Cambridge Study: AI Data Centers Raise Local Temperatures by Up to 16.4°F

A new study from the University of Cambridge reveals that data centers powering artificial intelligence are creating localized heat islands, raising land surface temperatures around them by an average of 3.6 degrees Fahrenheit, with extreme cases reaching up to 16.4 degrees Fahrenheit. The research, based on satellite data from 2004 to 2024, examined more than 6,000 facilities worldwide and found the warming effect extends as far as 6.2 miles, potentially affecting over 340 million people living nearby.

Led by Andrea Marinoni, an associate professor in the university's Earth Observation group, the team analyzed NASA's land surface temperature dataset to isolate the impact of data centers. They focused on remote locations to minimize influences from urban activity or other industries, filtering out seasonal variations and broader climate trends. The study detected a sharp temperature rise shortly after operations began at these sites, often called hyperscalers due to their massive scale, which can span over a million square feet and house thousands of servers.

The phenomenon, dubbed the "data heat island" effect, stems from the intense energy demands of AI training and...