AI data centres are heating the planet and millions may feel it

The rapid expansion of artificial intelligence is leaving a significant environmental footprint, with new research indicating that data centers are not only major electricity consumers but are also contributing to the warming of their surrounding areas. A recent study conducted by researchers from esteemed institutions, including the University of Cambridge, reveals that regions near AI data centers experience an average increase in land surface temperatures of approximately 2 degrees Celsius once operations commence. This phenomenon has been characterized by the researchers as the 'data heat island effect.' The effects mimic the well-documented urban heat island phenomenon, where densely built-up areas trap heat, but in this case, the rise in temperature is driven by computational activities. Utilizing two decades' worth of satellite-derived land surface temperature data, the researchers discovered a clear temperature gradient between data centers and their neighboring regions, indicating that the impact is substantial enough to create a localized 'data heat island effect.' The study noted a marked temperature increase that correlates directly with the initiation of data center operations, with average increases recorded at 2.07 degrees Celsius, and in some instances, temperatures soared by as much as 9.1 degrees Celsius. This positions AI infrastructure alongside traditional industrial heat producers. The warming impact is not confined to the immediate vicinity of data centers. The research suggests that elevated temperatures can extend up to 10 kilometers from data center sites, with increases of up to 1 degree Celsius detectable as far as 4.5 kilometers away. The researchers cautioned that the effects of data centers and the activities of AI hyperscalers on climate could be significant. They noted that energy-intensive AI tasks not only increase heat output but also contribute to higher emissions. Case studies conducted in regions such as Aragón in Spain, Bajío in Mexico, and northeastern Brazil demonstrated unusual temperature spikes corresponding with the clustering of data centers, reinforcing the connection between computational infrastructure and localized warming. According to the study, more than 340 million people worldwide may already be subject to these localized temperature increases, with many data centers located outside of densely populated urban areas. The authors emphasized that the data heat island effect is a significant phenomenon that should not be overlooked, as it could have detrimental effects on public welfare, healthcare, and energy systems. Looking ahead, data centers are anticipated to become some of the most energy-intensive sectors within the next decade, with computational energy consumption potentially surpassing that of the manufacturing sector. Moreover, the majority of AI infrastructure still heavily relies on fossil fuels, meaning that the rising demand for computing translates directly into increased emissions and heat output. However, quantifying the environmental impact of AI remains a challenge, with researchers pointing out that there are currently no reliable metrics available to assess the effects of existing AI applications.