AI Data Center Expansion Drives Water Depletion and Climate Target Failures
Rapid expansion of AI data centers in water-scarce U.S. regions is consuming vastly more water than tech companies publicly report, with indirect water use from power generation exceeding direct consumption by multiples. Companies' climate commitments focus on direct emissions and water use, leaving the larger indirect environmental impact unaddressed as drought conditions worsen in affected areas.
What happened
Tech companies have expanded AI data center construction rapidly, with major new facilities planned for water-scarce regions. According to reporting citing research and company disclosures, Google's water consumption rose 34% to 10.9 billion gallons in 2025, nearly all for data center cooling.
A study by Alex de Vries of Vrije Universiteit Amsterdam found Google's indirect water consumption—needed for power generation to run the centers—is approximately three times its direct water use. Research from the Environmental and Energy Study Institute (EESI) indicates a large data center can use up to 5 million gallons daily, equivalent to a town with tens of thousands of residents. The Guardian reported that approximately two-thirds of planned new data centers in the United States are located in regions that experienced severe drought in the past year.
Multiple companies have made climate commitments. Google reported a 25% increase in carbon emissions while Microsoft has pledged 'zero water' data centers and to 'replenish more water than we consume.' Meta stated indirect water consumption as 19 billion gallons, over 20 times its direct use, but has no plan to address indirect water impacts. Meta set a 2030 net water positive goal through water restoration projects calculating only direct consumption.
Reported impact
The expansion occurs amid widespread drought affecting over 60% of mainland United States, with severe water and snow shortages in the southeast and west driving farmland desiccation. Research estimates a model like GPT-3 requires approximately 500 milliliters of water per 10 to 50 medium-length responses depending on deployment location.
Christopher Dalbom, water law expert at Tulane University, stated that approving additional data centers drawing from the same aquifer could cause 'chronic devastation' for farmers in affected regions, while companies compete for market dominance and water-stressed areas face surging demand.
Index analysis
Companies report direct water consumption but systematically undercount indirect water use from power generation. The reported gap between direct and indirect consumption—ranging from 3 times higher for Google to over 20 times for Meta—indicates incomplete environmental accounting and accountability.
Kenneth Gillingham, Yale environmental economics professor, noted companies historically do not report indirect water because data centers are often far from power plants and water scarcity was not salient in non-water-stressed regions. The clustering of new centers in drought zones exacerbates this misalignment.
Governance controls
The Data Center Coalition claims progress adopting closed-loop cooling systems to replace standard evaporative cooling, though such systems consume more electricity typically from fossil fuels requiring substantial water for generation. Microsoft publicly committed to community-first AI infrastructure and water replenishment; Google and Meta have announced 2030 sustainability targets calculating only direct water consumption.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.