AI embraces liquid cooling, but enterprise IT is slow to follow

The enthusiasm for generative AI is attracting serious investment, and the associated power and cooling requirements will pose a significant challenge for the data centers that house it. Upcoming AI training clusters will escalate silicon power and rack density to unprecedented heights, upending infrastructure design conventions and accelerating the adoption of cold plate and immersion cooling systems.

AI training will act as a torture test for cooling designs and refine the engineering of cooling products that follow. It may also drive convergence and standardization. Lessons learned in the AI space may lead the way for conventional business IT to embrace densified and liquid-cooled designs as well. However, these workloads come with markedly different resiliency, networking and thermal needs — while still underpinning many businesses’ primary revenue streams.