The environmental impact of artificial intelligence is currently defined by a sharp divide in technical and public discourse regarding data-center water consumption. While some critics argue that the rapid expansion of AI infrastructure threatens local water security and risks depleting regional aquifers, technology executives and some industry analysts maintain that such concerns are frequently overstated or lack a technical basis in reality. The reality remains obscured by a lack of standardized, transparent data, leaving local communities to navigate complex infrastructure projects without a clear understanding of the specific resource requirements involved in their jurisdictions.
The debate has intensified as high-performance computing facilities—often referred to as hyperscale data centers—expand across the United States. In July 2026, New York Governor Kathy Hochul signed the nation’s first statewide data-center moratorium, citing the “massive amounts” of water that hyper-scale data centers could require. This move underscores the growing friction between the rapid deployment of AI-ready infrastructure and the capacity of local municipal utilities to manage increased demand, particularly in regions already facing climate-related water stress.
Data Center Cooling Methods and Resource Trade-offs
To understand the environmental footprint of these facilities, it is necessary to distinguish between the two primary cooling technologies currently in use. Data centers generally rely on either evaporative cooling towers or closed-loop air-cooled systems. Cooling towers function similarly to human perspiration, using the evaporation of water to dissipate heat generated by server racks. This process is highly efficient but results in the consumption of large volumes of water.
Conversely, many newer, high-density AI facilities are adopting closed-loop systems that function more like automotive radiators. These systems circulate a coolant that transfers heat to an air-chiller, which vents the heat into the atmosphere without the same level of water evaporation. While this method significantly reduces the direct water footprint of the building, it comes with a substantial increase in electricity usage. Research indicates that air-cooled systems can require 10 to 65 percent more electricity than their water-cooled counterparts to achieve the same thermal regulation, according to estimates cited by Shaolei Ren, an AI and sustainability researcher at UC Riverside.
This creates a complex environmental trade-off: reducing direct water consumption on-site often necessitates higher power consumption, which in turn leads to increased “indirect” water use. Indirect water consumption refers to the water required by power plants—including hydroelectric and thermal facilities—to generate the electricity that sustains the data center. Meta, for example, reported that its indirect water consumption in 2024 totaled approximately 19 billion gallons, a figure 23 times greater than its direct operational water use.
The Challenge of Localized Impact Assessment
The question of whether data centers “waste” water cannot be answered with a single national statistic. While data centers consumed slightly more than 17 billion gallons of water for cooling in 2023—a total that represents less than a tenth of a percent of all water used by U.S. farms that year—the impact is felt most acutely at the county and municipal levels. A facility with a daily demand of 8 million gallons may be negligible in a major metropolitan area, but it can exceed the entire peak capacity of a smaller town, creating a potential bottleneck for local residential and commercial growth.
In Lebanon, Indiana, a Meta campus currently under construction has faced scrutiny regarding its projected water demand. While some local officials have indicated that planned infrastructure upgrades mean there is “nothing to indicate” that they will run out of water, other regions have reported different experiences. In Newton County, Georgia, local officials have noted that existing water infrastructure is insufficient to support the influx of proposed data-center projects, necessitating a rapid expansion of water-recycling capabilities to meet future demand.
The lack of standardized reporting makes it difficult for residents and regulators to assess these impacts in advance. Data-center operators often utilize shell companies to negotiate land use and utility agreements under strict non-disclosure contracts, which often leaves community members uninformed until construction begins. This lack of transparency has contributed to a broader sense of distrust, positioning the environmental impact of AI as a central point of contention in the current political landscape.
Future Regulatory and Technological Outlook
The future of AI infrastructure development will likely depend on three evolving factors: technological efficiency, the shift toward renewable energy, and increased regulatory oversight. Researchers are exploring methods to allow server hardware to operate at higher temperatures, which would reduce the cooling load, and some operators are increasingly looking toward renewable energy sources that require minimal water for power generation, as opposed to traditional natural-gas combustion turbines.
For now, the debate remains caught between conflicting narratives. The lack of “ground-truth” data, as noted by Eric Masanet, a sustainability researcher at UC Santa Barbara, prevents a consensus on the actual environmental footprint of the sector. As individual states and municipalities continue to evaluate the necessity of moratoria and stricter zoning laws, the primary challenge remains the development of a framework that balances the technical requirements of the AI industry with the long-term water security of the communities that host them.
Related reading