A new Gallup poll has laid bare a startling shift in American attitudes toward industrial infrastructure. Seven in 10 adults say they oppose construction of data centers for artificial intelligence in their local area. Nearly half register strong opposition. By contrast, just 53 percent object to a nuclear power plant nearby.
The numbers come from a March survey of more than 1,000 U.S. adults. They mark the first time Gallup has tested sentiment on AI data centers. The findings, released this week, arrive as tech giants race to build ever-larger facilities to train and run advanced models. Opposition runs across party lines, demographics and geography. Democrats show stronger resistance, with 56 percent strongly opposed. Yet 39 percent of Republicans feel the same intensity.
Half of those against data centers point to excessive demands on local resources. Water and electricity top the list. Eighteen percent specifically call out each. Another 16 percent raise pollution worries that include noise, air emissions and impacts on waterways. Quality-of-life complaints follow close behind. Traffic. Crowding. Land better used for other purposes. One in five opponents fear higher utility bills or broader cost-of-living increases.
Environmental anxiety runs deep. Forty-six percent of all respondents say they worry a great deal about the footprint of these facilities. Another 24 percent worry a fair amount. The pattern holds even among those who otherwise support AI development. People see the servers, the cooling towers, the constant hum. They do not see the connection to the apps and services they use daily.
Gallup framed its data-center question in language identical to one it has asked about nuclear plants since 2001. Opposition to nuclear power has never topped 63 percent in that time. The gap reveals something fundamental. Decades of safety improvements, familiarity and perhaps a clearer sense of what nuclear delivers have softened resistance. Data centers feel like pure consumption.
Those who favor local construction lean heavily on economics. Fifty-five percent of supporters mention job opportunities. Thirteen percent speak of added tax revenue. Yet analysts have long noted that once built, these facilities employ relatively few workers. Many also receive substantial tax abatements that drain local coffers. One analysis cited in industry reporting estimates some states lose more than $1 billion annually from such deals.
The Register captured the disconnect in a May 14 article. Emma Fryer, public policy director at data-center operator CyrusOne, told an industry audience last year that “people don’t make a connection between the digital services they depend on every minute of every day of their lives and the fact that providing them every minute of every day of their lives requires industrial-scale infrastructure.” Garry Connolly, founder of Digital Infrastructure Ireland, put it more bluntly: “Most people are fucking scared of AI, like we’re feeding a monster.” Telling residents their neighborhood must host the monster has proven a losing sales pitch.
Recent coverage underscores the scale. A single large AI data center can consume as much electricity as hundreds of thousands of homes. Water demand reaches five million gallons per day in some cases, according to Forbes reporting on the Gallup results. That volume rivals the daily usage of a small city. In northern Nevada, utility NV Energy has told nearly 50,000 Lake Tahoe-area residents they must find new power providers by 2028 so capacity can flow to data centers instead.
Tech companies have responded with ambitious plans. Microsoft struck a deal to restart a unit at the Three Mile Island nuclear plant in Pennsylvania. Amazon, Google, Oracle and others pursue everything from small modular reactors to massive grid connections. Yet these moves have not eased public concern. If anything, the visible push for new power sources highlights how much these facilities require.
Other recent surveys echo the tension. A Consumer Reports poll from late 2025 found 78 percent of Americans worried that data centers would drive up electricity bills. Polling from Climate Power this month showed voters rank higher utility costs and energy consumption as their top fears. Support flips positive only when facilities are explicitly tied to clean power sources. Even then, local siting battles persist.
The industry now faces a classic infrastructure paradox. Demand for AI compute grows without pause. Hyperscalers project data-center electricity use could double or triple by 2030. Projections from the Electric Power Research Institute and others put data centers at 9 percent of national generation within four years. Meeting that load while keeping rates stable and grids reliable demands billions in new generation and transmission. But voters increasingly say the facilities themselves cannot sit near them.
Local governments have begun to feel the heat. In Maine, federal housing funds were pulled from a town after officials considered redirecting them toward data-center development. Rural counties in Virginia, Georgia and Texas that once welcomed the tax base now field organized opposition. Lawsuits, zoning fights and moratorium proposals multiply. What began as scattered complaints has hardened into a coherent NIMBY movement, one more potent than the antinuclear campaigns of past decades.
Demographic breakdowns offer little comfort to developers. Opposition shows no significant variation by age, race, income or whether respondents live in cities, suburbs or countryside. Women register stronger objections than men. Concern about environmental quality predicts resistance more powerfully than worry about energy prices. Those who already fret about the environment oppose data centers at a 78 percent clip. The gap shrinks but does not disappear among those focused on energy affordability.
Politicians notice. State and local races this year could turn on data-center policy. Candidates who champion new facilities risk grassroots backlash. Those who block projects may face pressure from business groups warning of lost economic opportunity. The White House has encouraged data-center developers to build their own power plants behind the meter, a move that sidesteps some grid constraints but raises fresh questions about emissions and permitting.
And yet the need will not vanish. Training a single frontier AI model can require months of nonstop computation on tens of thousands of specialized chips. Inference at global scale multiplies the load further. Without adequate infrastructure, progress stalls. Applications in medicine, scientific research and industrial efficiency remain theoretical. The social license that tech companies once took for granted now hangs in the balance.
Some in the sector hope better communication will help. Others bet on technological fixes: more efficient chips, advanced cooling that recycles water, co-location with nuclear or geothermal sources that deliver carbon-free firm power. A few talk of siting facilities on decommissioned industrial land or even offshore. Success will require more than engineering. It will demand addressing the genuine trade-offs communities face.
The Gallup data leaves little room for wishful thinking. Americans have drawn a line. They accept nuclear plants with all their historical baggage more readily than windowless halls packed with servers. The facilities that power the AI future look, to many eyes, like neighbors that take far more than they give. Bridging that perception gap stands as one of the defining challenges for the technology industry in the years ahead. Failure to do so could slow the very advances proponents promise will transform society for the better.