AI's Hidden Heat Problem: New Research Warns Data Centers Are Baking the Land Around Them — and a Massive Utah Project Is a Test Case

You probably don't think much about where your AI chatbot actually lives. But it lives in a data center — a massive warehouse packed with servers that run hot, consume enormous amounts of electricity, and need constant cooling. And according to new research out of the University of Cambridge, those facilities are doing something that's flown largely under the radar: they're cooking the land around them.

The study, published in late March, looked at 20 years of satellite temperature data mapped against the locations of so-called "hyperscale" data centers — the giant facilities that power AI at scale. What researchers found was striking. On average, surface temperatures around these facilities rose by about 3.6 degrees Fahrenheit after a data center opened. In extreme cases, the jump was as high as 16 degrees. In total, the researchers estimate that more than 340 million people are living in areas made hotter by nearby data centers.

Andrea Marinoni, an associate professor at Cambridge's Earth Observation group and one of the study's authors, was direct about what's at stake: the planned expansion of data centers "could have dramatic impacts on society" across the environment, public health, and the economy. Outside experts called it an important first look at a badly under-researched issue — though some said the numbers seem high and that the study, which hasn't yet been peer-reviewed, needs more scrutiny.

That debate now has a very real-world test case: rural Utah.

In Box Elder County, just north of the shrinking Great Salt Lake, a proposed "hyperscale" data center called the Stratos Project would be, by some accounts, the largest of its kind on Earth. The project is backed by Kevin O'Leary — yes, "Mr. Wonderful" from Shark Tank — and would span more than 40,000 acres, roughly two and a half times the size of Manhattan. At full power, it would require 9 gigawatts of electricity, which is more than the entire state of Utah currently generates or uses.

The scale of the heat problem alone is alarming to local scientists. Rob Davies, a physics professor at Utah State University, calculated that the project's total thermal output — from both power generation and the data center itself — could reach roughly 16 gigawatts of waste heat. He compared the daily heat release to the equivalent of 23 atomic bombs.

And then there's the water. Data centers need enormous amounts of water for cooling. Estimates suggest the Stratos facility could consume around 2 billion gallons per year just to run its gas-powered generators — enough to fill about 3,000 Olympic swimming pools. All of this is happening next to a Great Salt Lake that is already at or near record-low water levels after an unusually dry winter.

The county commission approved the project anyway in early May, to the fury of hundreds of residents who packed the public meeting. The developers have claimed the facility will use "zero water turbine" technology to minimize water consumption — but environmental advocates point out there's been no independent technical review or named manufacturer behind that claim.

O'Leary, for his part, pushed back on critics, saying the project "is not going to destroy air quality" and "not going to drain the Great Salt Lake." But with a project this large, many scientists and residents aren't willing to take that on faith.

The broader story here isn't just about Utah. It's about what happens when the AI boom meets a planet that's already running hot. The Cambridge researchers say there's still time to chart a different course — one that doesn't require choosing between technological progress and a livable climate. But that window, they warn, isn't open forever.