Iceotope raises $26m as AI rack densities push past what air cooling can do

"Iceotope has closed a $26 million Series B round to expand its product line and patent portfolio as the AI hardware cycle pushes rack power densities past the point where air cooling can keep up. Two Seas Capital and Barclays Climate Ventures led the round, with existing backers Edinv, ABC Impact, Northern Gritstone, and British Patient Capital all participating again, the company said on Wednesday."

"Next-generation Nvidia accelerator platforms are driving per-rack power densities toward and beyond 1 MW, a level at which both conventional air cooling and direct-to-chip liquid loops fail to remove heat fast enough to keep silicon at sustainable operating temperatures. Iceotope's approach, which the company calls chassis-based precision liquid cooling, replaces air with a dielectric fluid circulated directly around server components inside a sealed chassis, eliminating the fans, hot aisles, and water-intensive cooling towers that traditional data-centre architecture has relied on for decades."

"Iceotope cited SemiAnalysis figures projecting that the installed base of liquid-cooled AI accelerators will grow from roughly 3 GW today to 40 GW within two years, a more-than-tenfold expansion driven by hyperscaler and colocation operators forced to adopt liquid cooling for AI workloads that conventional architectures cannot sustain. The thesis is becoming visible at the operator end of the stack as well: Nvidia-backed Firmus is building modular, fully liquid-cooled AI factories across Australia to host 36,000 GB300 Grace Blackwell chips, financed in part by a $10 billion Blackstone debt facility."

Iceotope closed a $26 million Series B round led by Barclays Climate Ventures and Two Seas Capital, with existing backers participating again. The funding will expand the product line and patent portfolio as AI hardware increases rack power densities beyond what air cooling can handle. Next-generation Nvidia accelerator platforms are pushing per-rack power toward and past 1 MW, where conventional air cooling and direct-to-chip liquid loops cannot remove heat fast enough to keep silicon within sustainable operating temperatures. Iceotope uses chassis-based precision liquid cooling, circulating a dielectric fluid around server components inside a sealed chassis. This design removes fans, hot aisles, and water-intensive cooling towers. Liquid-cooled AI accelerator capacity is projected to rise from about 3 GW to 40 GW in two years, driven by hyperscalers and colocation operators adopting liquid cooling for AI workloads.

Read at TNW | Investors-Funding