SDSignal Desk

The data center backlash is about power plants, not a culture war, reporting from Ashburn suggests

· Updated · Signal Desk newsroom

Analysis of reporting by MIT Technology Review. How we work

Image: MIT Technology Review

OpenAI’s Chris Lehane framed opposition as unease about affordability and corporate power. MIT Technology Review’s reporting in Virginia’s Data Center Alley points to something more physical: gigawatts of demand, grid costs and turbines nobody voted for. That changes what fixes work.

Key takeaways

  • The average data center in Loudoun County draws 54 megawatts, more than Denver International Airport.
  • Sites attractive for AI training start around 300 MW; ideal ones approach 6,000 MW, about Chicago’s estimated draw.
  • Virginia now requires data centers to pay fully for new transmission; developers expect peak-demand cutoff rules.
  • Onsite gas turbines swap a bill problem for an air-quality problem; the EPA said xAI’s Memphis turbines violated air law.
  • Siting may shift to sparse regions or abroad, such as Google’s $15 billion Malaysia announcement.

Two readings of the backlash

Two days after the midterms, OpenAI chief global affairs officer Chris Lehane wrote that data centers had become a proxy for unease about affordability, corporate power and people losing control of their futures. MIT Technology Review’s James O’Donnell spent days in Ashburn, Virginia, the world’s largest data center hub, and came away with a simpler read: the backlash is mostly about electricity. Activist Elena Schlossberg told him the fight was always about power.

The distinction decides the fix. A culture war can be met with messaging and community benefits, the instinct behind renaming facilities (AI compute campuses). A physical problem, gigawatts of demand and the turbines or grid upgrades to meet it, can’t be spun away. Opposition has become a cross-party force at the ballot box (uniting America, Texas).

The scale

The average data center in this part of Virginia draws 54 megawatts, more than Denver International Airport. The smallest sites attractive for AI training are around 300 megawatts; ideal ones approach 6,000, roughly the estimated draw of Chicago. Loudoun’s centers were built at a steadier pace than the current national expansion and mostly connect to Dominion Energy, which relies heavily on natural gas. The piece also notes misinformation fuels some opposition; most facilities use far less water than many think.

Two paths, two backlashes

Grid-connected sites, like the Equinix facility O’Donnell toured, strain shared infrastructure, which is why Virginia now requires data centers to pay fully for new transmission rather than spreading the cost. Developers told him they expect rules cutting them off during peak demand, which is why flexible-load efforts are growing (Emerald AI, NVIDIA DSX).

Bringing your own gas turbines to rural land swaps a bill problem for an air-quality problem. The EPA said xAI’s Memphis turbine setup violated air quality laws, and a Clean Air Act suit is proceeding (NAACP case). Anthropic has reportedly agreed to pay $3.4 billion for gas turbines, described as the largest power purchase by an AI company.

The honest path

Many advances that improved modern life needed enormous electricity, and AI capacity has real value. But when private companies build generation at utility scale, communities have always expected a say in where it goes. Skipping that is what makes them, in the article’s phrase, horrible neighbors.

If onsite power becomes the norm, facilities go where the fewest people can resist: sparse regions like the Permian Basin, or abroad. Google announced $15 billion for data centers in Malaysia the day after the election. That exports the backlash to people with less leverage. The honest path is boring: pay the full cost of grid upgrades, take emissions seriously and give host communities real power over siting, which is likelier to work than talk of data centers in space, which Elon Musk floated hours after the results. More in our business coverage and regulation coverage.

Who feels it

Local communities
Electricity and air quality are the core trade; organized pressure works.
AI labs and hyperscalers
Power is the binding constraint, and PR won’t loosen it.
Utilities and regulators
Curtailment and cost allocation become central tools.
Sparse regions and markets abroad
Next in line for onsite-power campuses.

What to watch

  1. Whether anti–data center candidates follow through in office.
  2. Virginia peak-demand cutoff rules.
  3. More behind-the-meter turbine deals.
  4. Siting shifts to sparse US regions and Malaysia.

FAQ

Why are people opposing data centers?

Reporting from Ashburn suggests it is mostly electricity: grid strain, costs and onsite power plants, more than a culture war.

How much power does a data center use?

The average Loudoun County site draws 54 MW; AI training sites start around 300 MW and ideally approach 6,000 MW.

Who pays for data center transmission in Virginia?

Virginia now requires data centers to fully pay for new transmission lines.

Read the original

Continue at the source.

MIT Technology Review

More on AI

All AI coverage →