Nvidia has unveiled a new generation of AI data center hardware built around "100% liquid cooling," a design the company says could nearly eliminate the enormous volumes of water that today's facilities consume.
According to Moneycontrol, Nvidia says its Rubin-generation AI infrastructure can run on a fully liquid-cooled, closed-loop system. That approach lets data centers in favorable climates run hotter and skip the water-hungry cooling towers that conventional facilities rely on to shed heat.
The potential savings are large. According to MSN, the architecture can cut facility cooling water use from roughly 2.6 million gallons per megawatt per year under conventional cooling-tower systems to near zero.
The platform is arriving with industry partners already on board. Yahoo Finance reports that Super Micro Computer (SMCI) has been named a global system builder and launch partner for Nvidia's new Vera Rubin NVL4 AI platform, unveiling next-generation, liquid-cooled rack designs to support it.
The shift reflects a basic engineering trade-off. Air cooling and evaporative cooling towers are cheap and familiar, but they consume water and struggle as chips run hotter and denser. A closed-loop liquid system recirculates coolant rather than evaporating water away, so very little is lost — though it requires racks and servers built specifically to be plumbed for liquid.
Why it matters: AI's explosive growth has put data centers on a collision course with local water supplies and power grids, and a design that slashes water use to near zero could ease one of the industry's most contentious environmental pressure points.