Your neighborhood hates your data center

There are more than 4,000 AI data centers operating across the United States, with another 3,000 planned or under construction. The buildout is massive, accelerating, and increasingly impossible to ignore. Because the physical cost of AI is no longer abstract. It is showing up in your electricity bill, your water supply, and your neighbor's front-yard protest sign. For years, data centers were the invisible backbone of the internet, quietly humming in industrial parks. Now they are political flashpoints. Communities from Virginia to Arizona are pushing back, and for the first time, the industry is being forced to answer a question it has long avoided: what happens when the infrastructure behind AI becomes too expensive for the places that host it?

The scale of the problem

To understand the backlash, you first need to understand the scale. In 2023, U.S. data centers consumed roughly 176 terawatt-hours of electricity, about 4.4% of the nation's total. By 2028, that number could triple to 580 TWh, or 12% of national consumption. The International Energy Agency reported that data center electricity demand soared 17% in 2025 alone, and according to Fortune, data centers drove half of all new U.S. electricity demand growth that year. Stanford's 2026 AI Index report put it in starker terms. AI data center power capacity has risen to 29.6 gigawatts, roughly what it takes to power the entire state of New York at peak demand. Training a single frontier model like Grok 4 generated an estimated 72,816 tons of CO2, the equivalent of driving 17,000 cars for a year. And the water used just for GPT-4o inference may exceed the drinking water needs of 12 million people. These are not projections for some distant future. This is happening now.

Your electricity bill is the front line

The most immediate way most people feel the impact of data centers is through their power bills. In Virginia, data centers consumed about 26% of the state's total electricity supply in 2023. That concentration has consequences. Bloomberg reported that wholesale electricity costs near data center clusters have risen as much as 267% compared to five years ago. A Consumer Reports investigation highlighted residents like John Steinbach in Manassas, Virginia, whose electricity bill spiked from around $100 to $281 in a single month. Utilities are planning massive infrastructure investments to keep up. A PowerLines report found that U.S. utilities plan to spend $1.4 trillion over the next five years, more than 20% above their 2025 projections, with much of that driven by data center demand. Those costs are typically passed on to ratepayers. The Energy Information Administration forecasts the strongest four-year growth in U.S. electricity demand since 2000, fueled almost entirely by data centers. For residents who never asked for a data center in their neighborhood, the bill arrives anyway.

The water nobody talks about

Electricity gets the headlines, but water is the underreported story. Data centers run hot. Rows of servers operate around the clock, and most facilities rely on constantly circulating water to prevent overheating. A large data center can consume up to 5 million gallons of water per day, as much as 16,000 average U.S. households. The World Resources Institute estimates that by 2028, AI-related data centers could require up to 32 billion gallons of water annually, enough to support roughly 360,000 households. The problem is not just the volume. It is where these facilities are being built. In Newton County, Georgia, a single Meta data center uses 500,000 gallons per day, about 10% of the entire county's water consumption. The county continues to field requests for new permits, some of which would use up to 6 million gallons daily, more than doubling what the whole county currently consumes. In Nevada, more than half of the state's groundwater basins are already over-allocated, meaning more water is promised on paper than actually exists underground. Arizona faces a similar bind, with 150-plus data center projects advancing in a state where drought is a permanent feature of life, not a temporary disruption. A 2026 study highlighted by KUNC found that data center water usage remains remarkably opaque. Most facilities do not publicly disclose how much water they consume, making it nearly impossible for communities to plan. As one researcher put it, "We cannot write policy if we don't have data."

NIMBYism meets AI

The pattern is familiar. Cell towers, wind farms, nuclear plants, all faced community resistance when they moved from concept to construction. Data centers are following the same trajectory, but faster. Data Center Watch reported that opposition rose 125% in the second quarter of 2025 alone. An estimated $98 billion in projects were blocked or delayed, more than the total for all previous quarters since 2023. CBRE confirmed that new data center capacity under construction in primary U.S. markets fell for the first time since 2020, even as demand for compute power surged. The resistance is not fringe. A nationwide poll found that only 44% of Americans would welcome a data center in their community, making them less popular than gas plants, wind farms, or even nuclear facilities. In Virginia, a Washington Post-Schar School poll found that voters have turned sharply against data centers after years of welcoming them. Specific battles tell the story. In Prince George's County, Maryland, a proposal to build a data center on the site of the old Landover Mall drew 20,000 petition signatures and prompted the county to impose a six-month moratorium on all data center permits. In Prince William County, Virginia, homeowners sued to block the Digital Gateway mega-data center project, citing threats to the Manassas Battlefield and permanent environmental harm. In the Midwest, residents of one community removed four local council members and are now aiming to recall the mayor over a planned facility. At the state level, New York has proposed a three-year moratorium on new data centers. Georgia, Maryland, and Oklahoma have introduced similar pause legislation. Senator Bernie Sanders called for a national moratorium, arguing it would help America "catch up" and "make sure that the benefits of these technologies work for all of us." This is not anti-technology sentiment. It is a rational response to real costs being imposed on communities that see little in return. As Harvard researcher Laura Hecht-Felella noted, data centers "don't bring meaningful economic development, especially in the form of jobs." Her research concluded that they are, on balance, "a bad deal for communities on the local level."

