In the last two years, the United States has embarked on the largest infrastructure buildout in modern history. Driven by artificial intelligence, cloud computing, and the insatiable appetite for digital services, data centers are springing up across the country at a staggering clip. An estimated $6.7 trillion in global investment is projected through 2030 to meet compute demand. In the U.S. alone, data centers consumed roughly 183 terawatt-hours of electricity in 2024—4.4 percent of the nation's total—and that share is expected to nearly triple by 2030.
This is not a niche issue. This is the defining land-use question of the next decade. And right now, we are getting it wrong far more often than we are getting it right.
The Backlash Is Real—and Earned
Between May 2024 and March 2025, approximately $64 billion worth of U.S. data center projects were either blocked or delayed by local opposition. By the second quarter of 2025, that resistance had surged by 125 percent, stalling or halting an estimated $98 billion in planned development. These are not fringe protests. These are county commissioners, town councils, and ordinary residents saying enough.
In Virginia—home to the world's largest concentration of data centers—forty-two activist groups joined forces under the Data Center Reform Coalition. At least 25 planned projects were cancelled in 2025, nearly four times the number scrapped the year before. In Warrenton, voters unseated town council members who had supported a proposed Amazon facility. In Pittsylvania County, a rural community successfully defeated a large-scale proposal that would have industrialized agricultural land.
Georgia has followed a similar trajectory. At least ten municipalities enacted moratoriums on data center construction in 2025, including Coweta, Clayton, DeKalb, Pike, Lamar, and Troup counties. In Coweta County, a proposed $17 billion hyperscale facility—codenamed "Project Sail"—drew such fierce public outcry that county officials delayed a vote on new regulations. In rural Twiggs County, residents demanded answers about impacts to water, power, and the local environment. A state lawmaker introduced a bill in January 2026 for a potential statewide moratorium.
The pattern is unmistakable: communities are not opposed to progress. They are opposed to being steamrolled.
Water: The Crisis Nobody Budgeted For
A medium-sized data center consumes approximately 110 million gallons of water annually—comparable to the yearly usage of about 1,000 households. Larger facilities can draw up to 5 million gallons per day, equivalent to the water needs of a town of 10,000 to 50,000 people. In 2023, direct water consumption by U.S. data centers reached an estimated 66 billion liters. Indirect consumption—primarily from the electricity generation required to power these facilities—was roughly 800 billion liters.
And yet, roughly two-thirds of data centers built since 2022 have been sited in water-stressed regions. Texas, one of the most concentrated data center markets in the country, was projected to use 49 billion gallons of water for data centers in 2025, with estimates ballooning to 399 billion gallons by 2030. In Tucson, Arizona, residents have protested proposals that would divert millions of gallons of drinking water for cooling purposes.
This is not a hypothetical problem. It is happening now, in communities where aquifers are already declining and municipal water systems are already strained.
Power: Straining a Grid That Was Already Fragile
Data centers are the single fastest-growing category of electricity demand in the United States. S&P Global's 451 Research division projects that U.S. data centers will require 61.8 gigawatts of grid power by the end of 2025—a 22 percent increase over the prior year—expanding to 75.8 gigawatts in 2026. By 2030, that figure is projected to reach 134 gigawatts, nearly tripling in five years.
The consequences are already visible. In Virginia, data centers consumed 26 percent of the state's electricity in 2023. In Ireland—a cautionary tale for any region pursuing aggressive data center recruitment—the sector accounted for 21 percent of national electricity consumption in 2022, with forecasts pointing to 32 percent by 2026. Capacity prices in the PJM Interconnection, which manages the grid across much of the eastern U.S., surged 833 percent between the 2024–25 and 2025–26 delivery years.
When a single industrial user can consume more electricity than a mid-sized city, siting decisions become energy policy decisions. And they should be treated as such.
What "Right Place" Actually Means
We at the EPR Foundation believe the data center boom is not inherently destructive. Digital infrastructure is essential to the modern economy. But essential does not mean exempt from scrutiny—and it certainly does not mean that every greenfield site, every rural community, and every water-stressed county is an appropriate location.
