Microsoft’s data centres in West Des Moines, Iowa, have come under scrutiny for their substantial water consumption. According to a report by the Associated Press, these centres used significant amounts of water last year to keep cool while training OpenAI’s ChatGPT-4, a language model backed by Microsoft.
What’s particularly concerning is that this water usage occurred during a prolonged three-year drought in the region, putting additional stress on an already strained water system. The severity of the drought was such that even recreational activities like canoeing in local rivers became impossible.
Critics argue that diverting water resources for AI training, especially in times of drought, is unjustifiable. Jake Grobe, an organizer for Iowa Citizens for Community Improvement, pointed out that ChatGPT is not a necessity for human life, and diverting water to support AI computations raises ethical questions.
The water demand for data centres has been a concern even before the AI boom, particularly in arid regions like Arizona. With the rapid growth of AI, this problem is likely to be exacerbated. Microsoft alone increased its global water consumption by 34 per cent in a year, totalling nearly 1.7 billion gallons, largely attributed to AI training. Google, too, recorded a 20 per cent increase, consuming 5.6 billion gallons of water for machine learning.
This means Google’s data centres consumed more than 3 times what Microsoft consumed.
Moreover, Microsoft’s water consumption peaks during the hottest days of the year, compounding the strain on local water resources. This trend is worrisome because many American data centres depend on already stressed water systems, potentially disrupting local watersheds.
Data Centres Affecting Water Quality
Iowa’s water system is under moderate stress, and climate change is expected to worsen the situation. This will lead to increased demand from agriculture and data centres. The data shows that Microsoft’s water usage in West Des Moines has been consistently high. Further, the company has more data centres in construction.
The result is not only increased water consumption but also poorer water quality, affecting residents’ access to clean water. Local authorities are facing the challenge of balancing the water needs of residents, environmental concerns, and economic growth. This is especially challenging with the continued expansion of AI-related projects.
As AI technology advances, the ethical and environmental implications of its resource consumption will remain topics of debate. Prioritizing access to water for communities over supercomputing needs is a sentiment expressed by some local officials.
In Africa, we have seen a push for green data centres that focus on sustainability. At the African Climate Summit, an investment guideline for green data centres was presented to the attendees. This came a week after the ground breaking for EcoCloud, a green data centre powered completely by Geothermal.