AI's impact on the environment

The rapid rise of artificial intelligence comes with a growing environmental cost. As AI systems become more powerful, the infrastructure behind them is putting increasing strain on our limited resources.

Companies like Amazon, Google, and Microsoft are building enormous data centers to train AI models. These facilities run specialized hardware around the clock, generating extreme heat that requires constant cooling. That cooling, in turn, consumes vast amounts of water and energy.

Water use has surged alongside AI development. Google’s data centers used about 8.1 billion gallons of water in 2024, nearly double what they used in 2021. A single Google facility in Iowa consumed roughly 1.3 billion gallons in one year. Microsoft and Amazon show similar patterns, with Amazon’s data centers in water-stressed regions like Aragon, Spain licensed to withdraw hundreds of thousands of cubic meters of water annually. Much of this water is drawn from the same supplies communities rely on for drinking. Most of it never returns, as evaporative cooling releases around 80% into the atmosphere. Worryingly, many new data centers are being built in regions already facing water scarcity, and projections for places like Texas suggest water use could reach extreme levels within the decade.

Energy demand is rising just as fast. Data centers already consume about 1.5% of global electricity, and usage is growing rapidly. By 2030, their energy consumption is expected to more than double, driven largely by AI workloads. To meet this demand, fossil fuel plants slated for retirement are staying online, and some tech companies are securing exclusive access to nuclear power. Researchers estimate this expansion could raise U.S. electricity bills nationwide, with especially sharp increases in high-demand regions.

These pressures translate directly into higher emissions. In 2024, data centers produced about 105 million metric tons of carbon emissions in the U.S., roughly 2% of the national total and triple their 2018 level. Much of their power still comes from fossil fuels, giving them a carbon intensity well above the national average.

Tech companies often respond with promises to become “water positive” or more sustainable by 2030, but these pledges rely heavily on offsets that do little to solve local problems. Water scarcity is a local issue, and restoring water elsewhere does not help communities whose supplies are being diverted today.

Efficiency gains alone are unlikely to fix the problem. Data center capacity could triple by 2030, and water use is projected to rise sharply over the same period.

This is not an argument against AI itself. The technology holds real promise, especially in areas like medicine and scientific discovery. But advancing AI without fully accounting for its environmental foundation risks long-term damage. Meaningful transparency, regulation, and genuine sustainability are needed so innovation can move forward without exhausting the resources future generations will depend on.