These days it feels like you can’t escape hearing about Artificial Intelligence (AI) — especially generative AI, like ChatGPT and CoPilot — which generate original text or images when prompted by the user. However, there is one major repercussion of AI that I don’t hear talked about enough: the amount of freshwater it uses.
How can typing something on a computer waste water? Data centers.
MIT News defines a data center as “a temperature-controlled building that houses computing infrastructure, such as servers, data storage drives, and network equipment.” As that infrastructure is used to perform your requests, it gets hot, like how your laptop may after using it for a while. To keep data centers from overheating, chilled freshwater is often used. It’s called an evaporative cooling system: water absorbs the heat and evaporates.
One data center can use millions of gallons of water in a single day, says the BBC. A study in the World Economic Forum estimates that by 2027, AI-driven data centers could use up to 1.7 trillion gallons of freshwater, only returning some of it to the water table.
To put this into context, a medium-sized data center (15 megawatts) uses as much water as three average-sized hospitals, according to The Wall Street Journal.
Data centers are necessary for regular computing too, like searching or sending an email. But because of the rise of artificial intelligence, the demand for data centers has risen significantly, and so has their construction. Artificial intelligence is also especially energy-intensive.
That’s partially because training new AI models consumes huge amounts of water. Noman Bashir, computing and climate impact fellow at MIT Climate and Sustainability Consortium, says that “a generative AI training cluster might consume seven or eight times more energy than a typical computing workload.”
Shaolei Ren, associate professor of electrical and computer engineering at the University of California, estimated that it took hundreds of thousands of liters of freshwater for OpenAI to train GPT-3, the third generation Large Language Model that powers ChatGPT.
Some states are trying to regulate the high water usage. In Connecticut, Senate Bill 1292 would require AI data center owners or operators to submit quarterly reports to the Commissioner of Energy and Environmental Protection.
However, the nation is still building data centers quickly. Shortly after his inauguration, President Donald Trump announced Stargate, a plan to build 20 new data centers in the U.S. in collaboration with OpenAI — the creator of ChatGPT.
Living next to these data centers, like Georgia resident Beverly Morris, can be really difficult. Morris is afraid to drink the water at her house because she believes construction of the data center disrupted her private water well, causing her tap water to be filled with sediments. Meta, who owns the center, denies it.
Even with all this data, it is difficult to put an exact figure on how much energy and water AI uses, since there are currently no state or federal regulations that require tech companies to publish this information.
Using up water at such a high rate is especially concerning when 1.1 billion people in the world do not have access to viable water, according to Forbes. The water that is being used in these cooling centers is freshwater, which can be used for drinking or farming. Some of it may end up back in the water table, but some of it is lost completely to evaporation or becomes too dirty to use.
Companies like Microsoft, Meta and Google have made promises in the past to become water positive by 2030, meaning they will replenish the water they use up and then some. However, at the current rate data center construction is going, this does not seem viable.
Potential solutions for the high water usage include more effective cooling systems, harvesting rainwater and more efficient infrastructure. At the user level, we can be more mindful about how we’re using AI.
Ultimately, it would be unrealistic for me to say we should all stop using AI – it does have really beneficial applications across many different fields. And like Pandora’s Box, now that it has been released to the public, it would be almost impossible to take back.
But the next time you use it, think about how much water it costs and ask yourself whether you really need its help or if you’re just being too lazy to perform a basic search.
