President Donald Trump’s stated desire to build data centers while lowering utility bills is butting up against the costly realities of an outdated electric grid.
A new study, however, points to one solution: building data centers in rural areas where solar energy is already flourishing.
Trump has done his best to weaken the renewable energy market. But California has continued to see a massive growth in solar power — so much, in fact, that its grid has not kept up.
Traffic jams on transmission lines mean that some power produced in the Central Valley doesn’t reach the cities where it’s needed. So researchers at the University of Pennsylvania considered whether data centers could sop up that extra power.
“Rather than moving the energy through the wires, let’s move the data center to the energy,” said Stephanie Leonard, director of research for the California-based environmental nonprofit Next 10, which commissioned the report. “It’s a solution that not only avoids transmission costs, it’s one that doesn’t have to be paid for by the ratepayers.”
Although the study focuses on California, authors say the idea of “curtail to compute” zones could be applied more broadly in the United States. Policymakers are already grappling with how to pay for the expensive grid upgrades that data centers often require, while tech companies have pledged to pay any extra electricity needs in a nonbinding White House agreement.
Computing cost
The amount of solar power that California’s grid operator curtailed in 2024 could have powered 500,000 California households for a year — a loss of revenue for energy developers and a massive inefficiency for an already-strained grid.
Building new transmission lines to serve the San Francisco Bay Area could cost $700 million to $1.1 billion — a cost that’s typically spread out on ratepayers’ bills.
Putting data centers near the rural solar farms would make at least some of that investment unnecessary, the study found. It would also be roughly 40 percent cheaper for developers than building in Silicon Valley (when also accounting for the comparably cheaper real estate). Adding batteries to the rural site would raise the cost, but would also allow for more computational work and create another source of revenue by selling excess power later in the day.
The rural divide
That said, developers might not want to build in Fresno. Data centers prize proximity to cities to reduce the lag time for end users.
Study author Benjamin Lee, a professor of electrical and systems engineering at the University of Pennsylvania, said rural areas make the most sense for facilities specializing in background data analysis and processing.
“Such computations have deadlines of hours or days,” Lee said in an email. “This means the extra delays of communicating with a more remote data center that computes opportunistically with otherwise curtailed energy may not be perceptible.”
Another potential barrier: local residents. Community opposition to new construction has been growing in urban and rural areas alike, and a POLITICO Poll found that support for new data center construction drops the further away from existing development one gets.
That suggests urban environments may just be easier places to build — even if the grid presents a hurdle.