Study Finds AI Data Centers Raise Power Bills For Poorer Households

A new study published this week in the journal Nature Energy finds that the rapid buildout of artificial intelligence and cloud data centers has sharply driven up power costs on several major U.S. grids, with lower-income and rural households paying a disproportionately large share of the bill.^[1]

Researchers analyzed data from 2020 through 2025 across four large U.S. power grids and roughly 4,000 data centers, estimating that transmission spending on those grids rose from about $2 billion in 2020 to $34.2 billion in 2025, with data centers responsible for around $8.25 billion of the 2025 total. Residential customers were charged more than $16 billion for those grid costs in 2025, compared with about $3 billion billed directly to data centers, and wholesale power prices tied to data center demand were up nearly 20% in California's Southern Bay Area and about 17% in the Dallas-Fort Worth region that year.^[1] The study also concludes that rural and lower-income households saw energy-cost increases up to 76% higher than those for urban residents, even in many states where reported use of generative AI is relatively low.^[1]

The Nature Energy paper focuses on the 2020-2025 period, when tech companies raced to build and expand data centers to support cloud computing and, more recently, generative AI applications like large language models and image generators.^[1] To handle that surge in electricity demand, grid operators and utilities invested heavily in new transmission lines and upgrades, costs that regulators typically allow them to recover through fees built into customers' power bills. The study's authors argue that the way those costs are spread across customer classes leaves households and small businesses covering a much larger share of the grid spending tied to data center growth than the data center operators themselves, particularly in regions where big server farms cluster near cheaper land or favorable tax treatment outside major cities.^[1]

Researchers highlight an equity gap that they say is often obscured by the national conversation around AI innovation. While much of the tech industry and policymaking debate centers on AI's productivity gains and new services, the study notes that some of the steepest relative bill increases have fallen on rural and lower-income communities that report some of the lowest adoption of AI tools, effectively making them cross-subsidizers of infrastructure that primarily benefits large technology firms and heavy digital users.^[1] In regions like Northern California and North Texas, where dense clusters of data centers sit on already stressed grids, the added demand has not only pushed up wholesale prices but also required billions in new transmission spending that flows through in higher retail rates, hitting price-sensitive customers hardest.^[1]

The findings arrive as regulators and utilities across the country debate how quickly they can expand the grid to meet soaring electric demand from AI, crypto mining, electric vehicles, and industrial reshoring, and how fairly those costs should be divided among customers. The Nature Energy study suggests that without reforms to rate design or direct cost-sharing requirements for large data centers, households—especially those in poorer or rural areas—will continue to shoulder an outsized share of the infrastructure burden created by the tech sector's AI boom.^[1] Policymakers now face pressure to balance ambitions for AI leadership with consumer protection, as the social benefits and direct use of AI remain unevenly distributed while the costs of powering it are already landing in monthly electric bills nationwide.

1. CBS News