a small city, and future AI data centers could require even more power.

Power infrastructure takes many years to develop through regulatory approvals, construction, and capital investment, whereas semiconductor manufacturing capacity can be built in about three or four years per Intel, creating a new bottleneck.

Companies like Constellation Energy (the largest nuclear power operator in the United States) and Applied Digital (which has contracted access to electricity for its AI data centers) may benefit as technology companies increasingly compete for access to electricity and data center capacity rather than GPUs alone.