Nvidia is pressing back against criticism of AI data centers by highlighting its Rubin generation reference design for a fully liquid-cooled facility, claiming it “eliminated massive amounts of power usage and pretty much all water usage.” The company says the
For years, the public pushback against AI data centers has circled the same two worries: energy and water. Now Nvidia is stepping into that argument with a specific counterclaim—one that leans hard on cooling design.
In a blog post about its Rubin generation reference design for a fully liquid-cooled data center. Nvidia says the system has “eliminated massive amounts of power usage and pretty much all water usage.” The headline problem is familiar. and so is the skepticism. While Nvidia emphasizes near-zero water use. it doesn’t directly address all of the concerns that critics raise—especially what happens during construction and the power generation requirements of facilities at this scale.
There’s also the question of tradeoffs. As Gizmodo points out. Nvidia’s post doesn’t mention the cost of building this style of data center compared with one that relies on less efficient air cooling. Nvidia’s message. instead. is framed around momentum: it claims “every cloud provider and data center operator building for [Rubin] is making the transition.”.
Part of Nvidia’s efficiency case is simple, if counterintuitive: run the machines hotter. The company says its system can operate AI servers at temperatures as high as 113 degrees Fahrenheit (45 degrees Celsius). The pitch is that higher-temperature operation changes how heat can be moved and rejected.
With Nvidia’s approach, “heat is captured directly at the chip and transported through liquid loops operating at much higher temperatures,” letting outdoor dry coolers reject heat efficiently for much of the year. Nvidia also says the setup offers more flexibility as ambient air temperature changes.
Amazon made a similar move in a separate push, highlighting higher heat tolerances as part of making its mostly air-cooled data centers more efficient.
Nvidia’s strongest numerical claim targets water directly. Its head of sustainability, Josh Parker, says the reference design reduces 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.”
Still, the argument that matters most to communities and operators doesn’t end at operation. Even if a design can reduce water dramatically once servers are running. the broader pressure around AI data centers includes the effort required to build them and the electricity those facilities demand. Nvidia’s blog post leans on what its cooling system can do; critics want the full picture: what the transition costs. and what the knock-on effects look like before the first server is ever switched on.
Nvidia Rubin AI data centers liquid cooling water usage sustainability data center energy Josh Parker dry coolers
Near zero water use is great and all but who’s paying for all that “liquid cooled” stuff? Sounds like they’re just selling a dream.
So they run it at 113 degrees?? lol that feels like frying computers. Also “eliminated massive power usage” sounds sus. Did they forget the power to make the liquids and run the pumps?
I don’t get how they can say “pretty much all water” if the thing needs liquid loops. Like where does the water go, magically into the air? They mention dry coolers but that still sounds like water stuff to me.
This is another Nvidia PR thing. They didn’t say construction costs, and that’s the part people care about when companies want to stick these monsters in our neighborhoods. And electricity is the real issue anyway, not gallons per megawatt. Also “every cloud provider” yeah ok, sure.