Study finds AI data center costs fall hardest on poorer, rural households

Cornell researchers found that AI data centers raised power costs unevenly across parts of the United States in 2025.

Their study examined energy demand and infrastructure on four major power grids from 2020 to 2025, along with 4,000 existing and proposed data centers.

Areas with already strained grids saw the largest wholesale price increases, including nearly 20% in California’s southern Bay Area and almost 17% near Dallas-Fort Worth.

Households were billed more than $16 billion for infrastructure costs, while data centers were billed about $3 billion, the study found.

Researchers said older cost-sharing formulas, based on past electricity use rather than new demand, help shift the expense of building new power infrastructure onto residents.

Lower-income and rural households faced the greatest burden, with some rural residents seeing energy cost increases 76% larger than those in urban communities.

The findings highlight who may pay for the power systems needed by data centers, though the study does not say what policy changes will follow.