Do Data Centers Redistribute Power Reliability?

• In Event: A Critical Examination of Intersecting Policy, Socioeconomic, Energy, and Environmental Health Concerns About Data Centers

Abstract

The rapid expansion of large-scale data centers, driven by cloud computing and artificial intelligence, is reshaping electricity demand across the United States. Yet utilities and regulators typically evaluate grid performance using aggregate reliability metrics, which can mask important inequalities within utility service territories. This paper investigates whether data center siting redistributes power reliability spatially, improving service near large, high-priority loads while increasing outage risk elsewhere on the grid.To address this question, we assemble a novel national dataset that combines proprietary information on data center locations and capacity with high-resolution power outage data collected at 15-minute intervals across most U.S. utilities since 2023. Leveraging hyperlocal spatial variation and monthly temporal variation in data center commissioning, we implement near-far and near-near staggered difference-in-differences designs to identify the causal effects of data center growth on outage frequency and duration. We also examine whether these distributional effects are obscured when reliability is measured only at the utility or county level, and whether they intensify during extreme weather events, when grid stress is especially acute.Our analysis provides some of the first national evidence that data center expansion can reallocate power reliability rather than uniformly degrade or improve system performance. Preliminary findings suggest that communities located close to data centers experience localized reliability gains, while more distant communities within the same utility territories face higher outage risk. These disparities appear to become more pronounced during weather-related grid disruptions. We further explore how these patterns vary across institutional settings, including vertically integrated utility systems common in the Southeast and organized wholesale market regions elsewhere in the country. These findings highlight the need to move beyond aggregate grid metrics and consider the spatial distribution of reliability outcomes within service territories. The paper has direct implications for energy regulation, infrastructure planning, and equity-centered grid resilience policy in an era of rapidly accelerating AI-driven electricity demand.

Author

• Brian Y. An, Georgia Institute of Technology
  Presenting Author