Data Center Agglomeration and Power Infrastructure Capacity in the United States
Yutong Ma
Abstract
The continuous expansion of digital infrastructure especially data centers has imposed unprecedented and geographically concentrated stress on existing power grids. However, rigorous causal evidence quantifying how these facilities reshape local power infrastructure supply remains limited. This study constructs a comprehensive county-level panel dataset spanning the United States from 2010 to 2024, uniquely integrating individual data center building area footprints with power plant additions. To overcome endogeneity and high-dimensional confounding, this study employs Double Machine Learning (DML) with five-fold cross-fitting, controlling for county and year fixed effects, along with regional socioeconomic conditions. The baseline estimates reveal a significant positive effect: The positive temporal association effect is highly robust across a series of rigorous treatments: it persists after 1% and 99% winsorization to exclude mega-clusters such as Loudoun County. Increasing the data center per unit area will significantly increase the electricity demand of nearby power generation facilities (ATE=1.982, P=0.000, CI: [1.25, 2.72]). It remains significant when using one-year lagged treatment to address reverse temporal association; and it holds under alternative state-level fixed effects. An event study framework, augmented with historical agglomeration controls, validates clean parallel trends and reveals a sharp rise in the point estimate precisely at t = 3 (coinciding
with the typical utility-scale grid construction lag), although the confidence interval is wide and includes zero. The estimates become more precise in later periods. Furthermore, a placebo test confirms that the related estimates cannot be replicated by random chance. Regional heterogeneity further reveals stark spatial divergence, with strong positive effects concentrated in the West and significant negative displacement effects observed in the resource-constrained Northeast. These findings offer critical, timely insights for urban planners and energy regulators, underscoring the necessity of forward-looking grid infrastructure investments to sustain the ongoing digital transformation.