Data Centers Are Creating 'Heat Islands,' Warming Surrounding Areas by Up to 16 Degrees

The Study

Researchers led by Andrea Marinoni, an associate professor with the Earth Observation group at the University of Cambridge, published a preprint study on March 30, 2026, examining the thermal footprint of the world's largest data centers. The paper, titled "The data heat island effect: quantifying the impact of AI data centers in a warming world," is available on arXiv and ResearchGate. It has not yet undergone formal peer review.

The research team mapped the locations of approximately 8,400 hyperscale data centers — massive facilities offering cloud computing and AI services, many of which span up to a million square feet — against 20 years of satellite-measured land surface temperature data. To isolate the effect of the data centers from the well-documented urban heat island phenomenon, the researchers focused on facilities built in relatively remote or less-populated areas.

The pattern was consistent. After a data center began operations in an area, nearby land surface temperatures rose by an average of 3.6 degrees Fahrenheit. In the most extreme cases, temperatures surged by 16 degrees Fahrenheit. Crucially, the researchers noted that these measurements reflect the temperature of the ground surface itself — not ambient air temperature — as captured by satellite thermal imaging.

How Far the Heat Reaches

The temperature increases were not confined to a facility's immediate footprint. The study found that warming effects extended up to 6.2 miles from a data center, though they diminished with distance. Across all facilities studied, the researchers estimated that more than 340 million people live within the thermal impact zone of a hyperscale data center.

The researchers identified consistent regional patterns. In Mexico's Bajio region, which has become a major data center hub, temperatures near facilities climbed by approximately 3.6 degrees Fahrenheit over two decades, while surrounding areas that lacked data centers remained largely unchanged. A similar pattern appeared in Aragon, Spain, another rapidly growing hyperscaler corridor.

The Scale of Energy Behind the Heat

The physical mechanism behind the temperature spikes is not yet fully understood. Data centers consume enormous amounts of electricity for computation and cooling, and that energy is ultimately released as waste heat. Meta's recently constructed "Hyperion" data center, for example, cost $27 billion to build and has an expected computing capacity of five gigawatts — an appetite that, as Fortune reported on March 27, 2026, requires approximately ten gas-powered plants to feed.

However, some experts have questioned whether the heating measured in the study comes primarily from computational waste heat or from the physical properties of the buildings themselves. Chris Preist, a researcher at the University of Bristol, told New Scientist that it would be "worth doing follow-up research to understand to what extent it's the heat generated from computation versus the heat generated from the building itself," noting that sunlight absorbed by the large, dark structures could amplify the heating effect — a process already documented in urban settings.

Ralph Hintemann, a senior researcher at the Borderstep Institute for Innovation and Sustainability in Germany, described the figures as "interesting" but "very high," emphasizing the need for independent verification of the results.

Criticism and Pushback

The study has generated sharp debate even before peer review. Andy Masley, a writer who focuses on AI environmental claims, published a detailed critique calling the paper the "single worst writing and research on AI and the environment that I have read." Masley argued that the albedo effect — sunlight reflecting off and absorbing into building surfaces — could account for much of the measured temperature increase, independent of any computational heat output. He contended this would make the findings less about AI's unique energy footprint and more about the general effects of large-scale construction in previously undeveloped areas.

Marinoni, the lead author, told CNN that the rapid expansion of data centers "could have dramatic impacts on society" in terms of the environment, people's welfare, and the economy. Deborah Andrews, a sustainability researcher not involved in the study, told CNN that "the 'rush for AI-gold' appears to be overriding good practice and systemic thinking and is developing far more rapidly than any broader, more sustainable systems."

Context: The AI Infrastructure Boom

The study arrives as AI data center construction is accelerating worldwide. The construction of hyperscale facilities has surged over the past decade, driven by cloud computing and, more recently, by the explosive demand for AI training and inference. These facilities are already under scrutiny for their enormous water consumption — used in cooling systems — and their demand on local electrical grids.

The International Energy Agency has projected that global data center electricity consumption could double by 2030, with AI workloads accounting for a growing share. In the United States, utilities in Virginia's Loudoun County — home to the densest cluster of data centers in the world — have reported growing complaints from residents about noise, water use, and quality-of-life impacts from nearby facilities.

The Cambridge study, if its findings hold up under peer review, adds a new dimension to these concerns: the possibility that the physical heat output of data centers is measurably warming the areas where hundreds of millions of people live and work.

What Happens Next

As a preprint, the study's findings carry the caveat that they have not yet been validated by independent reviewers. The research methodology — comparing satellite temperature data against known data center locations over a 20-year baseline — is well-established in remote sensing, but the attribution of cause (computational heat versus structural albedo effects) remains an open question that the authors acknowledged.

Marinoni suggested that "there still might be time to consider the possibility of a different path … without affecting the demand of AI and its ability to provide progress for mankind." What form that path might take — whether better building design, mandatory setback zones, waste-heat capture systems, or stricter siting regulations — is a policy question that the study does not attempt to answer.