Rapid data center expansion is straining utilities and testing community support, as regions weigh the costs of infrastructure built for generative AI.
Data centers have moved from behind-the-scenes infrastructure to a visible source of concern. The facilities that store and process information for online services are expanding quickly to support generative artificial intelligence, and their scale is prompting questions about electricity, water and the effects on nearby communities.
The largest facilities, called hyperscale data centers, can contain hundreds of thousands of servers under one roof. Their electricity demand can reach hundreds of megawatts, comparable to that of a medium-sized city. That need can add pressure to electrical grids. Large facilities also require substantial water supplies, and residents nearby may contend with noise and light.
AI expansion brings local costs into view
For years, data centers attracted little public attention even as digital services depended on them. That changed as the number and size of facilities grew, particularly after ChatGPT launched at the end of 2022. The technology industry’s rapid construction push has made the infrastructure behind AI harder for communities to ignore.
One August poll cited in the source found that 75 percent of Americans opposed having such projects developed near where they lived. That finding suggests a gap between the industry’s demand for computing capacity and the willingness of some communities to host the facilities that provide it. Opposition has focused on immediate effects, including higher electricity costs and the experience of living close to a large site, as well as wider concerns about AI’s effects on work and relationships.
Those concerns have drawn support from people with different political views. In December, Bernie Sanders called for a nationwide pause on data center construction, arguing that the pace of change was too fast. He connected the growth of AI to potential job losses and weaker human connection, and said that a pause could give democratic processes more time to respond. His criticism also focused on the influence of wealthy technology executives, including Jeff Bezos, Elon Musk and Mark Zuckerberg.
Ron DeSantis, Florida’s governor, also criticized the industry that month. He proposed giving local governments more power to reject data center projects and developing standards to limit AI-related risks. He framed the issue around protecting workers, communities and society from job loss, encroaching construction and fabricated or low-quality AI content. The different arguments point to a broad dispute: people may disagree about AI’s promise while sharing concerns about who bears the costs of building the infrastructure.
Energy constraints shape development decisions
By 2026, rising electricity prices had become a widely discussed concern, with technology companies’ demand for power linked to their pursuit of generative AI. Hundreds of local construction pauses had been adopted, and New York introduced the first statewide pause on hyperscale projects in July. These actions reflect local efforts to slow development while communities and utilities assess the demands new facilities place on existing systems.
The tension isn’t limited to the United States. Singapore and the Dutch municipalities of Amsterdam and Haarlemmermeer paused new data center development in 2019. Ireland restricted projects in some areas by 2021 after energy demand began to outstrip supply. Some earlier pauses were later lifted when tighter rules were introduced, amid concern that restrictions could mean missing out on investment tied to AI. The latest expansion has renewed the debate about whether those safeguards can keep pace with demand.
Recent decisions in Denmark and Thailand show how directly power availability can affect plans. Denmark lowered data centers to the lowest priority for grid connections. Thailand paused further construction and suspended 49 projects that had already begun work. Such decisions highlight the tradeoff facing places seeking to attract AI development while managing limited electricity supplies and community concerns.
Investment goals meet community resistance
European countries have also sought to reduce reliance on U.S. technology companies. After Sam Altman presented the Stargate plans, European governments announced billions of euros in planned data center investment of their own. At the time, the focus was on keeping pace with the United States; the social and environmental costs received less attention. Those costs are now harder to set aside as communities confront the infrastructure needed to support the investment.
In the United States, pressure is also changing the debate in states with concentrations of data centers. Donald Trump has praised the industry and criticized communities that oppose projects. Yet politicians in areas with many facilities are facing complaints about power bills and accounts from people living nearby. Technology companies, in turn, are spending millions of dollars on political races in an effort to head off tighter limits.
The central issue is how quickly to build and who should have a say when a project would place new demands on local utilities and neighborhoods. AI services depend on computing infrastructure, but that dependence does not resolve questions about where facilities should go or how their costs should be handled. The coming midterm contests may influence the future of data center construction in the United States; the practical test for communities will be whether expansion can fit within available resources and earn local support.
This article has been adapted from Paris Marx’s book, Hyperscale: The Ambition and Excess of Big Tech’s Data Empires.
