AI Data Center Boom Strains US Power Grids
US regulators and grid operators are implementing emergency tariffs and permitting rules as AI data centers drive electricity demand toward a reliability crisis.
Rapid expansion of artificial intelligence data centers is outpacing electrical grid capacity across the United States, prompting urgent regulatory interventions to prevent widespread blackouts. In Pennsylvania, the Pennsylvania Public Utility Commission warned that the regional power grid, operated by PJM Interconnection, could fail reliability standards between 2027 and 2030. An independent analysis found that electricity supply could fail to meet demand an average of six times per decade by 2030, with worst-case scenarios predicting 13 days of power loss annually.
To mitigate these risks, Governor Josh Shapiro issued an executive order requiring data center developers to provide their own power sources to qualify for streamlined permitting. The commission also established a large-load tariff for customers requiring 50 megawatts or more to ensure these facilities bear their own system costs. PJM Interconnection has proposed an Interim Resource Adequacy Service to allow large-load connections without compromising residential service.
Similar pressures are mounting in other regions. The Northwest Power and Conservation Council released a draft plan for a multi-billion dollar grid build-out across Washington, Oregon, Idaho, and Montana, forecasting demand increases of up to 100% over two decades. Meanwhile, researchers from the University of Iowa and other institutions released the Iowa Climate Statement 2026, urging the state to meet AI-driven demand through wind and solar expansion rather than new fossil fuel infrastructure. Nationally, Moody's Corporation reports that data center consumption could exceed 426 TWh by 2030, potentially requiring $110 billion in new generation capacity.