A recent study, titled “The data heat island effect: quantifying the impact of AI data centers in a warming world,” by the Earth Observation group, University of Cambridge (UK), has rung alarm bells across the world.
The study undertaken by a research team of international scientists, led by Professor Andrea Marinoni, has unveiled a new environmental threat emanating from the artificial intelligence (AI) data centres, the “data heat island effect”.
What is the “Data Heat Island Effect”?
It is now sufficiently known that AI data centres are hugely resource-intensive, requiring copious amounts of energy and water for computing and cooling.
The United Nations University (UNU) estimates that the water needed for AI and data centres globally will nearly double to 9.3 trillion litres annually by 2030. This amount of water is enough to fulfil the yearly water needs of roughly 1.3 billion people. Likewise, UNU projects AI data centres’ electricity consumption to skyrocket to 945 Terawatt-hours (TWh) by 2030 on a yearly basis.
Amidst this, the Earth Observation group researchers have put forth a new ecological concern, AI data centres cause the surrounding land areas to heat up. The team examined at least 6,000 data centres worldwide and discovered that the land on which the AI data centre functioned, that land parcel witnessed its surface temperature increase by an average of 2°Celsius (3.6°Fahrenheit).
Hence, this warming effect of the AI data centres was termed by the researchers as the “data heat island effect”.
Some land areas, in extreme examples, also observed a gigantic rise of 9.1°C (16.4°F) in their surface temperatures due to the waste heat released by the all-round the clock operating servers of these data centres.
The team of researchers had deployed NASA’s MODIS Land Surface Temperature dataset to analyse and assess approximately 2 decades of satellite data, from 2004 to 2024. The researchers also isolated the data centres from urban environments, to strictly determine that the excess heat was being generated by these data centres, not the nearby industries, factories or city traffic.
The Land Heating Effect Spreads Far Beyond
The data heat island effect is not just confined to the area on which the AI servers are located, its impact was felt in regions situated almost 10 kilometres (6.2 miles) away. Therefore, the researchers forecast that upto 340 million people globally, living within the 10 kilometre radius of the data centre facility, could possibly be directly impaired by the data heat island effect.
Other Parts of the World Experiencing AI Heat Island
Additionally, the excessive warming of the land surface temperature by virtue of the AI data servers infrastructure was identified at Bajio area of Mexico as well as Aragon region of Spain.
Both the Bajio region and Aragon experienced an increase of about 2°C (3.6°F) in their land temperatures, especially after the expansion of the hyperscale data centres’ construction, the research study pointed out.
Another region that underwent the data heat island effect is Brazil’s Teresina. This municipality witnessed a spike of 2.8°C (5.0°F) in its land surface temperature around the AI data centre, which was an aberration from the local weather prevalent there.
Are the Research Findings entirely Accurate?
It is to be noted that the study has only been published as a research preprint, and has not been formally peer-reviewed. Moreover, other experts have pressed for more verification and are taking the study of the AI data heat island with cautious scepticism.
Lastly, the researchers have themselves admitted that further evidence and testing is necessary to confirm exactly how much of the local land warming is solely caused by the AI server infrastructure.