The Spatial Incidence of Hyperscale Data Centers

• In Event: Assessing data center policy design and impacts

Abstract

This paper estimates the local incidence of U.S. data center openings using a stacked difference-in-differences design and nationwide housing transaction data. Hyperscale data centers—developed and operated by large cloud and AI providers—have larger footprints, higher rack density, and greater power intensity than non-hyperscale facilities. We find that hyperscale data centers reduce nearby house prices by 6.8 percent, with effects highly localized and fading beyond 14 km. The implied aggregate capitalization loss is approximately $3.6 billion. By contrast, non-hyperscale facilities show little evidence of capitalization on average.

We provide evidence that the local incidence of data center development depends not only on scale, but also on siting context, local residential exposure and how burdens are perceived by surrounding communities. First, housing price declines increase with facility power capacity and are more pronounced when the project represents the first data center in a county. Second, hyperscale sites tend to locate in areas with greater residential exposure. Although non-hyperscale facilities have little average price impact, those located in more residentially exposed environments also reduce nearby housing values. Third, we find that the effects attenuate for properties located outside the data center’s zoning jurisdiction, suggesting that shared regulatory context shapes how households perceive and price these developments.

We also examine additional incidence channels. We document a modest but persistent increase in residential electricity prices in host areas after commissioning, consistent with load-driven procurement costs and grid investments being partially passed through to households. Using water-service boundary data, we show that absorbing water-service-by-time variation substantially attenuates the estimated effects, pointing to utility-level channels. Consistent with hyperscale facilities being capital intensive and labor light, we find little evidence of sustained local labor-market effects.

Overall, the results suggest that the costs and benefits of hyperscale development are spatially misaligned: localized burdens are capitalized into nearby housing markets, while fiscal benefits likely accrue over broader jurisdictions.

Authors

• Liu Ee Chia, National University of Singapore
  Non-Presenting Co-Author

• Siyuan Hu, University of Wisconsin-Madison
  Presenting Author