Image Courtesy: Getty Images

Nvidia has introduced a new liquid-cooling technology for AI data centers that operates at temperatures hotter than a typical hot tub, a design the company says could dramatically reduce both water consumption and electricity use while supporting its next-generation AI chips.

The system, announced for Nvidia’s upcoming Rubin architecture, circulates a coolant mixture consisting of 75% water and 25% propylene glycol at approximately 113 degrees Fahrenheit (45 degrees Celsius). Despite the unusually high temperature, Nvidia says the liquid is still cool enough to absorb heat from its powerful AI processors before exiting the system at around 131 degrees Fahrenheit (55 degrees Celsius). The company believes the approach could help address growing concerns about the environmental impact of AI infrastructure.

Cooling has become one of the largest operational challenges for modern data centers as AI workloads continue to drive power consumption higher. Traditional liquid-cooling systems often rely on chillers that can account for nearly 40% of a facility’s energy use. Many of these systems also lose water through evaporation, increasing their environmental footprint.

Nvidia’s solution takes a different approach. Because the coolant operates at a much higher temperature than conventional systems, facilities can often use outdoor dry coolers to release heat directly into the environment instead of relying on energy-intensive chillers. The company also says the design functions as a closed-loop system that is filled once and reused throughout the life of the facility, potentially reducing water consumption by up to 100%.

The technology is expected to be most effective in cooler climates where outdoor temperatures remain below the coolant’s operating temperature. In warmer regions, chillers may still be required during periods of extreme heat, though far less frequently than in conventional cooling systems.

While the announcement could help address concerns that have delayed dozens of proposed data center projects, experts note that cooling represents only part of the sustainability challenge. AI facilities still require enormous amounts of electricity, much of which is generated by fossil fuel power plants that consume significant water resources themselves.

Even so, Nvidia’s high-temperature cooling system marks a notable step toward improving the efficiency of AI infrastructure as demand for computing power continues to accelerate.