As tech companies pour billions into AI, they face a growing risk: it’s incredibly unpopular. The backlash isn’t surprising. Do we really want technology that eliminates jobs, uses huge amounts of energy and water, erodes our ability to think, and poses a wide range of other risks, from making it easier for the government to surveil citizens to encouraging teen suicide? AI could obviously also be a useful tool—potentially helping develop better drugs and climate solutions, for example—but the haters have a point. And while AI may seem unstoppable, the lack of support means that it’s getting much harder for tech companies to build the new data centers that they desperately want.
Americans don’t want to live by data centers
In less than a year, opinions about data centers have quickly shifted. In a Heatmap survey last August, 24% of respondents said that they “strongly opposed” a data center being built near where they live. In Heatmap‘s latest survey in May of more than 4,000 voters, that number had jumped to 55%. In the last nine months, strong opposition more than doubled.
In a Gallup poll in May, 71% of Americans said they would be opposed to a new AI data center being built in their area, with nearly half of them strongly opposed. (It’s worth noting that in a same survey, only 53% of respondents said that they would oppose a new nuclear power plant in their area.) It’s hard to untangle how much of the data center opposition is tied specifically to the fact that a project powers AI versus more immediate concerns like electric bills; in the Gallup survey, half of opponents cited impacts on resources such as water and energy, and a smaller fraction mentioned their dislike of AI. But it’s also true that many proposed data centers wouldn’t need to exist if not for AI, and voters know that. The fact that AI creates slop and can seem unnecessary isn’t helping.
In Virginia, a hotspot for data centers, support for local projects dropped from 69% in 2023 to 35% this year, and one county recently abandoned plans to build a massive campus with as many as 37 data centers. A city in California recently became the first to ban new data centers. Others across the country have passed temporary moratoriums. In a town in Missouri, four city council members were voted out after they approved a $6 billion data center. For data center developers that were already struggling to source energy and water and get permits, it’s going to continue to get harder to build.
AI is less popular than ICE
Only 26% of voters view AI positively and 46% view it negatively, according to a national survey of 1,000 voters in March by NBC. With a net favorability of minus 20, AI was less popular than US Immigration and Customs Enforcement or President Trump. (Since that survey, Trump’s low approval ratings have gotten even worse, so it’s possible AI may now have an edge.) Younger voters, age 18-34, gave AI a lower favorability rating of minus 44.
Fifty-seven percent of respondents said that they thought the risks of AI outweighed the benefits. A third of those polled also said that neither political party was doing a good job of with AI policy.
Americans are becoming more concerned than excited about AI
In a series of Pew surveys that started in 2021, public enthusiasm about AI has been shrinking. In 2021, 37% of Americans said that they were more concerned than excited about the technology. Now, 50% say that they’re more concerned; only 10% are more excited than concerned.
Fifty-seven percent say that the societal risks of the technology are high. Around half of respondents in the 2025 survey said that they thought AI would make it harder to think creatively or form meaningful relationships with other people.
Voters think AI is moving too fast
The majority of Americans, 65%, think that AI is developing too fast, according to a May poll from YouGov and the Economist. (In a poll earlier in the month, that number was slightly higher, at 71%.) That’s true for both Republicans and Democrats, though Democrats are a little more worried. A majority of respondents said it’s unlikely that economic gains from AI will benefit everyone. Around three-quarters were worried—ranging from slightly worried to very worried—about AI’s impact on jobs.
It’s obviously not the first technology to face a backlash. In the 15th century, some critics argued that new printing presses would spread misinformation and immorality. In the late 1800s, a journalist warned that the proliferation of new magazines would rot brains so people would be unable to concentrate and would “think like birds, in little broken thoughts.” Some risks might be overblown, but others might not be.
Any AI booster will point to the fact that when other technologies threatened jobs, that eventually led to new careers, but it’s possible that the magnitude of changes from AI will be different. (Even OpenAI CEO Sam Altman has warned about the job apocalypse, though recently changed his tune.) We’re building gas plants to power data centers when we’re already out of time to cut emissions. Anthropic just argued that AI labs may need to slow down because of the risk that AI could start to improve itself in ways that could harm society.
The pace of development may be the biggest challenge. As tech companies race to build more powerful systems, can governments react in time?