How exactly do AI and data centers use water? It all comes down to two-pronged consumption: direct and indirect water demand.
The nonprofit Environmental Law Institute shared in an October 2025 report that high-performance servers and data centers require large amounts of water, not only to generate electricity but also to manufacture the materials. More than a decade ago, in 2014, data centers used 21.2 billion liters of water for their operations. In 2023, that amount will increase to 66 billion liters.
Inside a data center, servers on site use electrical energy to store data and perform calculations. That effort is expended in the form of heat. To avoid overheating and mechanical failure, data centers previously used air conditioners and fans to maintain a cool environment. Currently, water evaporation is gaining popularity as a means of cooling large-scale operations and higher heat quantities.
“Although this evaporative cooling process uses less energy, significant water is lost due to evaporation from waste heat,” the Environmental Law Institute writes. “Essentially, optimizing for energy efficiency can actually make your water less efficient.”
So where does that water come from? It is typically obtained through purchase with a local water utility to draw from local surface water or groundwater. Tensions are rising across Texas, especially as the state battles drought conditions already exacerbated by ongoing development across large swaths of the state.
Meanwhile, focusing on indirect water use, US-based data centers account for approximately 800 billion liters of indirect water use, drawn through power generation and manufacturing of servers, chips, and other related necessities. According to an analysis by the Environmental Law Institute, making one microchip requires about 2.1 to 2.6 gallons of water to cool the machine and remove contaminants.
“While some data centers are located in areas with abundant water and easy access without competing with other users, others may be built in drought-prone areas with degraded infrastructure,” the report said. “Over the past three years, more than 160 new AI data centers have opened in water-poor locations across the U.S. The strain often peaks during the hot summer months and periods of high electricity demand, when cooling systems are running and local power companies are already stretched to their limits.”
While concerns about water access and control remain, some companies, including both Google and Microsoft, have announced commitments to return more water to the environment than is used in manufacturing and cooling processes. Most technology industry leaders also publish data on water usage in AI and data centers, but their specific practices vary based on both what is included in the reporting and how often it is reported.
