The AI thirst is bleeding us dry
Artificial Intelligence (AI) is resource-hungry. From OpenAI’s ChatGPT to Google Gemini, the data centers running these AI services consume heavy amounts of electricity. In fact, according to the International Energy Agency (IEA), data centers accounted for around 1.5 percent of global electricity consumption in 2024. Not to mention, consumption was reported to have grown by 12 percent per year over the last five years, a rate of use that will only continue to go higher.
It also requires heavy processing power. Rapidly expanding AI infrastructure has put a strain on memory manufacturers. Random Access Memory (RAM) that would normally go to laptops, PCs, phones, and other consumer electronics is instead going to AI data centers, leading to shortages and increased prices.
Apparently, it’s quite thirsty as well.
The AI thirst
In a blog post by Sam Altman last year, the OpenAI CEO disclosed that AI datacenters use roughly one-fifteenth of a teaspoon per ChatGPT query:
“As datacenter production gets automated, the cost of intelligence should eventually converge to near the cost of electricity… The average query uses about 0.34 watt-hours, about what an oven would use in a little over one second… It also uses about 0.000085 gallons of water; roughly one-fifteenth of a teaspoon.”
He also went on to explain that these costs are but a necessary part of an inevitable technological evolution. “There will be very hard parts like whole classes of jobs going away, but on the other hand, the world will be getting so much richer so quickly that we’ll be able to seriously entertain new policy ideas we never could before,” says Altman. “We probably won’t adopt a new social contract all at once, but when we look back in a few decades, the gradual changes will have amounted to something big.”
But these only refer to water consumption during the use of AI services. What about training and maintaining these?
According to the Environmental and Energy Study Institute (EESI), medium-sized data centers consume up to roughly 110 million gallons of water per year. Meanwhile, larger data centers consume up to five million gallons per day.
Even manufacturing the chips and servers necessary to establish these AI services consumes plenty of water. And we’re not talking just about any kind of water—but rather, specialized ultrapure water that won’t contaminate or damage sensitive electronic chips.
Miguel Yañez-Barnuevo of EESI explains that approximately 1,500 gallons of piped water are required to produce 1,000 gallons of ultrapure water. “A single chip installed in a data center has already consumed thousands of gallons of water by the time it reaches the site.”
Environmental impact? Present, but not the main issue
Millions (and even billions!) of gallons of water sound a lot, particularly when we’re talking about the use of freshwater, which only accounts for about three percent of the world’s total global supply. Not to mention, according to EESI, approximately 80 percent of the freshwater used by data centers evaporates, meaning that it cannot be reused. In short, it’s a finite resource.
However, when compared to other uses of freshwater, particularly when agriculture accounts for roughly 70 percent of freshwater withdrawals, and industrial use under 20 percent, according to the 2024 UN World Water Development Report, AI water consumption seems minuscule in comparison.
Should AI’s thirst be monitored? Yes. But when municipal water that would normally be going to homes is instead shared with data centers—similar to the ongoing memory shortage—perhaps we should be more concerned about how the AI boom could affect the cost of water for regular citizens.
After all, it wouldn’t be the first time the bill is passed off to the consumer.
