AI Data Centers Use Far More Water Than Most Tech Giants Report
AI Data Centers Outpace Tech Giants in Water Consumption
AI Data Centers Use Far More - Major tech firms such as Microsoft, Google, and Amazon are allocating over $1 trillion to AI infrastructure in recent years, yet their water usage exceeds what is disclosed in sustainability reports. This discrepancy arises because the companies often overlook the indirect water consumption tied to energy generation, which varies significantly based on power sources.
Sustainability Reports Miss the Mark
While these corporations publish annual sustainability reports detailing on-site water use at data centers, only Meta includes water consumed at the power plants that supply their energy. This omission leaves a gap in understanding the full water footprint of AI operations.
"Indirect water consumption can vary a great deal, depending on how power is generated—coal and nuclear need a lot of water, while natural gas needs less." – Alex de Vries-Gao, VU Amsterdam researcher
A 2024 analysis by Lawrence Berkeley National Laboratory highlights that in the U.S., indirect water use for data centers is roughly 12 times greater than direct use. This trend could spark local disputes over water access as demand escalates.
Google’s Water Use Surpasses Expectations
Google’s 2025 sustainability report reveals a 34% surge in water consumption, totaling 10.9 billion gallons—primarily for cooling data centers. However, the report does not account for the water required to generate the electricity that powers these facilities.
Renewable energy sources like solar and wind use minimal water, whereas coal and nuclear plants consume significantly more. Despite Google’s claim of sourcing 100% renewable energy, not every electron comes from these low-water sources, leading to criticism about the true impact on water availability.
Regional Impacts and Reporting Challenges
Companies often fail to track indirect water use due to the distance between data centers and power plants. In regions with ample water supply, this indirect use becomes less critical, according to Kenneth Gillingham, a Yale environmental economics professor.
A new data center complex in Homer City, Pennsylvania, exemplifies this issue. Built on a decommissioned coal plant site, it now houses a natural-gas facility, yet its water demand remains comparable to the previous coal operation, per Jonathan Burgess of the Pittsburgh Water Collaboratory.
Water-Stressed Expansion
Recent findings from the Guardian and Bloomberg indicate that two-thirds of U.S. data centers are located in areas facing water shortages, such as Phoenix. This trend underscores the growing strain on regional water resources as AI infrastructure expands.
Meta’s 2024 indirect water use reached 19 billion gallons—20 times its direct consumption. While the company aims to achieve water positivity by 2030 through restoration projects, it has yet to address indirect usage. Microsoft, meanwhile, pledges zero water use for its data centers and a focus on community-centric AI, but this targets only direct consumption.
Amazon claims its data centers operate seven times more efficiently in water use than the industry average, with the company 75% toward its goal of replacing each gallon drawn. However, this calculation excludes indirect water demands, which could still affect local water availability.
Critics Highlight the Gap
State attorneys general and environmental advocates argue that renewable energy credits do not resolve water scarcity locally. For instance, a drought in Nevada’s rivers cannot be mitigated by Michigan’s water abundance. This disconnect raises concerns about the sustainability of AI growth in water-stressed regions.
A 2025 analysis by C further emphasizes the scale of this challenge, illustrating how current practices may not align with the long-term water needs of the U.S. as AI infrastructure expands.