The AI industry's dirtiest secret is not how much energy its models consume. It is what happens to the neighborhoods where that energy gets consumed.

Senator Bernie Sanders has called for a nationwide moratorium on data center construction, citing environmental and ethical concerns about AI. He is not alone. More than 140 activist groups across 24 states are now campaigning to stop or slow data center development, according to a new report from DataCenterWatch. And two Massachusetts cities, Lowell and Everett, just became the latest to take action, with Lowell passing a yearlong moratorium and Everett campaigning for an outright ban.

The numbers behind the backlash are brutal. U.S. data centers consumed as much energy as the entire country of Pakistan in 2024. That demand is projected to more than double by 2030. Around 39% of natural gas-fired power plants built last year served data centers. And a Harvard analysis found that a single Virginia data center, operating within legal air quality limits, still projected $99 million per year in health damages to local residents.

Read that again. A facility operating legally was projected to cause $99 million in health damages annually. That is the gap between what the law permits and what communities can tolerate.

The NIMBY Problem That Big Tech Cannot Buy Its Way Out Of

Google, Microsoft, Amazon, and Meta are all facing the same problem: they need massive amounts of new compute capacity, and the communities where they want to build it are saying no. Loudly. The complaints are consistent across states: noise pollution from generators that run 24/7, air pollution from diesel backups, water consumption that strains local supplies, and electricity demand that drives up household bills.

This is not abstract policy debate. Lowell City Councilor Kim Scott said the moratorium was not about stopping development. It was about giving the city time to review zoning rules before a facility appears and cannot be undone. That is the sound of local government catching up to an industry that moved faster than regulation.

Here is the problem for Big Tech: AI models are getting bigger, compute demand is accelerating, and there is no cloud without physical buildings. Every data center moratorium is a brake on the AI buildout. And with 140 groups organized across 24 states, these brakes are multiplying faster than lobbying budgets can handle.

The AI industry promised that bigger models would change the world. It did not mention that the world would have to live next to the generators.