How much of a problem is AI’s water use?

Water usage by artificial intelligence (AI) systems has become a topic of concern and protest across the United States, with locals in states such as Texas, New Mexico, and Arizona voicing their worries about the environmental impact of the rapid expansion of data centers. The primary concern stems from the fact that many data centers rely on water to absorb heat from air or surfaces, subsequently causing cooling as it evaporates, much like the human body's cooling mechanism through sweating.

The processors inside data centers can reach internal temperatures of up to 176° Fahrenheit, a condition analogous to an overtaxed laptop overheating. However, determining the exact extent of water consumption by data centers is a complex matter. While there is no doubt that data centers consume significant amounts of energy, the exact water usage remains elusive due to the lack of precise and consistent data from large tech companies.

Online claims regarding water consumption by AI systems vary widely, with some articles suggesting that a single chatbot query could consume up to 500 ml of water. Conversely, other commentators dismiss the issue of AI's water usage as nonexistent. These conflicting reports have only served to further complicate efforts to gauge the true impact of AI on water resources.

The discrepancy in available data is not helped by the often incomplete and inconsistent reports provided by the tech companies themselves. This lack of clarity has led to a situation where even industry insiders struggle to provide definitive answers regarding the water usage of AI systems. As the demand for AI continues to grow, the issue of water consumption by data centers is likely to remain a contentious topic, with protesters and environmentalists urging for more sustainable practices and greater transparency from the tech industry.