The People vs. AI: Why America’s Growing Backlash Against Data Centers Signals a Broader Tech Reckoning

From Virginia’s megacampus communities to Mississippi’s courtrooms, a cross-partisan coalition is demanding that America slow down and ask who, exactly, benefits from the AI revolution—and at what cost.

One icy morning in February, nearly 200 people gathered in a Richmond, Virginia church before dawn. They came from rural farms and suburban subdivisions, from the valleys of Botetourt County and the exurbs of Washington, D.C. Republicans stood alongside Democrats. Pastors sat next to environmental engineers. And though they had arrived carrying different anxieties—higher electricity bills, fouled groundwater, the low industrial hum that now keeps rural families awake at night—they shared a single, galvanizing conviction: that the AI industry’s appetite for infrastructure had outpaced its accountability to the people who must live beside it.

“Aren’t you tired of being ignored by both parties, and having your quality of life and your environment absolutely destroyed by corporate greed?” state senator Danica Roem asked the crowd. The standing ovation that followed was the sound of something new crystallizing in American political life. What is causing AI backlash? The short answer: communities feel they are absorbing all of the costs—environmental, economic, democratic—while the profits flow elsewhere.

The activists marched to the state capitol, where state delegate John McAuliff offered what may be the most honest six-word summary of the public’s relationship with the AI boom: “You’re getting a sh-t deal.”

AI Pessimism Is Not a Fringe Position

Pundits frequently portray skepticism of AI as technophobia. The data tell a different story. According to Pew Research Center’s 2025 AI Attitudes Survey, five times as many Americans are concerned as are excited about the increased use of AI in daily life—a ratio that has widened over the past two years, not narrowed, as the technology has become more pervasive. Majorities believe AI will worsen creative thinking, erode meaningful human relationships, and degrade decision-making. More than half say AI poses a serious risk of spreading political misinformation. These are not marginal anxieties; they are mainstream ones.

Internationally, the United States is among the most skeptical rich nations, a finding that surprises many observers who assume American technological exceptionalism translates into enthusiasm. It does not. The country that houses the majority of the world’s AI compute infrastructure is also one of the most apprehensive about its consequences. The table below, drawn from Pew’s cross-national data, illustrates the divide.

Table 1: AI Optimism vs. Pessimism by Country (Pew Research, 2025)

Source: Pew Research Center, “AI Attitudes Survey” 2025. Net sentiment = % excited minus % concerned.

The pattern is stark: wealthy democracies with established labor protections and high wages view AI as a threat to existing quality of life; rapidly developing economies, where AI offers tangible prospects of economic leapfrogging, are markedly more enthusiastic. This is not irrational on either side. It reflects a fundamental asymmetry in who stands to gain from the present deployment trajectory.

Ground Zero: Why Virginia Became the Symbol of Bipartisan Resistance to AI Development

Virginia’s Loudoun County—nicknamed “Data Center Alley”—hosts more data center capacity than any comparable geography on Earth, accounting for roughly 70% of the world’s internet traffic at any given moment. The concentration has brought tax revenue and construction jobs. It has also brought something else: a relentless surge in electricity demand that is reshaping the state’s energy grid and the household budgets of people nowhere near a server rack.

As NPR reported, residential customers in Dominion Energy’s service territory—which covers much of northern and central Virginia—have seen bills climb as the utility pursues new generation capacity to feed data centers whose power purchase agreements are structured to benefit large commercial customers first. Rural residents, already stretched by post-pandemic inflation, are being asked to help finance infrastructure they will never use.

The activists in homemade shirts—“Boondoggle: Data Center in Botetourt County”—were not opposing innovation in the abstract. They were opposing a specific regulatory and financial arrangement in which local residents bear external costs while shareholders and cloud tenants capture value. This is a data center backlash in Virginia 2026 that has become a template: similar coalitions are emerging in Indiana, Arizona, Nevada, and rural Texas.

Stalled Projects and the $98 Billion Question

The activism is having measurable economic effects. According to industry trackers, approximately $98 billion in planned U.S. data center projects were stalled or subject to significant regulatory delay in Q2 2025, with activism and permitting challenges cited as primary factors. The table below breaks down the stalls by state.

Table 2: Stalled U.S. Data Center Projects by State (Q2 2025, est.)

Source: Industry estimates, state regulatory filings, Q2 2025. Figures rounded.

The delays are not killing AI development—they are redirecting it, to jurisdictions with cheaper power, laxer environmental oversight, and weaker community organization. This is the classic spatial arbitrage of industrial capitalism: the factory moves when the community pushes back. Whether that dispersal is good or bad depends on whether you are in the community that succeeds in pushing or the one that inherits the factory.

The Legal Front: xAI in Mississippi and the Clean Air Act Test

The backlash has found its way into federal courts. Litigation against Elon Musk’s xAI facility in Memphis, Mississippi alleges violations of the Clean Air Act, with plaintiffs arguing that the company’s backup generators—operated as primary power sources during periods of grid stress—emit pollutants at levels requiring permits the company does not possess. The case is being watched nationally as a potential precedent for whether AI companies can claim de facto exemptions from environmental law by classifying their continuous operations as “emergency” use.

If plaintiffs succeed, the implications for the industry would be significant: hundreds of facilities across the country rely on similar generator arrangements. Environmental lawyers note that the xAI case may open the door to Clean Air Act enforcement against data centers at a scale the sector has never faced. “This is not a fringe environmental argument,” one former EPA enforcement official told The Guardian. “These are the same rules every other industrial emitter has to follow.”