"Right place" means, at minimum:
Sites with existing industrial infrastructure. Brownfields, retired manufacturing facilities, decommissioned power plants, and former industrial corridors offer electrical capacity, road access, and zoning compatibility that greenfield sites do not. The EPA published guidance in September 2025 promoting data center development on remediated brownfield and Superfund sites. In North Carolina, AI computing firm Bit Digital is converting a former factory in Madison into a 24-megawatt data center, leveraging existing power and fiber infrastructure. In Northwest Indiana, an initiative has identified at least eight brownfield sites along a Quantum Fiber Corridor for potential data center redevelopment. These are models worth replicating.
Regions with surplus power generation or renewable energy access. Siting data centers in areas with abundant wind, solar, or hydroelectric resources—or near retiring fossil fuel plants with existing grid connections—reduces the strain on overburdened transmission networks and can accelerate the transition to cleaner energy. The worst possible outcome is building new natural gas generation specifically to feed data centers in regions that are already struggling to decarbonize.
Areas that are not water-stressed. This should be obvious, but the track record suggests it is not. Water Usage Effectiveness (WUE) metrics matter, but they are meaningless if the facility is sited atop a declining aquifer. Closed-loop cooling systems, immersion cooling, and air cooling technologies can reduce water consumption by up to 91 percent compared to traditional evaporative methods. Microsoft has announced zero-water cooling designs for pilot facilities in Arizona and Wisconsin, expected online by late 2027. The technology exists. The question is whether developers will deploy it—or whether they will continue choosing the cheapest option until a community forces them to do otherwise.
What "Right Way" Actually Means
Siting is only half the equation. The process by which data centers are proposed, permitted, and built matters just as much as the location.
Transparency from day one. Communities should not learn about a proposed data center from a leaked zoning application or a cryptic public notice. Developers should be required to disclose projected water consumption, electricity demand, noise levels, and traffic impacts before a single permit is issued. New Jersey has begun requiring quarterly public reporting of data center electricity and water usage—a baseline that every state should adopt.
Genuine community benefit agreements. Tax incentives for data centers have become so routine that they are rarely questioned. Virginia's data center sales tax exemption was projected to cost the state $1.6 billion annually before state senators voted to end it in March 2026. Communities that host these facilities should receive tangible, enforceable benefits: infrastructure improvements, broadband access, local hiring commitments, and revenue-sharing arrangements that reflect the scale of the burden imposed.
Grid impact assessments before approval, not after. A data center that requires its own substation or necessitates transmission upgrades should bear those costs—and those costs should be part of the public permitting record. The current practice of approving projects and then scrambling to find the power is backwards.
Environmental review that includes cumulative impacts. A single 50-megawatt data center may seem manageable. Five of them in the same county are a different matter entirely. Permitting processes must account for cumulative water, energy, noise, and land-use impacts—not evaluate each project in isolation.
The Opportunity We Keep Missing
The frustrating truth is that data centers, done right, could be engines of genuine community revitalization. A facility built on a brownfield site brings investment to land that was otherwise abandoned. A data center powered by landfill gas or paired with a solar installation turns waste into economic value. A campus designed with wildlife corridors and stormwater management can improve the ecological baseline of a degraded site rather than diminish it.
Iron Mountain has operated an underground data center in a former Pennsylvania limestone mine for years—a facility that uses the earth itself for cooling, reducing both energy and water consumption. These are not theoretical concepts. They are proven models that the industry has largely chosen to ignore in favor of speed and scale.
The question is not whether we build data centers. The question is whether we build them as if the communities around them matter.
At the EPR Foundation, we believe the answer to that question will define the next era of American infrastructure development. The communities pushing back are not anti-technology. They are pro-accountability. They are asking for what every citizen has a right to demand: that the industries operating in their backyards respect the water they drink, the air they breathe, and the land they call home.
We can build the digital future without sacrificing the places that make it worth building. But only if we start asking the right questions before the concrete is poured—not after.