Solving for AI’s power problem

What happened: A panel of energy and data center executives discussed how electricity shortages, not GPU supply, are becoming the bottleneck for AI data centers, and outlined approaches including microgrids, spare grid capacity, and fewer power conversion stages. Key numbers: Grid is underused by about 45% on average, per Janous. Why it matters: The panel framed "speed to power" as the industry priority, with leaders arguing that optimizing existing grid capacity and building microgrids as bridges can move faster than waiting on utility connections. Coverage: 5 outlets in 24h

A panel of energy and data center executives discussed how electricity shortages, not GPU supply, are becoming the bottleneck for AI data centers, and outlined approaches including microgrids, spare grid capacity, and fewer power conversion stages.

The panel framed "speed to power" as the industry priority, with leaders arguing that optimizing existing grid capacity and building microgrids as bridges can move faster than waiting on utility connections.