Artificial intelligence (AI) is increasingly shaping our everyday decisions and becoming an integral part of our daily lives. From virtual assistants to complex climate simulations, the AI is powering everything. While the technology continues to advance at a rapid pace, the expanding infrastructure required to support AI brings an often-overlooked consequence, that is, a growing demand for water.
So far, the focus has been on data centres’ high electricity use, but a new study shows they are also becoming major users of freshwater, at a time when water is already scarce. Several regions are facing water scarcity worldwide due to rising instances of droughts and lack of rainfall. Notably, India is one of those countries. We are relating the use of AI to possible scarcity of water, because the report suggests that data centres supporting AI systems like ChatGPT or Google's Gemini are increasingly concentrated in regions facing high water stress. According to experts, this could put local communities, ecosystems, and even the AI industry itself at risk.
---Advertisement---
How ChatGPT & other AI apps are making water scarce on Earth?
Interestingly, AI models run on large data centres filled with servers that operate 24/7. These servers generate a lot of heat and require constant cooling. This cooling mostly relies on freshwater. In the cooling process, much of this water is lost to evaporation or becomes wastewater that is hard to reuse. A recent study noted that about 45% of data centres worldwide are situated in river basins already facing high water stress. The report highlighted that data centres in the United States consumed about 66 billion litres of water in 2023 alone, with facilities accounting for the bulk of this usage.
---Advertisement---
Why is it concerning?
The matter is concerning as the issue involves not only direct water use, but also indirect consumption from electricity generation. This can constitute up to three-quarters of a facility’s total water footprint. The findings come amid a global surge in AI investment, which is paving the way for numerous AI chatbots and applications to enter the market. According to scientists, a 100-word AI prompt consumes roughly one bottle of water.
From Asia to the United States and Europe, large data centres are being constructed in regions already facing droughts, groundwater depletion, and pollution. This situation is concerning as currently there are no mandatory global regulations that require AI or data centre companies to disclose water usage, water-related risks, or local environmental impacts.