The rapid expansion of artificial intelligence infrastructure has sparked intense debate across the United States regarding who should bear the financial weight of the necessary energy upgrades. A recent study provides new clarity on this issue, revealing that lower-income and rural households have shouldered a disproportionate share of the rising energy costs associated with data centers. This finding highlights a significant disparity between the entities driving demand and the consumers paying for the infrastructure.
Researchers from Cornell University published these findings this week in the journal Nature Energy. The analysis examined trends in energy demand and infrastructure projects across four major U.S. power grids from 2020 to 2025. The team also evaluated data from 4,000 existing and proposed data centers nationwide. Their work demonstrates that the proliferation of data centers supporting the AI boom is increasing energy costs in ways that place the greatest economic strain on those who derive the least benefit from the technology.
The study found that data centers have unevenly driven up wholesale power prices. These prices are established in open marketplaces where electricity is traded in bulk between generators and utilities before reaching consumers. According to the researchers, power grids that were already under strain experienced the most significant price hikes. For instance, data centers in California's Southern Bay Area were estimated to raise wholesale energy prices by nearly 20% last year, while the increase in the Dallas-Fort Worth area of Texas approached 17% in 2025.
In contrast, regions with fewer data centers, such as New York, did not see substantial changes in wholesale energy prices. The higher wholesale prices created downstream effects for residential customers, who ended up paying a large fraction of the costs for transmission infrastructure built specifically for data centers. Between 2020 and 2025, spending across the four analyzed grids surged from $2 billion to $34.2 billion. Data centers accounted for $8.25 billion of these infrastructure costs last year, yet residential users were billed more than $16 billion, compared to approximately $3 billion for data centers.
The researchers attributed this imbalance to outdated formulas used to allocate energy costs. Historically, these formulas divided costs based on previous energy consumption rather than new growth, causing households to absorb much of the expense for developing new power infrastructure. Fengqi You, one of the study's authors, described this as a chicken-and-egg problem. He noted that building a data center requires expanding power infrastructure to ensure electricity supply, but the facility is not using energy before it is constructed.
Residents in rural and lower-income areas saw their bills rise sharply, with some facing energy cost increases 76% larger than those in urban communities. The study explains that although data centers are concentrated near metropolitan hubs, the financial burden often spreads to surrounding suburban and rural areas due to policies governing how energy costs are shared. The researchers argued that these current policies require review to ensure fairer distribution.
A paradoxical mismatch was also identified in states with the greatest energy price increases linked to data centers. States like West Virginia and Ohio, which reported the lowest usage of generative AI, experienced the largest cost increases. Additionally, the study found that some states paid for infrastructure supporting data centers located far away. For example, residents in New Jersey and Pennsylvania helped fund power grid updates tied to data center growth in Virginia, effectively paying for infrastructure they neither contributed to nor directly benefited from.
To address these issues, researchers suggested that data centers reduce energy use during peak demand hours or pay states directly for the new demand they add to the grid. With regulating data centers being a major focus for lawmakers in November, the team hopes their findings will inform legislative decisions. You emphasized the need for science-based evidence to help all stakeholders understand the full picture and weigh their options in these ongoing debates.
