Journal Article

Spatially distributed air quality and health impacts of data center electricity demand in the United States

Chen, Y., S. Wang, J.R.L. Senga, A.S. White, A.Y.H. Wong, L.M. Freese, M.L. Bell and N.E. Selin (2026)
Environmental Research Letters, 21(21), 214046 (doi: 10.1088/1748-9326/aebbd4)

Abstract / Summary:

Abstract: The rapid expansion of artificial intelligence is driving substantial growth in electricity demand from data centers in the US, yet how the resulting air quality and public health impacts are distributed remain poorly quantified. To trace the ambient air pollution and associated premature mortality impacts of grid-supplied data center electricity demand in 2023, we couple the United States Electricity Generation Optimization model with the GEOS-Chem chemical transport model. We compare hourly, unit-level electricity dispatch with and without data center demand to quantify changes in power-sector emissions, ambient PM and ozone, and associated premature mortality. 

Our results show that electricity generation attributable to data center demand is largely supplied by fossil fuel units across multiple regions, resulting in additional annual mean PM concentrations of 0.02–0.08 g m (local peak concentrations up to 0.2 g m), and regionally varying ozone responses characterized by wintertime depletion and summertime enhancement. 

These impacts are most pronounced in the Midwest United States, where marginal emissions from coal-fired units and ammonia-rich atmospheric conditions enhance secondary particulate formation. We estimate that this grid-supplied demand is responsible for 513 (95% CI: 469–557) annual premature mortalities from PM and an additional 137 (95% CI: 90–183) annually from maximum daily 8 h average ozone (MDA8-) in the contiguous US. State-level analysis reveals a systematic geographic mismatch: states hosting the largest data center loads do not necessarily experience the highest attributable mortality rates, while some states with little data center demand bear substantial burdens. This mismatch arises because of dispatch across interconnected electricity regions, regional differences in generation and emissions, and atmospheric chemistry and transport, which together separate the locations of demand from those of exposure and mortality. These findings highlight the importance of considering both air quality impacts and their spatial distribution in data center and macro-energy system planning.

Citation:

Chen, Y., S. Wang, J.R.L. Senga, A.S. White, A.Y.H. Wong, L.M. Freese, M.L. Bell and N.E. Selin (2026): Spatially distributed air quality and health impacts of data center electricity demand in the United States. Environmental Research Letters, 21(21), 214046 (doi: 10.1088/1748-9326/aebbd4) (https://eartharxiv.org/repository/view/13648/)