The Tetra Tech High Performance Buildings Group’s latest white paper, “Taking the Heat Out of AI: Sustainable Solutions for Liquid-Cooled AI Data Centers,” helps data center clients sustainably power the rapid advance in AI computing.
The next generation of high-powered AI servers are becoming commonplace in data centers to accommodate increasing AI use across many industries. Pushing the bounds of computational power comes at a cost: increasing power consumption.
The liquid-cooled servers needed to power AI computing will need to work alongside air-cooled IT equipment, leading to a hybrid environment. Direct-to-chip and immersion cooling provide great opportunities for increased heat rejection efficiencies and better parameters for heat reuse. Liquid cooling of AI servers does not require a fundamental change to facility water systems, but the cooling systems will need to evolve to support both liquid- and air-cooled requirements that will exist in a hybrid environment.
Today, new liquid-cooled data centers are being designed and constructed globally to support the new 2.7 kilowatts (kW) NVIDIA Blackwell GB200 GPU and the new 97.2 kW NVL72 liquid-cooled servers. The increased power demand of AI should also drive the development of renewable and low-carbon technologies to meet decarbonization targets. Data center developments will need to focus on increased usage efficiencies and delivering effective heat reuse.
Our latest white paper, “Taking the Heat Out of AI: Sustainable Solutions for Liquid-Cooled AI Data Centers,” discusses challenges and sustainable solutions of designing for this new technology.
Learn more from our High Performance Buildings Group about how to sustainably support the next generation of AI technology.