Nvidia says its AI data center design runs hotter to use a lot less water

Nvidia claims its next-generation AI data center design can slash water consumption to near zero by switching to 100% liquid cooling and running servers at higher temperatures, according to a company blog post [1]. The Rubin generation reference design captures heat directly at the chip and transports it through liquid loops, allowing outdoor dry coolers to reject heat efficiently for much of the year [1]. Servers in the system operate as hot as 113 degrees Fahrenheit (45 degrees Celsius), a shift that Nvidia says eliminates massive amounts of power usage and nearly all water use [1]. Josh Parker, Nvidia’s head of sustainability, stated the 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]. The announcement comes as Nvidia dominates the market for AI data center hardware. The company controlled more than 80% of the GPU market for training and deploying AI models as of 2025 and provided chips for over 75% of the world’s TOP500 supercomputers [8]. Nvidia’s valuation surpassed $5 trillion in 2025, driven by soaring global demand for AI infrastructure [8]. The company said “every cloud provider and data center operator building for [Rubin] is making the transition” to the liquid-cooled design [1]. Liquid cooling has become a focus for hyperscale operators managing escalating power densities. Amazon recently touted higher heat tolerances as part of making its mostly air-cooled data centers more efficient [1]. The broader context includes rapid growth in cloud services: Amazon Web Services is the world’s largest cloud computing provider, and Amazon Luna, its cloud gaming platform, uses Nvidia Tesla T4 graphics cards to stream games to users [7][9]. Cloud gaming demands minimum bandwidth of 10-20 Mbps, compared to 100-200 kbps for a typical console game, placing additional strain on network and data center infrastructure [6]. Nvidia’s blog post did not address the cost of building fully liquid-cooled facilities compared to those using air cooling, nor did it tackle concerns around power generation requirements or construction impacts [1]. The AI boom, accelerated by products such as OpenAI’s ChatGPT, has driven rapid investment and public attention toward artificial intelligence, intensifying scrutiny of data center resource consumption [2].