Power Access for AI: The Flexibility Compact

In brief

Arushi argues that access to the power grid is a central bottleneck for affordable AI, because planners assume data centers will always draw their maximum. She describes work under way to verify how data centers can adjust their consumption in real time, proposes policy price signals that reward flexible operation, and calls for activating flexibility early so that grid studies and capital follow realistic load profiles. Her flexibility compact is an enforceable agreement among developers, utilities and regulators that trades faster interconnection for dependable flexibility.

Key points

1. 01Worst-case assumptions about data center demand delay access to power.
2. 02Verifiable operating profiles can show how data centers adjust their use in real time.
3. 03Five policy moves would put a price on flexibility and speed up interconnection.