AI Companies Are Thirsty for Data Centers, but Americans Oppose Them Nearby

There’s a growing backlash against the construction of local data centers, which consume massive amounts of electricity for compute and water for cooling in order to power artificial intelligence. According to a newly released Gallup survey, 7 out of 10 (71%) Americans oppose new AI data center construction in their area, with nearly half (48%) saying they’re strongly opposed.

Data centers have drawn increased scrutiny and resistance, with concerns about water shortages, noise and air pollution, and the depletion of land and energy resources. Communities across the US have protested new construction, and calls for moratoriums and bans have been growing. 

This is the first poll that Gallup has conducted about data centers. The results highlight some of the reasons Americans oppose them, with environmental impacts, such as energy and water consumption, topping the list. Other reasons include quality-of-life concerns, the effects data centers might have on utility builds, pollution and negative views of AI itself. 

Another survey question asked respondents about their attitudes toward the construction of a nuclear power plant to generate electricity. Nuclear power plants are viewed more positively than data centers, with 53% opposing them. 

Those in favor of data centers in their area cited economic benefits such as job opportunities and tax revenue, and technology benefits, such as meeting the demand for AI tech.

The data center race

The data center race is driven by companies at the forefront of AI technology — OpenAI, Microsoft, Google, Meta, Anthropic and others — as they compete for control of the industry. For its part, Nvidia dominates the AI chip market, providing high-performance GPUs and CPUs that populate data centers.

The AI race has become so competitive that there is talk of building data centers in space (as SpaceX’s Elon Musk has proposed), at sea and even in people’s backyards. Anthropic and SpaceX recently signed a major data center deal that would advance data-in-space efforts, and Google is also rumored to be working on a partnership with the rocket-maker.

These deals are driven by the increasing demand for apps, streaming TV and other data, but primarily by the processing power required for AI, including popular large language models, or LLM chatbots, like OpenAI’s ChatGPT and Anthropic’s Claude. 

To accommodate that, data centers are expanding capacity or being newly built, in some cases at an enormous scale. One proposal in Utah that has drawn public outcry would be twice the size of Manhattan and would require more electrical power than the entire state uses. That $100 billion proposal, called the Stratos Project, is backed in part by Shark Tank’s Kevin O’Leary, who said it will create 10,000 jobs. Local residents have pushed back against O’Leary’s comments about the project, including his claim that protesters against the project were bused in. 

A recent Politico story reported that a data center in Georgia used 30 million gallons of water without initially paying for its usage.

Doubts about the economic benefits of data centers

Another recent poll by YouGov found that 71% of Americans believe the pace of AI development is too fast, with 64% expressing doubt that AI will create economic gains that benefit everyone.

A Brookings report released earlier this month studying the employment effects found that data centers can create new jobs, but the employment potential is vastly overestimated by local governments and AI companies. Unlike factories, data centers operate more like warehouses for computers running on expensive chips, and many jobs associated with data centers are temporary during the buildout phase.

Many AI critics say that AI infrastructure proposals should include more energy-efficient measures to mitigate the heavy environmental footprint, and that more research is needed on the long-term health and economic impacts on local communities.