Primakov/Shutterstock
Most Americans (75%, to be precise) consume only 2.5 cups (591ml) of water per day. This is well below the recommended amount, which means that three-quarters of the people you see on a daily basis are likely dehydrated. Artificial intelligence, on the other hand, doesn't suffer from that problem. We already know how much energy a single ChatGPT prompt consumes, but a new finding from the Association for Computing Machinery (ACM) shows that a "medium-sized" GPT-3 query alone requires 500ml of water, which is about the same as that 16-ounce bottle of water sitting on your desk. For the record, a medium-sized AI request is considered an input of around 800 words and a response of less than 300 words — just think about how many words and responses it takes to have AI write an email for you.
And that trend isn't slowing down. A 2025 report from McKinsey & Company shows that companies are expected to spend $5.2 trillion by 2030 to build more data centers in order to keep pace with the growing global AI demands. In 2023, a single Google-owned data center's onsite cooling consumption exceeded 6.07 billion gallons (23 billion liters) of fresh water, and more data centers means even more water consumption in the future. At that rate, the water used by AI data centers will eventually rival the annual water withdrawal requirements of some countries.
How much water does AI require?
Gorodenkoff/Shutterstock
Generative AI models require a lot of power, and while electricity requirements and associated greenhouse gas emissions are often easy to grasp, the "thirst" for water is less talked about. How much water exactly? The ACM research shows that training Microsoft's AI data centers in the U.S. consumed 185,000 gallons (700,000 liters) of onsite water and accounted for the total consumption of 1.4 million gallons (5.4 million liters) of water. To help you visualize that better, a million gallons is equivalent to over 25,000 bathtubs.
These massive data centers are filled with an equally massive number of servers — it's estimated that the largest of such facilities are big enough to hold 2.6 million servers! These servers need to be cooled, and local water supplies are tapped for that purpose. Even when a data center uses dry cooling, there is still significant amount of water consumption, and these systems are less effective when exposed to hot ambient air temperatures or high humidity.
Naturally, GPT-3 is old news at this point, but even if future models become more efficient, researchers warn that the water consumption may not go down, as the growing demand for AI will offset any improvements. Though many people despise AI (partly for its obvious environmental impact), companies are increasingly relying on it, meaning that despite AI eliminating white-collar jobs and causing computer prices to rise, it's unlikely that the trajectory of AI growth will slow down any time soon.
Is the problem getting worse?
Richard Newstead/Getty Images
Yes. Despite all of these signs and warnings, there have been no substantial shifts in how data centers approach the problem of water consumption. Case in point: two-thirds of new AI facilities are built in the driest parts of the U.S. More precisely, over 500 of them will find their home in areas that have had severe droughts. This includes Texas, which is likely to see data centers account for nearly 9% of the state's water consumption by 2040.
The researchers who wrote the noteworthy paper on AI's water expenditure have suggested multiple ways to make AI more sustainable. More transparency would help. Reporting and tracking water consumption could certainly be beneficial, both for cooling requirements and the often-overlooked costs of off-site electricity generation. A similar thing applies to finding more appropriate areas for training AI models (such as data centers that have better water efficiency) and training models during more water-efficient hours.
Can you change anything? Hardly. Skipping on writing that ChatGPT email could help you sleep better at night, but the impact of that choice will be microscopic at best. After all, over a billion people use AI daily, so even if you quit AI cold turkey, you're still essentially bringing a knife to a fight that involves drones and fighter jets. Think about that next time someone gives you a hard time for using a plastic straw.