Markets
AI Data Centers Shift Water Burden from On-Site Cooling to Power Plants
AI data centers are increasingly adopting advanced cooling technologies that, while minimizing on-site water usage, inadvertently shift the water demand burden to the power plants supplying their electricity. Major players in the industry, including Nvidia, Microsoft, Amazon Web Services, and Meta, are utilizing closed-loop and immersion cooling systems to achieve this reduction.
Traditionally, data centers relied on evaporative cooling, which consumes significant amounts of water. However, the newer methods, such as closed-loop systems that recycle cooling fluids and immersion cooling that submerges hardware in liquid, can eliminate on-site evaporation. The downside is that these systems can increase electricity consumption by 6-35% during peak periods, leading to higher water usage at thermal power plants that cool their operations with water.
According to the Lawrence Berkeley National Laboratory, U.S. data centers consumed approximately 17.5 billion gallons of water directly in 2023, but the indirect consumption related to electricity generation was about 12 times higher. As AI technology advances, server racks are now exceeding 100 kW, which could significantly expand the total water footprint of data centers by 2030, with electricity-related water use expected to account for 50-72% of total consumption.
The industry measures water efficiency using a metric called water usage effectiveness (WUE). AWS reported a WUE of 0.12 liters per kWh, while Microsoft aims for 0.27 liters per kWh by fiscal 2025. However, these metrics focus solely on water used at the facility, not accounting for the water consumed at the power plants.
This shift in water consumption patterns is likely to spark regulatory discussions, particularly in water-stressed regions like Texas. Recent market pricing indicates a growing interest in a potential moratorium on new data centers in Texas by December 2027, with support rising from 16% to 20%. Stakeholders should monitor developments from the Texas Commission on Environmental Quality and the Electric Reliability Council of Texas, as well as any legislative actions from Texas Governor Greg Abbott that could signal regulatory changes.
FAQ
What cooling technologies are AI data centers using to reduce water consumption?
AI data centers are increasingly adopting advanced cooling technologies such as closed-loop systems and immersion cooling, which minimize on-site water usage by recycling cooling fluids and submerging hardware in liquid.
How does the shift to advanced cooling systems affect water usage at power plants?
While advanced cooling systems reduce water usage on-site, they can increase electricity consumption by 6-35% during peak periods, leading to higher water usage at thermal power plants that require water for cooling.
What is water usage effectiveness (WUE) and how is it measured?
Water usage effectiveness (WUE) is a metric used to measure water efficiency in data centers, expressed in liters of water used per kilowatt-hour (kWh) of electricity consumed. For example, AWS reported a WUE of 0.12 liters per kWh.
What are the potential regulatory implications of the water consumption patterns of data centers?
The shift in water consumption patterns may lead to regulatory discussions, especially in water-stressed regions like Texas, where there is growing interest in a potential moratorium on new data centers due to increased water demand.
How significant is the indirect water consumption related to electricity generation for data centers?
In 2023, U.S. data centers consumed approximately 17.5 billion gallons of water directly, but the indirect consumption related to electricity generation was about 12 times higher, highlighting the substantial impact of electricity-related water use.