Tue 10 Nov 2026 / 12:43 ET
Kernel
AI 3 min read

Data center backlash centers on electricity, but not only electricity

Power costs, grid upgrades and on-site generators are making data centers a local political fight, alongside broader concerns about AI and control.

Felix Aranda

By Felix Aranda / Silicon Editor

The data center electricity backlash is not reducible to anxiety about artificial intelligence, though that anxiety is plainly part of the mix. Reporting from Virginia and elsewhere shows a more immediate dispute: who supplies the power for computer campuses, who pays to expand the system, and what nearby residents get in return.

MIT Technology Review reported from Ashburn, Virginia, in the Loudoun County cluster often called Data Center Alley, where most facilities connect to the existing grid. One Equinix site it visited receives power from Dominion Energy and keeps diesel generators for backup. The outlet put the average capacity of operating data centers in that part of Virginia at 54 megawatts.

That is the machinery underneath the political argument. A data center is a building full of servers, networking equipment and cooling systems that turn electricity into computing capacity. How a data center works is also why power supply becomes a local issue: the computers need continuous electricity, and most of that energy eventually becomes heat that must be removed.

Why is electricity driving data center opposition?

Electricity demand can produce visible local consequences. Grid-connected projects may require transmission lines and other infrastructure. BBC News reported accounts of residents worrying that utility bills could rise when new facilities require system upgrades. MIT Technology Review reported that Virginia adopted rules after protests requiring data centers to pay the full cost of new transmission lines, rather than allocating those costs to other customers.

Developers also told MIT Technology Review they were preparing for rules that could cut facilities off from grid power during periods of peak demand. That prospect puts the usual industry language about “the cloud” back where it belongs: on wires, substations and power plants.

The alternative does not make the argument disappear. Projects built in rural areas without nearby grid infrastructure may generate electricity on site, often described as behind-the-meter power. MIT Technology Review reported that the model can use natural-gas turbines, with diesel generators for backup. Those machines avoid drawing directly from the local grid, but they can bring noise and air-pollution concerns. NPR identified a Loudoun County facility powered by gas turbines that create noise.

The scale being sought for AI training makes the supply question harder. MIT Technology Review characterized 300 megawatts as the smallest facilities attractive for that work and reported industry estimates of projects seeking far more. Those figures are not a verdict on any individual proposal, but they help explain why a small-town zoning hearing can become an argument about regional power infrastructure.

Electricity is one part of a broader fight

It would be too neat to call every opponent a ratepayer with a calculator. NPR reported that local opposition crosses party lines and includes fears about job losses, AI risks, opaque development deals and a loss of control over technology. The same report found that some communities support data centers for jobs, lower land taxes and tax revenue. In Loudoun County, NPR reported, Democratic officials have split over whether those benefits outweigh effects on nearby residents.

Electricity is therefore a central mechanism, not a complete explanation. It converts an abstract debate over AI into concrete questions about bills, infrastructure, generator noise and emissions. Companies can promise to bring their own power, as seven Big Tech CEOs did in a non-binding pledge reported by the BBC. Communities still have to live next to the equipment.

This story draws on original reporting from MIT Technology Review.

More AI/

view all ↗