Amid growing concerns about water usage in AI data centers, industry leaders are arguing that liquid cooling systems do not necessarily lead to high water consumption. Tuan Hoang, Head of Cooling Technology and Product Development at Schneider Electric, emphasized at a recent press event that, "We do not need to consume water to operate data centres." This statement responds directly to investigative reports highlighting the environmental impact of cooling technologies in data centers.
Hoang clarified that while liquid cooling is essential for managing heat loads in high-capacity systems, it does not inherently require large amounts of water. "Zero water is needed to cool AI data centres. Liquid cooling is required but it’s for the load and radiators," he stated. This distinction shifts the narrative about water consumption in data centers from necessity to choice, depending on geographical and operational factors.
A key point raised during the Schneider Electric and TeraWulf Global Press Event was the potential for significant reductions in water use when moving from traditional air cooling to liquid cooling. Case studies conducted by Schneider Electric on facilities in Dallas and Paris illustrate this potential. In Dallas, annual water usage could decrease from 382,000 cubic meters under air cooling to 197,000 cubic meters with liquid cooling—a reduction of about 48%. Similarly, the Paris facility could see its water consumption drop from 108,000 cubic meters to 51,000 cubic meters, representing a 53% decrease.
The Efficiency of Liquid Cooling Systems
These findings highlight how liquid cooling systems can improve thermal efficiency while minimizing water usage. Hoang noted, "It’s a choice to how you reject the heat and the myth that all data centres with liquid cooling are using lots of water isn’t true. We focus on high-efficiency chilling." The Uniflair XCA line from Schneider Electric exemplifies this approach, utilizing air-cooled chillers designed for data center environments. These systems operate without water by effectively radiating heat, demonstrating how advanced technologies can address environmental concerns.
The design of these cooling systems features closed-loop engineering, relying on a sealed volume of high-quality cooling fluid that does not evaporate or discharge into the environment. This innovative method allows for sustained cooling capabilities—up to 2.4MW per unit—without tapping into local water supplies. As Hoang explained, "The fluid supply is built to endure the entire operational life of the data centre."
Liquid Cooling: A Necessity for the Future
Rich Whitmore, CEO of Motivair by Schneider Electric, reinforced the necessity of liquid cooling in modern AI infrastructures. He stated, "Liquid cooling has been around since the 1980s but now it isn’t an option, it’s mandatory." With the rise of advanced AI systems, effective cooling solutions are key for operational stability and performance. Whitmore added, "People don’t have a choice – if you want advanced AI systems going in, you have to cool them."
This shift not only redefines resource consumption in data centers but also underscores the importance of thermal efficiency in climate and resource management. As demand for AI capabilities increases, so does the need for sustainable practices that do not compromise performance. The implementation of closed-loop liquid cooling systems represents a strategic move towards a more eco-friendly approach to data management.
The conversation around water usage in AI data centers is evolving. With advancements in liquid cooling technology, data centers can effectively manage thermal loads while significantly reducing their water footprint. As the industry continues to change, the focus will likely remain on balancing efficiency with sustainability, making sure that the future of AI infrastructure aligns with environmental considerations.
