Introduction

Data centres form the physical backbone of cloud computing, AI, streaming, and enterprise software. Their rapid expansion has raised questions about local economic and environmental impacts. This paper examines the effects of data centre growth on U.S. counties, studying employment, establishments, income, house prices, and electricity prices using long differences from 1995 to 2005, 2010, 2015, and 2020.

Data centres accounted for around 1.5 percent of global electricity consumption in 2024, or about 415 TWh (p.3). In the United States, electricity use rose from 58 TWh in 2014 to 176 TWh in 2023 — a 300% increase representing about 4.4 percent of U.S. electricity consumption (p.3). Impacts are highly local, as data centres are spatially clustered.

Data

The analysis uses facility-level data from the 451 Research Datacenter Knowledge Base (DCKB), a proprietary S&P Global Market Intelligence database covering more than 12,960 data centre facilities across over 2,700 providers and 131 countries (p.5). County-level outcomes are drawn from County Business Patterns, IRS Statistics of Income files, the FHFA House Price Index, and electricity price data from EIA Form 861.

To address endogenous siting, two shift-share instruments are constructed: a county’s pre-existing proximity to long-haul InterTubes fibre nodes interacted with Chinese data centre revenue growth; and a county’s 1980 share of national urban college population interacted with rest-of-world revenue growth.

Industry Background

Data centre growth accelerated sharply after the mid-2000s, particularly in Asia-Pacific, the United States and Canada, and Europe (p.7). The U.S. market is highly concentrated, with Virginia and Texas playing an outsized role (p.7). Critical capacity — the electrical and mechanical infrastructure supporting IT load — accounts for 53% of construction costs (p.8), explaining why data centres can generate large construction demand even when permanent employment is modest.

Descriptive Statistics For Main Outcomes

Treatment is sparse: by 2020, 7.6 percent of counties had positive data centre presence over the horizon, with a mean change in log cumulative data centre revenue of 0.190 (p.10). The share of counties with positive data centre presence rose from 3.0 percent in 2005 to 7.6 percent in 2020 (p.11).

Main IV Results

A one-unit increase in cumulative data centre revenue is associated with a 3.9 percent increase in total employment between 1995 and 2020 (p.12). Data-processing employment rises by roughly 27 percent in 2015 and 29 percent in 2020 (p.12). Construction employment effects are strongest at short horizons, consistent with a build-out phase (p.12). Establishments also rise at every horizon, with coefficients between 0.026 and 0.062 (p.12).

House prices increase at every horizon, with a 2020 coefficient of 0.177 (p.12–13). Electricity prices rise at every horizon, with a 2020 coefficient of 0.009 (p.13). Tax returns, adjusted gross income, and wages are positive at every horizon; the 2020 coefficients imply increases of about 5.9 percent in tax returns, 8.2 percent in adjusted gross income, and 5.9 percent in wages (p.13). Annual payroll responds less robustly.

Robustness

With state fixed effects, positive effects on data-processing employment, establishments, and house prices remain statistically significant. The total employment coefficient falls from 0.039 to 0.026 and is no longer statistically significant (p.16). Construction and electricity-price estimates are also no longer statistically significant with state fixed effects (p.16).

Conclusion

Data centres create measurable economic activity, especially in directly related sectors and during construction, and are associated with larger county-level income aggregates. They also raise electricity prices and are associated with higher house prices, which may benefit property owners while increasing costs for renters and prospective homebuyers (p.17).

Counties considering data centre projects should weigh employment, establishment, and income gains against pressures on electricity markets and local housing affordability (p.17).