Trying to End the Power Suck

Over 230 national and community groups across the U.S. have called on Congress for an immediate nationwide halt on building new data centers until strong new safeguards against the environmental, economic and social impacts of artificial intelligence are in place. Such a ban would stop the AI boom in its tracks and reshape the entire tech industry.

Resistance to the next wave of AI data centers has been growing as the scale of Silicon Valley’s AI build-out — and their massive energy and water demands — becomes increasingly clear. 

The new coalition — including Food & Water Watch, Physicians for Social Responsibility, Greenpeace, Friends of the Earth and local groups in 35 states — says the number of U.S. data centers will triple by 2030, consuming enough energy to power 28 million households, with 56% of that power coming from fossil fuels.

These data centers will further require as much water as 18.5 million households. Many will be in states with acute water shortages, such as Texas and Arizona, a problem certain to get worse with rising temperatures.

Alongside the environmental harms, it may be AI’s impact on electricity bills that receives the most attention in Congress — and causes more problems for AI companies. The coalition says electricity bills have risen 21%, mostly because of AI data centers, as 46% of Americans now say the cost of living is “the worst [they] can remember.”

“At a time when millions of Americans are already struggling with soaring utility costs, the sudden explosion of the Big Data industry represents an existential threat for communities ill-equipped to handle the massive environmental and economic hardships these data centers inflict,” said Wenonah Hauter, executive director of Food & Water Watch. “The only prudent action is to halt the unfettered expansion of this dangerous industry … before it’s too late.”

Six big tech companies are building massive data centers that will consume between 1 gigawatt and 5 gigawatts of electricity; 1 gigawatt being enough to power a city of 750,000 people. This year these six companies alone have spent around $400 billion on AI data centers, and next year they intend to spend even more. With demand from tech companies seemingly limitless, dozens of other companies and investors are building more huge AI data centers wherever they can find the requisite supplies of land, energy and water.

AI data centers are also driving the explosive growth of semiconductor factories, or “fabs,” that make the microchips to fill them. These massive industrial plants not only require vast amounts of energy and water, but also spew out toxic waste, including PFAS “forever” chemicals.

Despite the huge business interests driving the AI boom, campaigns against data centers have had notable successes, as Zeb Larson wrote in Truthdig this week:

Sixty-four billion dollars of data center construction was halted or slowed between May 2024 and March 2025; that figure grew to $98 billion between March and June of this year alone. But it’s going to need to keep growing and even accelerate.

AI companies face further headwinds as they scramble to find energy for their new data centers. By 2028, a shortfall of 19 gigawatts of power is forecast for data centers, equivalent to about twice the power needed by New York City.

AI companies are resorting to extreme measures just to power their data centers. Microsoft plans to reopen the Three Mile Island nuclear power station, closed in 1979 after a radioactive meltdown, while Elon Musk has trucked in dozens of unlicensed mobile gas generators to power his “Colossus” data center in Memphis, Tennessee, causing protests over noise and pollution from nearby residents. Other data centers are even deploying converted old jet engines, which are twice as polluting as conventional gas turbines, to generate electricity.

As more people realize the full cost of the AI boom, Silicon Valley’s “AI race” may in the end simply run out of gas.