Global Pressure: The AI Impact Summit 2026 and Trade Deal Disruptions

The U.S. backlash is not occurring in isolation. At the AI Impact Summit 2026 in New Delhi, delegates attempting to finalize a framework for AI-driven trade agreements—covering data localization, intellectual property, and labor displacement provisions—were disrupted by Youth Congress activists protesting what they called a “digital colonialism” framework that would concentrate AI-derived wealth in American and European technology companies while requiring developing nations to provide low-cost data and labor. The protests did not collapse the summit, but they delayed a planned joint communiqué and forced a revision of language around benefit-sharing mechanisms.

The New Delhi disruptions signal that AI skepticism is globalizing even as AI enthusiasm in some emerging economies remains strong. The distinction, activists argue, is between optimism about AI as a technology and skepticism about the terms on which it is being deployed. These are separable positions, and conflating them—as advocates for the industry often do—obscures the legitimate grievance at the heart of the backlash.

Bernie Sanders and the Case for a Moratorium

Senator Bernie Sanders has proposed what he calls a “moratorium on AI data center development” to “slow down the revolution and protect workers,” arguing that the pace of deployment has deliberately outrun the capacity of democratic institutions to govern it. The proposal, greeted with skepticism by economists who note that unilateral moratoriums invite capital flight, has nonetheless reframed the debate: instead of asking “how do we govern AI?,” it asks “should we be allowed to pause and decide?”

Sanders’ intervention illustrates the unusual political geography of AI resistance. As The Washington Post has documented in its polling analysis, concern about AI does not sort neatly along partisan lines. MAGA Republicans who distrust Silicon Valley’s cultural influence and democratic socialists who distrust its economic power converge, awkwardly but consequentially, on the same demand: slow down.

The AI Environmental Impact on Communities: What the Data Show

Beneath the politics lies a set of empirical disputes that deserve more rigorous public attention than they typically receive. The AI environmental impact on communities operates along three axes:

• Energy: A single large language model training run can consume as much electricity as several hundred U.S. homes use in a year. The inference costs—running the model millions of times daily—are ongoing and growing.
• Water: Cooling systems for major data centers can consume millions of gallons of water annually, a serious concern in drought-stressed regions like Arizona’s Phoenix metro, where several proposed facilities face water-availability challenges.
• Noise: Industrial cooling equipment operates continuously, producing low-frequency noise that affects nearby residents. Unlike construction noise, it does not stop; it is the permanent ambient condition of living near a data center campus.

None of these harms are, in principle, unmanageable. They are, however, being managed poorly—or not at all—under current regulatory frameworks that were not designed for facilities of this scale or this permanence.

AI Job Displacement: The Other Fear Nobody Talks About Plainly

Community opposition to data centers is partially a proxy for a deeper anxiety: public concerns about AI job loss. When residents object to a data center, they are often also expressing a fear that they are watching the physical infrastructure of their economic replacement being built in their backyard. Data centers employ relatively few people for their footprint—a facility consuming hundreds of megawatts may have a permanent workforce of dozens—while the AI systems they power are actively displacing white-collar and creative jobs in ways the public perceives, even if economists debate the magnitude.

A 2025 McKinsey analysis estimated that generative AI could displace 12 million workers in the United States by 2030 in occupations ranging from customer service to legal research to graphic design. Meanwhile, the TIME investigation into public AI pessimism found that workers in affected industries are not merely worried about losing their jobs; they are worried about losing the sense of purpose and mastery that skilled work confers. This is not easily compensated by a retraining voucher.

What Good Policy Would Look Like

The backlash is real, its grievances are legitimate, and it will not be resolved by dismissing protesters as technophobes or promising trickle-down prosperity from the AI economy. Several policy directions merit serious attention:

• Community benefit agreements: Require data center developers to negotiate directly with affected municipalities before permitting, covering utility cost guarantees, noise mitigation, water use limits, and local hiring commitments.
• Energy cost isolation: Regulatory reform to prevent data center power purchase agreements from socializing costs to residential ratepayers. Industrial customers that drive demand spikes should pay their proportional share of grid expansion costs.
• Environmental permitting reform: Close generator loopholes that allow data centers to operate industrial combustion equipment under emergency-use classifications. Require full Clean Air Act permits for any facility operating generators more than a defined annual threshold.
• AI worker transition funding: Establish a dedicated federal fund—potentially capitalized by a small levy on AI compute revenues—for worker retraining, wage insurance, and economic transition support in communities demonstrating displacement.
• International benefit-sharing frameworks: Pursue multilateral agreements that require AI platform companies to contribute to development funds in countries where their systems are deployed and their training data was sourced.

The Reckoning Is Already Here

The people who gathered in that Richmond church in February were not anti-technology. Most of them use smartphones, stream video, and google their symptoms before seeing a doctor. What they object to is a specific power arrangement: one in which transformative decisions about infrastructure, energy, water, and labor are made by a small number of corporations and ratified by governments responsive to lobbying, with communities consulted—if at all—after the cement has been poured.

AI will not be stopped. The economic incentives are too powerful, the competitive pressures too acute, and the genuine benefits in healthcare, scientific research, and educational access too real to dismiss. But “AI will not be stopped” is different from “the current deployment model is optimal or just.” The backlash against data centers is the most visible symptom of a reckoning the industry has been avoiding: that legitimacy, in a democracy, must be earned—not assumed.

As The New Republic argued in its analysis of local AI rebellions, data centers have become “the enemy we’ve all been waiting for” not because they are the worst thing that corporations do to communities, but because they are immediate, visible, and undeniable. You can see the construction. You can hear the cooling fans. You can open your utility bill.

The AI industry’s best advocates understand this. They know that social license, once forfeited, is very expensive to recover. The question is whether the companies building this infrastructure will engage with the communities affected before they are forced to—or whether they will wait for the lawsuits, the moratoriums, and the legislative backlash to compel them to a table they could have come to voluntarily.