"Artificial Intelligence & AI & Machine Learning - 30212411048" by mikemacmarketing is licensed under CC BY 2.0.

The United Nations University has compiled a report estimating the impact that AI data centres have on the environment, globally.  

Dr. Miriam Aczel, the lead author of the report, says carbon-only metrics are no longer sufficient for estimating the impacts of AI data centres, as they do not capture the whole picture. Far from it, actually. 

She explains that this report is focused on three of the many footprints being left by data centres: water, energy, and carbon. 

She says it was difficult to collect information due to a widespread lack of transparency; however, it is estimated that in 2025, 4.5 trillion litres of water were used b data centres just to generate electricity. This is projected to rise to 9.3 trillion litres by 2030.  

“When we say the water footprint, we were only looking at the water that was used to generate electricity to run the data centres. So, we’re not looking at, for example, cooling water.” 

She adds that, globally, these centres consumed the energy that could have powered 6900 square kilometres of adjacent land in 2025. This is expected to rise to 14,500 square kilometres by 2030.  

“That’s twice the size of metropolitan Jakarta; about 10 times the size of Mexico City. These numbers are really quite staggering, and that’s just for the electricity usage to power the data centres.” 

When it comes to carbon emissions, Aczel says data centres are expected to 399 million tonnes of C02 equivalent by 2030, which would require approximately 6.7 billion trees grown over 10 years to offset.  

Aczel adds that ChatGPT is the fastest adopted technology in history, receiving an estimated 2.5 billion daily prompts.  She assures that she is in favour of AI and all its benefits, but unless either users or developers start putting some safeguards in place, it could end up taking a detrimental toll on the world’s resources. 

Next steps for her research include exploring further impacts of data centres, such as cooling-water usage, e-waste emissions, and the big-picture impacts of using that much water on local ecosystems and industries.