Please enable JavaScript.
Coggle requires JavaScript to display documents.
"Making AI Less 'Thirsty'", Globally, AI systems are…
"Making AI Less 'Thirsty'"
The authors of this article suggest a few
solutions to AI's massive water usage
Simply raising awareness of AI
systems' insane water usage
Users can determine to use AI under water-efficient systems & at water-efficient times
To use AI when the sun is done to avoid high temp times when AI systems have to use more water but this is being revised because it conflicts with the "follow the sun" concept that helps reduce carbon emission
There are a lot of pros to AI usage, but we cannot overlook the cons of it too
AI has helped with climate change
AI uses a lot of energy & water
U.S. data centers (including AI ones) could
end up using 280 billion liters
of water a year by 2028
In the U.S., AI systems are predicted to use well over 300 TWh (1 trillion watts/hour) of electricity by 2028
When people highlight problems, those problems can be solved quicker & more efficiently
AI's carbon footprint would probably
be a concern today too, but a lot of people
started to notice it so it was controlled
AI's water usage continues to increase and not a lot of people are paying attention to it until it may become dangerously high
Some solutions for AI excessive usage has been put into play like the "Water Positive by 2030" project that includes major companies trying to slow down their water usage
AI systems mainly use water
two different ways (WWF & WCC)
Water Consumption accounts for water that is completely lost after water withdrawal to evaporation & etc.
Water withdrawal is the amount of water that is directly taken from a water source like our freshwater, temporarily or permanently
There's three different reasons that AI uses excessive amounts of water
Electricity generation using
off-site water
Server manufacturing using
supply-chain water
Cooling using onsite water
Globally, AI systems are predicted to use up to 134 TWh (1 trillion watts/hour) of electricity by 2027
Reference: Li, P., Yang, J., Islam, M., & Ren, S. 2025. Making AI less 'thirsty'. Communications of the ACM, 68(7), 54–61.
https://doi.org/10.1145/3724499