A Thirsty Future: The Hidden Water Cost of AI

As artificial intelligence weaves ever more deeply into the fabric of our daily lives—from chatbots answering our questions to algorithms guiding our online searches—the resources needed to power it often remain out of sight. We talk a lot about AI’s hunger for electricity, the race for renewable energy, and the carbon footprints of data centers. But behind these conversation-starters is a quieter, equally pressing issue: the enormous amount of water these massive computer operations require to keep themselves cool and running smoothly.

Every time we fire off a request to an AI model, a chain reaction of calculations takes place in data centers—rooms stuffed with thousands of computers working around the clock. These machines produce intense heat, and preventing them from overheating typically demands large quantities of water. Worse, the electricity they use is often generated in ways that also rely on significant amounts of water. In 2023, Microsoft’s data centers and other operations alone used enough water to fill more than 15 billion bottles. While that’s easy to say, it’s harder to find clear, straightforward reports from tech companies about their total water usage or how this compares from year to year. Many companies are simply not required to report these numbers, and some choose not to do so, leaving the public in the dark.

Yet, lack of information doesn’t change the facts on the ground. Imagine a region suffering from drought. Even if the water drawn for a data center is eventually released back into the environment, that short-term removal can worsen local shortages at critical times. The stress on water supplies is set to grow as AI becomes more central to business, entertainment, and even essential services. If we ignore this problem, we risk allowing the very technologies that promise to solve society’s biggest challenges to create new ones.

There are ways forward. First, the tech industry and governments can step up and establish clear, common rules for reporting water usage. We need honest numbers if we want to understand the scale of the problem. Second, data center designers can invest in better cooling systems—ones that rely less on water, use recycled water, or use methods like air cooling or more efficient heat exchangers. Third, it’s time to think more carefully about where we build these facilities. If sunny, water-scarce areas are best for solar power, maybe data centers shouldn’t cluster there unless they adopt advanced cooling methods. Meanwhile, regions with plentiful water might benefit from data center investments, provided companies can work alongside local communities to ensure sustainability.

The push for greener energy in tech is already underway, but it needs to be matched by an equally determined effort to cut back on water use. AI has the potential to help us address complex problems—from climate modeling to managing natural resources—yet it can also strain those same resources if we aren’t careful. As we enter an era defined by data-driven intelligence, let’s also commit ourselves to responsible stewardship of something far simpler and more precious: water.