Data center expansion and community power resilience in the USA

• In Event: The Infrastructure of Intelligence: Navigating Siting Conflicts, Community Equity, and Grid Resilience in the AI Era

Abstract

The rapid expansion of data centers is reshaping electricity demand patterns, introducing large and geographically concentrated loads that raise new challenges for power system reliability and distributional equity. As artificial intelligence workloads drive data center electricity consumption toward an estimated 426 terawatt-hours by 2030, the distributional consequences of this infrastructure boom for community-level grid reliability remain poorly understood. While existing research has examined electricity costs and capacity markets, less is known about how data centers affect community-level power outages at scale across diverse regulatory and grid environments. This paper provides the first nationally representative empirical analysis of how data center deployment affects county-level power outage outcomes, covering 14 US states that collectively account for the vast majority of national data center capacity.We construct a panel dataset of 148,169 county-month observations combining facility-level data center records with high-frequency outage data from the EAGLE-I program at Oak Ridge National Laboratory, supplemented by county socioeconomic characteristics from the American Community Survey. The dataset encompasses 1,575 operational facilities, representing approximately 133 gigawatts of total installed capacity. Two-way fixed effects estimates show that a 1% increase in county-level data center stock is associated with a 5.6% reduction in monthly cumulative customer outage hours. This finding is robust across alternative treatment variable specifications, including log capacity and log facility area, and replicates across both outcome measures: cumulative customer-hours and maximum percent of customers out. Heterogeneity by capacity size suggests that the reliability benefit may scale with the magnitude of infrastructure investment triggered by data center entry rather than facility count alone.These findings suggest that data center entry induces local grid infrastructure reinforcement, generating positive reliability spillovers for surrounding communities. The national scope of this analysis substantially strengthens external validity beyond prior single-state evidence, suggesting that infrastructure reinforcement effects reflect a general pattern across the U.S. electricity system. As regulators and utilities navigate the interconnection of unprecedented large loads, understanding whether data center entry improves or degrades local reliability is central to designing equitable cost-allocation mechanisms, siting approval processes, and grid investment frameworks. The findings also highlight that infrastructure investments driven by large-load interconnections may generate broader community reliability benefits.

Author

• Xiaofeng Ye, University of Maryland, College Park
  Presenting Author