Gone are the days of the Cold War, of nuclear anxiety, of those old animated “Duck and Cover” PSAs we were shown in school: Americans are now more opposed to an AI data center in their neighborhood than a nuclear power plant.
A Gallup survey conducted in March found that 71% of U.S. adults oppose the construction of an AI data center in their local area, with nearly half (48%) strongly opposed while only 27% are in favor. But perhaps the most surprising figure from the survey is that only 53% opposed a nuclear energy plant in their backyard instead, nearly 20 points lower than the data center opposition crowd.
Ever since Gallup began asking the nuclear question in 2001, opposition has never exceeded 63%. This year is Gallup’s first time asking about data centers, and they blew past that ceiling on their first appearance.
“Isn’t that insane?” asked Wannie Park, an energy industry veteran and CEO of PADO AI, an LG NOVA-backed platform that conducts energy management—for data centers. “I think it’s just uninformed stakeholders that aren’t really understanding what the opportunities are.”
Park said the opposition often stems from a lack of knowledge on data centers—and that’s something the industry needs to do a better job of explaining.
“It was just a lack of education. There’s a lack of proper marketing and communication of what this is gonna do. And, I would argue that we haven’t done a good job of that, right?”
Which is… better?
Park emphasized that education is the first step in making sure there’s community support at the local level—or at least, a smaller percentage of people who are vocally opposed.
Perhaps none other than nuclear energy best explains that. The irony is that nuclear power carries risks data centers don’t: meltdown potential, radioactive waste that remains dangerous for thousands of years, and again, the ever constant, growing fear of a countdown clock that remains the center of many a movie to this day. Despite this—perhaps due to a lack of recall from the Cold War days of old—Americans now fear data centers more.
That likely reflects the nature of the impacts. Data center effects are immediate and visible to nearby residents, exhibited in increased noise, traffic, utility bills, and water drawdowns. Nuclear’s worst-case scenarios feel abstract to most people, and its had decades to normalize. Data centers went from obscure to ubiquitous in just a few short years.
On carbon, nuclear is among the lowest-emission energy sources, emitting roughly 12 grams of CO₂-equivalent per kilowatt-hour and rivaling wind. A Cornell study in Nature Sustainability found that by 2030, AI growth alone could produce 24 to 44 million metric tons of CO₂ annually, or the equivalent of adding 5 to 10 million cars to U.S. roads.
Both are water-intensive, but in different ways. Nuclear plants discharge heated cooling water that can disrupt aquatic ecosystems, which is their most significant ongoing environmental impact. Data centers instead consumed an estimated 17 billion gallons in 2023, triple the 2014 figure. By 2030, AI-driven water demand could equal the annual household usage of 6 to 10 million Americans.
U.S. data centers now also account for roughly 4.4% of national electricity consumption, up from 1.9% in 2018, and that figure could reach 12% by 2028. A Bloom Energy report projects total U.S. data center demand will nearly double between 2025 and 2028 — from 80 to 150 gigawatts — like adding a country with Spain’s energy needs in three years.
When Gallup asked an open-ended follow-up in April, half of opponents cited resource strain on resources, including a strain on the water and energy systems as well as a loss of farmland. About 22% raised quality-of-life issues like traffic, while one in five pointed to higher utility bills. Sixteen percent flagged pollution, especially noise. A smaller percentage share expressed unease about AI itself.
Park sees the backlash as a communication failure. “What you hear is a lot of the doom and gloom. We’re gonna use up all the water, rates are gonna go up,” Park told Fortune. “But there hasn’t been a really strong articulated message of what this actually means.” He compared it to the Obama-era smart grid rollout, when residents showed up at city council meetings claiming smart meters would spy on them or cause cancer. “As you fast forward, none of that is there. It was just a lack of education.”
Moratoriums won’t stop the build
Sixty-nine U.S. jurisdictions have enacted moratoriums on data center construction, but Park cautioned that bans simply redistribute development. “The folks developing these sites, they kind of don’t care,” he said. “If you want to shut us down here, we’re gonna go somewhere else.” The economics, he adds, backs up the push: the value compute generates from electricity can be 20 to 100 times the cost of the power itself.
Oftentimes, it’s because the need and demand for compute has already been accounted
“Even if you track things where, oh, hey, I’m XYZ developer, and I’ve got this site that’s going to be developed, it’ll go online in three years, that compute is already pre-booked, like three years ahead of time. That is the amount of demand that’s there.”
Protests have been popping up across the country, with community members standing outside data center construction sites in opposition of the potential negative environmental and financial externalities. Some states, in lieu of a full moratorium, have started implementing “80/20” laws or similar, where hyperscalers would be required to pay more than what they use, in terms of water, electricity, and other low resources.












