How Do Data Centers Impact Grid Resilience? Evidence from Power Outages and Hurricanes

• In Event: Data Centers, Grid Infrastructure, and Energy Investment

Abstract

The rapid expansion of data centers is transforming electricity demand in the United States, introducing large and geographically concentrated loads that raise new challenges for grid reliability and energy justice. While prior research has focused on electricity markets and system-level impacts, less is known about how data center deployment affects community-level exposure to power outages—particularly the severity and recovery during extreme weather events. This paper asks whether data centers are associated not only with improved reliability, but also with enhanced grid resilience, measured through both outage intensity and recovery performance. We construct a novel county-level panel dataset for Florida spanning 2014 to 2025, combining facility-level data center information with high-frequency outage records, hurricane exposure measures, and socioeconomic characteristics. The dataset captures both routine outages and disruptions during 17 major hurricane events. To characterize reliability and resilience, we measure outage outcomes using cumulative customer-hours without power (total burden), peak outage rates (intensity), and outage duration based on restoration thresholds (recovery time). To estimate the relationship between data center deployment and outage outcomes, we employ multiple empirical strategies. First, we use panel fixed effects models to exploit within-county variation in data center expansion over time. Second, we analyze hurricane-induced outages using cross-sectional comparisons within events and two-way fixed effects models across events. We also implement event-study designs and reweighting approaches to assess identification assumptions and address non-random siting of data centers. We find three main results. First, counties hosting data centers experience higher outage burdens in absolute terms due to larger customer bases, but lower outage rates when normalized, indicating stronger baseline grid performance. Second, within counties, each additional data center is associated with a 4.4% reduction in monthly outage hours and a 0.4 percentage point decrease in peak outage rates, with larger improvements observed during extreme weather periods. Third, evidence from hurricane analyses is mixed: while within-county estimates suggest improved resilience, event-study results reveal that counties with data centers already exhibited systematically different outage patterns prior to deployment, indicating non-random siting in areas with distinct infrastructure characteristics. These findings suggest that data center expansion is associated with improvements in both reliability and recovery performance, potentially through infrastructure investment and grid reinforcement. However, the observed benefits may partly reflect pre-existing advantages rather than causal effects. The results highlight important policy considerations for utilities and regulators managing large-load interconnections, including infrastructure planning, cost allocation, and the equitable distribution of resilience benefits across communities in an era of rapidly growing digital infrastructure.

Author

• Yufan Ji, The Ohio State University
  Presenting Author