Big Tech Is Hungry: To fuel AI in 2022, Microsoft utilised over 22 billion litres of water, whereas Google used 7 billion
AI work has always required a lot of resources, particularly power and water. But as of 2022, Microsoft's water usage has increased by 34%, Google's by 22%, and Meta's by 3% due to the current AI boom.
According to a Financial Times report, big tech companies like Microsoft, Google, and Meta have been using a lot more water to cool down their data centres in recent years, which has raised concerns about the environmental effects of the rapid advancements in generative artificial intelligence (AI). Water use has increased for Microsoft, Google, and Meta, as a result of the reliance of millions of users on their internet services.
A study from the University of California, Riverside that was published in Nature states that, given the increasing scarcity of freshwater, the lengthening of droughts, and the ageing of water infrastructure, it is critical to identify and solve the hidden water footprint of AI models.
The worry grows as top tech firms compete to launch products that use generative AI, which is driven by large language models that can handle enormous volumes of data. Due to the high processing power requirements of these models, large server farms that use chilled water for cooling systems—some of which can be recycled and part of which evaporates during the process—have been established.
Tech companies have used water at far higher rates than other energy-generating entities, such as hydroelectric power and thermal power plants. Water is a vital resource.
Microsoft used roughly 22 million cubic metres of water in 2022—that is, 22 billion litres of water. That would fill a pool the size of the Olympic Games 8800 times over.
Conversely, Google used around 6-7 million cubic metres, or 6-7 billion litres, of water. Merely 2 billion litres were consumed by Meta.The usage increased even further in 2023. These corporations have established goals to replenish water resources; by 2030, they hope to restore wetland ecosystems and improve irrigation infrastructure to the point where more water is returned to systems like aquifers than is used.
Water withdrawal is expected to reach previously unheard-of heights due to the need for artificial intelligence (AI). By 2027, it is projected to be between 4.2 billion and 6.6 billion cubic metres, or over half of the UK’s yearly water usage.
West Des Moines, Iowa, locals expressed concerns in a recent lawsuit regarding a data centre cluster that was using a large amount of the district’s water supply. In order to highlight the possible water-intensive nature of AI models, Shaolei Ren, an associate professor at UC Riverside, compared the water consumption of well-known chatbots like ChatGPT to “drinking” a 500ml bottle per 10 to 50 encounters.
There have been calls for increased data disclosure and transparency from AI companies, calling for comprehensive analyses of water usage across various computing services. While some businesses, like OpenAI, have stated that they are committed to increasing productivity, other businesses, like Google, have remained silent on the matter.
Experts stress that thorough reporting on AI models’ effects on the environment is essential, especially in light of the world’s growing concerns about climate change and depleting water supplies. In light of the current climate crisis, USC Annenberg research professor Kate Crawford emphasises the significance of comprehending the full environmental implications of generative AI techniques. (Source : Firstpost.com)
