Does AI really consume a bottle of water per search? Recently, a major topic of discussion has been the environmental impact of AI as a reason to heavily criticize it. But how true is that- how much energy and water does it really use?
To understand the environmental impact of AI, first, it’s important to understand how AI works. For AI to function, it uses data centers just like phones do. CPU systems are generally used in a lot of technology such as phones. CPUs do cereal processing which is lots of tasks in a specific order, in contrast AI uses GPU systems. GPUs take larger tasks and divide them, executing them at the same time (parallel processing). GPU systems allow developers to develop at a faster rate and perform the needed functions.
James O’Donnell and Casey Crownhart, senior reporters at MIT Technology, conducted a study to determine exactly how much energy AI uses. The two colleagues paired with the University of Michigan to run GPU servers at different scales; there was no definitive answer. “It depends on the model, it depends on what you are asking,” said Casey Crownhart. A ChatGPT query uses, “on an average question, about 0.000085 gallons of water; roughly one fifteenth of a teaspoon (per search)” and around 0.34 watts of energy. However, this data does not include the large-scale environmental impact of training the AI models, and the rapid rate at which this is occurring. It’s about “making adjustments after the model has been deployed, it is not just the electricity you consume when you plug the computer in,” says MIT writer and professor Elsa A. Olivetti. Some high-emitting GPU services could use 1,000 to 2,500 kg of CO2 per server (most use more than 10,000 servers), although AI models are becoming more efficient. In comparison, “ a generative AI training cluster might consume seven or eight times more energy than a typical computing workload,” says MIT scientist Noman Bashir.
This was a concern of Golestan Radwan, Officer of the United Nations Environment Programme, who stated in 2024 “the long term effects of AI are questionable and its impact varies, but AI growth is inevitable.” While AI’s popularity and growth continues to rise, its negative environmental impact also continues to increase. It is clear that AI uses significant amounts of energy and water. As more research continues to be released it’s necessary to consider the global implications of AI usage.
