- New research shows temperature increases of up to 9.1°C around AI data centres.
- Up to 340 million people globally affected by elevated local temperatures.
- Industry explores heat recovery and low impact cooling solutions.
A new study led by Andrea Marinoni at the University of Cambridge has identified a growing environmental challenge linked to the rapid expansion of artificial intelligence infrastructure. The research highlights a phenomenon termed the data heat island effect, where high density AI data centres significantly increase surrounding land temperatures.
Unlike conventional urban heat islands driven by infrastructure such as roads and buildings, this emerging effect is directly tied to the intense thermal output generated by advanced AI chips and their associated cooling systems.
The findings, published as a preprint on arXiv, indicate that land surface temperatures near data centre clusters have risen by an average of 2°C, with extreme cases reaching up to 9.1°C. The impact extends well beyond facility boundaries, with measurable warming observed up to 10 km from source sites.
Researchers estimate that approximately 340 million people worldwide are currently living within zones affected by these elevated temperatures. Notable case studies include regions such as Aragon in Spain and Bajío in Mexico, where unexplained temperature increases have been linked to large scale data centre deployments.
The surge in heat generation is largely attributed to the rising energy intensity of next generation AI hardware. Compared to traditional cloud infrastructure, these systems consume significantly more power and produce higher volumes of waste heat. This heat is typically expelled into the surrounding environment via air based cooling or liquid systems, contributing to the formation of localised microclimates.
In response, industry stakeholders are assessing mitigation strategies aimed at reducing thermal impact while improving energy efficiency. These include the integration of waste heat recovery systems that redirect excess thermal energy into district heating networks for residential and public use.
Alternative siting strategies are also under consideration, including the development of data centres in cooler climates, as well as underground and underwater installations designed to leverage natural temperature regulation.
At the same time, advances in liquid cooling and green data centre design are expected to play a critical role in managing heat output as global demand for AI infrastructure continues to rise.
Author: Bryan Groenendaal
