Liquid cooling solutions gain momentum as AI workloads turn up the heat in data centers

Artificial intelligence is driving unprecedented request successful information centers, arsenic nan request for processing immense amounts of information continues to surge.

As tech giants title to grow their infrastructures to accommodate AI workloads, they are faced pinch nan increasing situation of really to sustainably and affordably powerfulness these operations - and this has moreover led companies for illustration Oracle and Microsoft to explore atomic power arsenic a imaginable solution.

Another captious rumor is managing nan power generated by powerful AI hardware. Liquid cooling has emerged arsenic a promising measurement to support optimal strategy capacity while handling rising power demands. In October 2024 alone, respective tech firms announced liquid cooled solutions, highlighting a clear manufacture displacement successful that direction.

Liquid-cooled SuperClusters

At its caller Lenovo Tech World event, nan institution showcased its next-gen Neptune liquid cooling solution for servers.

The sixth procreation of Neptune, which uses open-loop, nonstop warm-water cooling, is now being deployed crossed nan company’s partner ecosystem, enabling organizations to build and tally accelerated computing for generative AI while reducing information halfway powerfulness depletion by up to 40%, nan institution says.

At OCP Global Summit 2024, Giga Computing, a subsidiary of Gigabyte, presented a nonstop liquid cooling (DLC) server designed for Nvidia HGX H200 systems. In summation to nan DLC server, Giga besides revealed nan G593-SD1, which features a dedicated aerial cooling enclosure for nan Nvidia H200 Tensor Core GPU, aimed astatine those information centers not yet fresh to afloat clasp liquid cooling.

Dell's caller Integrated Rack 7000 (IR7000) is simply a scalable strategy designed specifically pinch liquid cooling successful mind. It's tin of managing early deployments of up to 480KW, while capturing astir 100% of nan generated heat.

“Today’s information centers can’t support up pinch nan demands of AI, requiring high-density compute and liquid cooling innovations pinch modular, elastic and businesslike designs,” said Arthur Lewis, president of Dell’s Infrastructure Solutions Group. “These caller systems present nan capacity needed for organizations to stay competitory successful nan fast-evolving AI landscape.”

Supermicro has besides revealed liquid-cooled SuperClusters designed for AI workloads, powered by nan Nvidia Blackwell platform. Supermicro’s liquid-cooling solutions, supported by nan Nvidia GB200 NVL72 level for exascale computing, person begun sampling to prime customers, pinch full-scale accumulation expected successful precocious Q4.

“We’re driving nan early of sustainable AI computing, and our liquid-cooled AI solutions are quickly being adopted by immoderate of nan astir eager AI infrastructure projects successful nan world pinch complete 2,000 liquid-cooled racks shipped since June 2024,” said Charles Liang, president and CEO of Supermicro.

The liquid-cooled SuperClusters characteristic precocious in-rack aliases in-row coolant distribution units (CDUs) and civilization acold plates for lodging 2 Nvidia GB200 Grace Blackwell Superchips successful a 1U shape factor.

It seems clear that liquid cooling is going to beryllium astatine nan bosom of information halfway operations arsenic workloads proceed to grow. This exertion will beryllium captious for managing nan power and power demands of nan adjacent procreation of AI computing, and I deliberation we're only conscionable starting to spot imaginable effect it will person connected efficiency, scalability, and sustainability successful nan years to come.

More from TechRadar Pro

• How information centers tin usage liquid supra crushed to enactment cool
• Evolving information halfway cooling for AI workloads
• The institution down your washing instrumentality now wants to cool information centers