Nvidia says its next-gen data center cooling could solve AI's water problem, but real-world validation is needed

Nvidia has claimed at London Climate Week that its next-generation AI infrastructure can be fully cooled using a recirculated liquid coolant running at 113°F, potentially eliminating the need for additional chilling equipment and addressing growing water-use concerns around data centers. The announcement comes as Google and Amazon defend their data center water practices amid local opposition to AI infrastructure. For AI builders planning new deployments, the promise of reduced water consumption is significant, but the real-world impact depends on independent validation across diverse facilities.

What happened

Nvidia introduced a cooling system that uses a recirculated liquid mixture of water and propylene glycol, similar to automotive antifreeze. The coolant can operate at 113 degrees Fahrenheit, which is warm enough to reduce the need for additional chilling equipment. Nvidia claims this method can largely address data center water concerns, which have intensified as AI infrastructure expands. Microsoft’s vice president of data center engineering, Steve Solomon, commented on the potential before learning of Nvidia’s announcement, stating that it "would be a big deal for everybody if we got all of the chips to do that".

Why AI builders should care

Data center cooling and water usage are persistent constraints for AI deployments. A credible method to reduce water consumption could influence design decisions for new AI infrastructure, especially at scale. Nvidia’s approach targets one of the key environmental criticisms of AI data centers, which could ease regulatory and community pushback for new facilities. For builders operating in water-stressed regions, any reduction in cooling-related water consumption is directly relevant to operational planning and sustainability commitments.

Practical implications

If validated across diverse deployments, Nvidia’s high-temperature coolant could lower water consumption and potentially reduce cooling-system capital and operating costs by eliminating or downsizing chiller plants. The use of a recirculated coolant mixture means less reliance on evaporative cooling, which is a major source of water loss in traditional data centers. However, the practical benefits depend on the specific cooling architecture, climate, and power sources of each facility. Microsoft’s Steve Solomon highlighted the potential impact, though he noted it would require broad chip adoption to realize.

Caveats

Nvidia’s claims are based on its own statements at London Climate Week and have not been independently verified. Axios reports that even if the technology dramatically reduces cooling-related water use, water concerns do not disappear entirely. Producing the electricity to run AI infrastructure can also require significant water, depending on the power source. Additionally, efficiency gains may reduce per-unit resource needs but could also accelerate overall AI infrastructure buildout, increasing the industry’s total environmental footprint. Builders should treat the announcement as a promising directional claim and await independent testing across real-world data center environments.