Skip to Main Content
QUL logo

Artificial Intelligence

Artifical Intelligence and the research process

Environmental Impact

AI's Water Use & Carbon Footprint

Artificial Intelligence models consume an enormous amount of water and emit large amounts of carbon in their production, training, operation, and maintenance.

AI's water usage and carbon emissions are both direct and indirect:​

Direct usage includes water required to:​

  • Cool data centres ​

  • Produce technical components used for AI models ​

Indirect usage includes water required to, and emissions created by:​

  • Generate electricity that powers data centres​

Model of AI water use as cooling towers on and off site

Li, P., Yang, J., Islam, M.A., Ren, S. (2023). Making AI less “thirsty”: Uncovering and addressing the secret water footprint of ai models. ArXiv. Pg.4. http://arxiv.org/abs/2304.03271

It is difficult to keep track of how much of a carbon footprint AI has and how much water is used because of the lack of reporting and transparency from AI developers​

  • The water and energy needed continually grows along with the sophistication, size, efficiency, and use of the AI model

Related Readings

Li, P., Yang, J., Islam, M. A., & Ren, S. (2023). Making AI Less “Thirsty”: Uncovering and Addressing the Secret Water Footprint of AI Models. arXiv.Org. https://doi.org/10.48550/arxiv.2304.03271

Luccioni, A. S., Viguier, S., & Ligozat, A.-L. (2023). Estimating the Carbon Footprint of BLOOM, a 176B Parameter Language ModelJournal of Machine Learning Research24(253).

Yokoyama, A. M., Ferro, M., de Paula, F. B., Vieira, V. G., & Schulze, B. (2023). Investigating hardware and software aspects in the energy consumption of machine learning: A green AI-centric analysis. Concurrency and Computation35(24). https://doi.org/10.1002/cpe.7825

Potential Grey Literature/Popular sources: