A surge in artificial-intelligence data centers has lifted power costs unevenly across U.S. grids, with low-income and rural households shouldering a disproportionate share, according to Cornell University research published in Nature Energy. The study, which examined four major power grids and 4,000 existing and proposed data centers from 2020 to 2025, found that data-center growth pushed wholesale electricity prices higher in already strained regions—nearly 20% in California’s southern Bay Area and about 17% in the Dallas–Fort Worth area last year—while areas with fewer facilities saw limited impact.
The buildout has also driven a spike in transmission spending, from $2 billion in 2020 to $34.2 billion in 2025. Researchers estimate data centers caused $8.25 billion of those costs last year but were billed about $3 billion, while residential users caused under $3.5 billion yet were billed more than $16 billion. Antiquated cost-allocation formulas that rely on historical consumption, rather than new demand, shift expenses onto households, the study says.
The burden has fallen hardest on rural and lower-income communities, where bills rose as much as 76% more than in urban areas. Cross-border cost sharing means customers in states such as New Jersey and Pennsylvania helped pay for grid upgrades tied to data-center growth in Virginia. The authors urge policymakers to reform cost allocation and consider measures such as curtailing data-center load during peak hours or requiring operators to directly fund the infrastructure they necessitate.



























