AI’s Thirst: Data Centers Face Growing Scrutiny Over Water Consumption
Table of Contents
As artificial intelligence rapidly expands, so does its hidden environmental cost: water. The massive data centers powering AI models require enormous amounts of electricity and, consequently, water – both to cool the servers directly and to generate the power thay consume.A new wave of concern is building as these facilities increasingly locate in regions already grappling with water scarcity, forcing challenging choices between technological advancement and essential resources.
The water footprint of AI is a two-pronged issue,demanding solutions both within data center walls and across the broader power grid.Direct water use stems from cooling systems,often relying on evaporation,while indirect use is tied to the electricity generation needed to run these energy-intensive operations. According to experts, this indirect consumption frequently accounts for 80% or more of the total water used.
Direct Water Demand: A Local Stressor
Approximately two-thirds of U.S. data centers constructed since 2022 are located in areas already experiencing high water stress, according to Bloomberg News. The situation is especially acute in locations like Newton County, Georgia, where proposed data centers have requested water volumes exceeding the county’s daily usage. “Officials there now face tough choices,” one analyst noted, “rejecting projects, mandating water-efficient cooling, investing in infrastructure, or risking water rationing for residents.”
The timing of water demand is also critical. Data center water usage spikes during hot whether, coinciding with peak residential and business needs. In Arizona,for example,a data center’s monthly water consumption during the summer can nearly double its average level.
Beyond direct cooling, the electricity powering data centers carries a significant, often overlooked, water cost. In many regions, electricity generation – whether for AI training or everyday use – relies on fossil fuel-based power plants that require substantial amounts of cooling water. The U.S. electric power sector withdraws approximately 11.6 gallons of water and consumes 1.2 gallons for every kilowatt-hour of electricity produced, making it one of the nation’s largest water users.
This water usage, while typically not potable and not drawn from municipal sources, can still strain rivers, aquifers, and ecosystems, particularly in water-scarce areas. A 2023 study estimated that generating a single text output of 150-300 words using GPT-3 consumed 16.9 milliliters of water in an average U.S. data center – 2.2 ml for onsite cooling and a staggering 14.7 ml for electricity generation. While model efficiencies are improving, indirect water use remains dominant.
Balancing innovation and Sustainability
Fortunately, solutions exist to minimize the water impact of data centers. Unlike electricity, cooling systems are a design choice. While evaporative cooling remains cost-effective,its burden on local supplies during heatwaves is undeniable. Data centers can mitigate this by building onsite water storage, installing thermal energy storage, or upgrading local water infrastructure.
Alternatives to evaporative cooling include air-based and liquid-immersion cooling, utilizing recycled water, and employing waste heat reuse to reduce cooling demand. “Zero-water” designs, which recycle cooling water in a closed-loop system, eliminate the need for external water sources. Though, these advanced designs frequently enough increase electricity demand, potentially exacerbating indirect water use. Notably, water-cooled data centers consume roughly 10% less energy than air-cooled facilities. Immersion cooling,where servers are submerged in a heat-conducting fluid,offers another promising avenue. [image of immersion cooling system]
In water-stressed regions, prioritizing low- to zero-water cooling systems alongside investments in renewable energy sources is crucial. Conversely, in wetter regions with carbon-intensive grids, reducing overall power consumption should be the priority, even if it means continuing to use evaporative cooling.
Ultimately, navigating the intertwined systems of water and electricity requires data center operators to make difficult trade-offs between global climate goals and local water needs. These choices are complex and, until renewable energy sources dominate electricity grids, may be unavoidable.
