A new study is raising fresh questions about the environmental cost of artificial intelligence.
Researchers say hyperscale data centers — the massive facilities powering AI systems — may be creating localized “heat islands” that raise surrounding land-surface temperatures and affect communities miles beyond the buildings themselves.
The paper, posted March 21 on arXiv and not yet peer-reviewed, analyzed temperature data collected by remote sensing platforms over roughly the last two decades and compared it with the locations of more than 6,000 data centers around the world. The researchers found that land-surface temperatures increased by an average of 3.6 degrees Fahrenheit, after an AI data center began operating. In the most extreme cases, nearby temperatures rose by as much as 16.4 degrees Fahrenheit.
The warming didn’t appear to stop at the property line. Researchers found temperature effects extending as far as 6.2 miles from some sites. They estimate more than 340 million people worldwide could be living within areas affected by that added heat.
The study’s authors describe this as a “data heat island effect,” driven by the heat released from energy-intensive computing and cooling systems. As demand for AI keeps climbing, they argue, this kind of local warming could become a bigger part of the conversation around sustainable tech infrastructure.
Researchers pointed to examples in places like Mexico’s Bajío region and Aragón, Spain, where areas with growing concentrations of data centers showed temperature increases not mirrored in nearby provinces. They focused on data centers outside dense urban cores in an effort to reduce interference from other heat sources such as heavy industry or home heating.
The study is careful not to claim that data centers are directly causing global warming on their own. However, it suggests these facilities may be worsening local heat in places already dealing with rising temperatures tied to climate change. The concern here is less about a new planetwide climate driver and more about another layer of heat stress landing on nearby communities.
The findings add to a growing list of concerns surrounding the rapid buildout of AI infrastructure, which has already drawn scrutiny over energy use, water consumption and strain on local power grids — and that’s not including the socioeconomic it’s creating, as well.
