How much water does AI really use? It’s a question that’s getting more attention as technology grows. AI, or artificial intelligence, needs huge amounts of water to keep its systems running. Globally, AI is projected to use between 4.2 billion and 6.6 billion gallons of water each year. That’s a lot of water, enough to fill thousands of swimming pools.
AI’s water consumption is staggering, projected at 4.2 to 6.6 billion gallons yearly, enough to fill countless swimming pools.
Big tech companies are behind much of this water use. Google’s data centers, for example, used about 5 billion gallons of fresh water in 2022. That’s 20% more than the year before. Microsoft’s water use jumped too, hitting nearly 1.7 billion gallons in 2022, a 34% increase from 2021.
These data centers are like giant computers that power AI. They need water to cool down the machines so they don’t overheat. A single large data center can use up to 550,000 gallons a day, while smaller ones average 18,000 gallons. This high demand often draws from already stressed water sources, especially in the western U.S. where water stress is critical.
This water use is causing problems in many places. In the U.S., over 5,300 data centers often pull water from areas already short on supply. Some regions are feeling the strain. In The Dalles, Oregon, Google’s data center takes a quarter of the city’s water. That’s raised concerns among locals. Additionally, protests in countries like Chile and Uruguay highlight growing tensions over water access due to data center demands.
In places like Chile and Uruguay, people have even protested against new data centers because of their high water needs. The future looks challenging too. As more businesses adopt AI, water demand is expected to grow. By 2026, 90% of companies might use AI, putting even more pressure on water resources. Moreover, the energy demands of AI exacerbate the issue, as data centers are projected to account for a significant portion of global energy demand by 2030.
Training AI models also takes a lot of water. For instance, training one model at the level of a human brain can use 126,000 liters a year. Even a simple string of AI prompts might use 16 ounces of water.
Climate change is making things worse. Rising temperatures mean data centers will need more water for cooling. Many worry this could lead to shortages for communities. With AI use growing fast, balancing tech needs with water supplies is becoming a big issue worldwide.
