Artificial intelligence is becoming an increasingly integral part of our daily lives. However, its growth comes at a significant cost: water.
As AI systems consume more and more water, it's essential to consider their impact on our planet's precious resources.
Here’s an example of what I’m talking about:
If sixteen million people in the United States were to send just one 100-word email per week using ChatGPT, a widely popular AI platform, it would require approximately one hundred fifteen million gallons of water annually.
That’s nearly equivalent to the total water consumption of Rhode Island in two days.
Researchers have also found that a single 100-word AI-generated email uses slightly more water than a standard 16.9-ounce bottle of soda. These seemingly small amounts can quickly add up, highlighting the need for mindful water consumption and water efficiency in the age of AI.
So, where is all this water going?
Most of this water is going for cooling AI servers. They need a lot more cooling – meaning a lot more water – than traditional servers.
This water consumption is further exacerbated by the substantial electricity demands of generative AI in regions like California, where a sizable portion of electricity is generated using water-intensive methods.
Google's data centers offer a stark example of AI's water footprint.
According to recent research by the University of California, these data centers withdrew nearly 6.5 billion gallons of water and consumed 5.3 billion gallons for cooling in 2022 alone, a 21% increase from the previous year. These are the latest stats available.
Microsoft's water consumption for AI purposes rose by an even more alarming 34% during the same period.
Today, some AI companies are drawn to Arizona because of the abundant sunshine for solar power generation. However, the need for water for cooling remains is a pressing concern. With Arizona already facing water scarcity due to climate change and population growth, the additional demand from AI technology could further strain its fragile water resources.
In response to these concerns, a ChatGPT spokesperson emphasized the company's ongoing efforts to improve efficiency. These initiatives include developing more water-efficient cooling systems, optimizing AI algorithms, and increasingly turning to renewable energy sources.
While these solutions hold promise, it's imperative to address the growing water challenges associated with AI's unchecked expansion. The potential threat to global water scarcity cannot be ignored. Responsible and efficient water management is essential to ensure that AI's benefits are realized without compromising the worlds growing needs for clean, healthy, potable water.
Klaus Reichardt is CEO and founder of Waterless Co, Inc., a pioneer in advancing water efficiency. Reichardt is a frequent author and presenter who discusses water conservation issues. He can be reached at klaus@waterless.com
