Recent research from the University of Cambridge reveals that AI datacenters are creating localized heat islands, raising temperatures more than expected. These heat islands form as server farms dissipate heat, a trend accelerated by the increasing global presence of these facilities in response to AI demands. The study observed temperature increases ranging from 0.3°C to 9.1°C around operating datacenters, with the average rise between 1.5°C and 2.4°C. The heat effect remains measurable up to 10 km away, although reduced past 7 km. This situation adds complexity to existing urban heat concerns, as server farm energy consumption rapidly grows, outpacing many sectors.

The paper highlights the environmental implications of datacenters and AI hyperscalers, suggesting they are unlikely to be insignificant. Most facilities rely on fossil fuels for power, suggesting a potential spike in emissions alongside AI expansion. In the United States, new power projects rose in tandem with datacenter demand, causing a resurgence of coal power.

With the possibility of impacting up to 343 million people globally, the data heat island effect could influence public health, welfare, and energy systems. The authors acknowledge that advances in technology, such as more energy-efficient systems, could mitigate such effects.

This early analysis has not yet undergone peer review, and more studies are needed to confirm its findings. The study’s lead author, Dr Andrea Marinoni, emphasizes caution in the design and development of new datacenters to prevent future complications. Experts note that the current findings primarily address surface temperature changes and urban heat effects rather than definitive evidence of significant neighborhood warming.