Microsoft's water consumption rises to 22bn liters amid AI boom
The rise of generative artificial intelligence (AI) has led to a surge in water consumption. As reported by the Financial Times, tech giants like Microsoft have substantially increased their water usage for cooling data centers, sparking concerns about the environmental impact. A study from the University of California, published in Nature, emphasizes the need to address the hidden water footprint of AI models, especially with increasing freshwater scarcity, ongoing droughts, and outdated water infrastructure.
Water consumption by different firms
In 2022, Microsoft's water usage jumped by 34%, reaching 22 billion liters (enough to fill 8,800 Olympic-sized pools), while Google's consumption increased by 22% to around six to seven billion liters. Meta's water usage rose by a modest 3%, totaling two billion liters. These numbers continued to grow in 2023, as companies aim to replenish water resources by 2030 through projects like improving irrigation systems and restoring wetlands.
AI demand expected to drive water withdrawal to unprecedented levels
Water consumption is anticipated to soar in the coming years as top tech firms compete to roll out products leveraging generative AI, fueled by expansive language models capable of processing immense data volumes. These models demand significant computing resources, prompting the creation of large-scale server farms reliant on chilled water for cooling. While some of the water evaporates during this process, only a slight portion of it can be recycled.
Heavy water withdrawal expected by 2027
Experts estimate that AI's rising demand could cause water withdrawal to reach between 4.2 billion and 6.6 billion cubic meters by 2027, nearly half of the UK's yearly water consumption. Recently, residents of West Des Moines filed a lawsuit over concerns that a data center cluster was using a significant portion of their district's water supply. Shaolei Ren, an associate professor at UC Riverside, compared water consumption of chatbots like ChatGPT to "drinking" a 500ml bottle for every 10-50 interactions.
Calls for transparency and comprehensive reporting
There is a growing demand for AI companies to offer detailed information on water consumption across various computing services and improve transparency. While some firms like OpenAI have pledged to increase efficiencies, others like Google have remained silent on the issue. Kate Crawford, a research professor at USC Annenberg, stresses the importance of understanding the true environmental impact of generative AI tools in the face of a climate crisis and shrinking water resources.
