NVIDIA says its next-generation Rubin AI infrastructure can operate using liquid coolant at temperatures of up to 45°C, significantly reducing power and water consumption while enabling more sustainable, high-density data centre operations.
NVIDIA has introduced a major redesign of data centre cooling technology with its upcoming Rubin AI server platform, claiming the new infrastructure can operate using liquid coolant at temperatures as high as 45 degrees Celsius. The company says the innovation could dramatically reduce both electricity consumption and water usage in large-scale data centres as demand for AI computing continues to surge.
The announcement marks a significant departure from conventional cooling approaches that rely heavily on chilled air, fans and refrigeration systems. According to NVIDIA, Rubin is the first generation of its AI infrastructure to adopt a fully liquid-cooled architecture, eliminating fans entirely and cooling every processor and networking component through a closed-loop system.
The company said the design is based on its DSX AI factory reference architecture, which has been developed to support the growing power requirements of advanced AI workloads while improving operational efficiency.
Reducing energy and water demands
As AI models become larger and more computationally intensive, cooling has emerged as one of the most significant operational challenges for data centre operators. NVIDIA noted that cooling systems can account for a substantial share of a facility’s overall electricity consumption, making efficiency improvements increasingly important.
By allowing coolant to enter the system at 45°C and exit at approximately 55°C after absorbing heat, Rubin servers can maintain full performance without relying on energy-intensive chilling infrastructure. The cooling liquid, a blend of water and propylene glycol, circulates through specialised cold plates attached directly to processors and other critical components.
NVIDIA estimates that hyperscale facilities deploying the new technology could achieve significant reductions in cooling-related operating expenses. In suitable climates, operators may be able to eliminate most water consumption associated with traditional cooling towers, helping address growing environmental concerns linked to data centre expansion.
Company executives said the architecture is designed to minimise reliance on refrigeration equipment, although limited use of chillers may still be required during extreme weather conditions.
Designed for the AI Factory era
The shift to full liquid cooling has also prompted a redesign of server infrastructure. NVIDIA said Rubin incorporates a denser rack architecture, streamlined liquid connections and sealed system designs that improve space utilisation while supporting higher computing performance.
Industry experts believe the transition reflects a broader trend within the data centre sector. As processor power levels continue to rise, liquid cooling is increasingly viewed as a necessity rather than an option for next-generation AI deployments.
Beyond efficiency gains, NVIDIA said the technology could enable new opportunities for waste-heat recovery, allowing excess heat generated by AI facilities to be repurposed for nearby buildings and industrial applications.
With cloud providers and hyperscale operators already preparing infrastructure for the Rubin platform, NVIDIA is positioning liquid cooling as a cornerstone of future AI data centres, where sustainability, density and performance must advance together.
See What’s Next in Tech With the Fast Forward Newsletter
Tweets From @varindiamag
Nothing to see here - yet
When they Tweet, their Tweets will show up here.
