Nvidia's Rubin-generation reference design uses 100% liquid cooling and higher server operating temperatures to dramatically reduce water consumption in AI data centers.
Nvidia is promoting a new reference design for AI data centers built around its Rubin-generation hardware that the company says eliminates nearly all water consumption by combining full liquid cooling with higher server operating temperatures. [1]
Conventional data centers rely on cooling-tower-based systems that consume significant amounts of water. According to Nvidia’s head of sustainability, Josh Parker, the new design takes water use “from roughly 2.6 million gallons per megawatt per year for conventional cooling-tower-based systems to near zero — up to a 100 percent reduction.” [1]
A key element of the efficiency gain is running AI servers at temperatures as high as 113 degrees Fahrenheit (45 degrees Celsius). [1] In Nvidia’s approach, heat is captured directly at the chip level and moved through liquid loops operating at elevated temperatures, which allows outdoor dry coolers to reject heat efficiently for much of the year and provides greater flexibility relative to ambient air temperature. [1]
Nvidia claims the design has “eliminated massive amounts of power usage and pretty much all water usage,” and says every cloud provider and data center operator building for the Rubin generation is making the transition to this style of cooling. [1]
The announcement comes amid growing public scrutiny of AI data centers over their water and energy footprints. [1] Amazon has separately highlighted higher heat tolerances as part of efforts to improve efficiency in its mostly air-cooled facilities, according to a recent report cited by The Verge. [1]
Nvidia’s blog post does not address the construction-phase environmental impact of large data centers, nor the power generation requirements of the facilities themselves. [1] It also does not disclose the cost of building a fully liquid-cooled data center compared with a conventional air-cooled one, a gap noted by Gizmodo. [1]
Sources
This article was drafted with AI from the cited sources and checked against them before publication. Spot an error? Let us know.



