Key facts
- Data centers use water for cooling, with processors reaching up to 176°F.
- Data center cooling systems consumed an estimated 66 billion liters of water in 2023, less than 1% of US total consumption.
- Projections indicate US data centers could consume 731 billion to 1,125 billion liters annually by 2030.
- Water is also consumed in generating electricity for data centers, particularly from fossil fuels.
- Strategic siting and renewable energy use could reduce AI's future water footprint by up to 86%.
- Liquid cooling is an increasingly common and water-efficient method for AI data centers.
Protests are emerging across the United States against the growing water consumption of AI data centers, with locals expressing concerns over water use, energy demands, and pollution. Data centers utilize water for cooling their high-temperature processors, a process that can lead to significant water evaporation. While claims about the exact water footprint of AI queries vary, and company data is often inconsistent, experts emphasize that overall data center water consumption is currently less than 1% of the nation's total.
However, this figure is projected to rise substantially with the rapid expansion of AI infrastructure. While water-rich regions may absorb this increase, drought-stricken areas like New Mexico and Arizona face heightened local water stress. Engineers and researchers are developing and implementing solutions to mitigate this impact. These include improving the efficiency of AI models and processors, as demonstrated by Google's estimate of minimal water use per query for its Gemini chatbot.
Beyond direct cooling, a significant portion of AI's water footprint comes from generating the electricity to power data centers, often through fossil fuel combustion that requires water for cooling turbines. A shift towards renewable energy sources like solar and wind, which require little to no water, is seen as a key strategy to reduce overall consumption. Furthermore, strategic siting of new data centers in areas with ample water and renewable energy resources, rather than in water-stressed regions, could drastically curb future water use.
Advances in cooling technologies are also contributing to water efficiency. While older data centers relied heavily on evaporative cooling, newer AI-focused facilities are increasingly adopting liquid cooling systems. These systems circulate water or coolant directly over the hardware, offering a more efficient and often closed-loop method that minimizes water loss. In some cases, water-based cooling is reserved for the hottest periods of the year, further optimizing usage.