CAMBRIDGE: Scientists have discovered that large AI data centers are creating “heat islands” and warming the land around them, sometimes by as much as 16 degrees Fahrenheit. The rise in temperature is affecting more than 340 million people globally.
The study, conducted by researchers including Andrea Marinoni, an associate professor at the University of Cambridge, analyzed temperature data over the last 20 years using remote sensors. They compared surface temperatures with the locations of more than 6,000 AI “hyperscalers,” which are massive data centers housing thousands of servers and covering up to a million square feet.
The research focused on data centers located away from dense urban areas to avoid interference from other heat sources like manufacturing or residential heating. The team also adjusted for seasonal effects, global warming trends, and other influences.
Findings showed that surface temperatures increased by an average of 3.6 degrees Fahrenheit after a data center began operations. In some extreme cases, temperatures rose by up to 16.4 degrees.
The impact extended beyond the immediate vicinity, affecting areas up to 6.2 miles away. Regions such as Mexico’s Bajio and Aragon in Spain, which have become hubs for AI data centers, saw temperature increases of 3.6 degrees over 20 years—changes not observed in neighboring areas without data centers.
Researchers warn that the expansion of AI data centers could have dramatic environmental, social, and economic effects. “The planned scale-up of data centers could significantly impact society,” Marinoni said, highlighting risks to people’s welfare and the planet.
Deborah Andrews, emeritus professor of design for sustainability at London South Bank University, said this study was the first to focus specifically on the heat produced by data centers. She added that the rapid expansion of AI is outpacing sustainable planning.
Marinoni hopes the research will spark discussions on reducing AI’s environmental impact. “There might still be time to explore alternative paths without compromising AI’s progress for humanity,” he said.