The sustainability paradox

Here is the irony that is hard to look away from: the same companies that publish annual sustainability reports and pledge carbon neutrality are building the most energy-intensive infrastructure in history. Stanford's AI Index flagged environmental costs as one of its 12 key takeaways for 2026. This is no longer a niche concern. It is a mainstream acknowledgment that the trajectory of AI development is on a collision course with climate goals. The Foundation Model Transparency Index, which tracks how openly AI companies disclose details about their models, saw average scores drop from 58 to 40 points year over year. The most capable models now disclose the least. When companies will not even tell you how much energy their models consume, it is hard to take their sustainability commitments at face value.

What Singapore can teach us

Singapore offers an interesting counterpoint. The city-state is land-scarce, energy-constrained, and sits in one of the worst climates on Earth for data center cooling, with average temperatures of 33°C and humidity above 80%. In 2019, Singapore did something no other major digital hub had done: it imposed a full moratorium on new data center construction. The ban lasted until 2022, when the government began selectively issuing permits under strict sustainability requirements. The latest round, DC-CFA2, allocates at least 200 megawatts of new capacity but mandates 50% green energy usage, among the strictest requirements in Asia-Pacific. Singapore cannot brute-force its way into AI infrastructure the way the U.S. can. With only 734 square kilometers of land and more than 1.4 gigawatts of existing data center capacity, every new megawatt has to be justified. Keppel, a local conglomerate, has proposed building a floating data center park using spare shipping docks, using seawater cooling to increase efficiency by up to 80% while avoiding potable water entirely. The lesson is not that Singapore has solved the problem. It is that constraints force better solutions. When you cannot simply build more, you have to build smarter. The U.S. has the land and energy capacity to keep building, but the question is whether it should, at least not without the kind of guardrails Singapore has implemented.

The question nobody wants to answer

All of this leads to an uncomfortable question: what if the environmental cost of AI is permanently higher than the value it creates for most use cases? Not all AI applications are equal. Training a frontier model to solve PhD-level physics problems is a different proposition from generating marketing copy or powering a chatbot. Yet both draw from the same pool of energy and water. The marginal cost of each additional query may be small, but at the scale of billions of users, the aggregate cost is enormous. A Cornell study found that locating data centers in regions with lower water stress and improving cooling efficiency could reduce water demands by about 52%. Combined with grid and operational improvements, total water reductions could reach 86%. The technology exists to make data centers significantly less harmful. The question is whether the economics and the political will exist to deploy it. Right now, the incentive structure points the other way. Speed matters more than efficiency. Being first to deploy matters more than being sustainable. Tax breaks flow to projects that promise jobs and revenue, even when the evidence shows that the jobs are minimal and the costs are real.

What this means going forward

The backlash against data centers is not going away. If anything, it is accelerating. The combination of rising electricity bills, water stress, declining transparency, and minimal local economic benefit has created a political environment where "more data centers" is no longer an easy sell. This does not mean AI development will stop. But it does mean that the where and how of building AI infrastructure will become as important as the what. Communities are demanding a say, and they are winning. The $98 billion in blocked or delayed projects is not a temporary speed bump. It is a signal. For the tech industry, the path forward probably involves some combination of better cooling technology, genuine investment in renewable energy (not just purchasing offsets), transparent reporting of resource consumption, and meaningful community benefit agreements that go beyond tax revenue. For the rest of us, it is worth paying attention to what is being built in our backyards, and what it costs. The cloud is not weightless. It sits on land someone owns, drinks water someone needs, and runs on electricity someone pays for. The communities pushing back are not anti-progress. They are asking a reasonable question: who bears the cost?

References