An AI data center moves in next door. Within months, the surrounding land is measurably warmer — and the effect stretches for miles.

A University of Cambridge study has found that data centers raise local land surface temperatures by an average of 2°C (3.6°F) after they begin operations, with extreme cases hitting increases of 9.1°C (16.4°F). The warming doesn’t stop at the property line. Significant temperature increases were detected up to 10 km (6.2 miles) away, according to the paper, which analyzed 20 years of NASA satellite data across more than 6,000 facilities worldwide.

The researchers call it the “data heat island effect” — a direct analog to the urban heat islands created by dense cities. Roughly 340 million people live close enough to a data center to be affected.

The study, which has not yet been peer-reviewed, maps temperature anomalies in regions including Aragón, Spain; the Bajío region of Mexico; and parts of Brazil — all areas where data center construction has intensified over the past two decades.

Not everyone is convinced. Vlad Galabov, senior research director at Omdia, told The Register that the study measures land surface temperature — how hot roofs and tarmac get in sunlight — rather than the air temperature people actually experience. “Most of the observed effect is likely driven by land-use and surface-cover changes, not by AI compute itself,” he said.

Ralph Hintemann, a senior researcher at the Borderstep Institute, separately told CNN the reported effects “seem very high” and argued that carbon emissions from power generation remain a more pressing concern.

Lead author Andrea Marinoni stands by the findings. “The message I would like to convey is to be careful about designing and developing datacenters,” he told New Scientist.

Power consumption for data processing has been estimated to exceed the amount required for global manufacturing within three to five years. That statistic is familiar. The idea that your neighborhood gets physically hotter to make it happen is newer — and harder to ignore.

Sources