Liquid cooling lessons from HPC enable AI factory scale

Motivair applies its exascale liquid-cooling expertise to enable AI factories at scale.

• Main announcement/action: Motivair is leveraging proven HPC/exascale liquid-cooling systems (used on Frontier, Aurora, El Capitan) to supply modular, repeatable cooling infrastructure for large-scale AI factories, including CDUs, ChilledDoors®, cold plates, and in-rack manifolds, scaling from 100 kW racks to multi-megawatt campuses.
• Background and technical details: Highlights key thermal variables — pressure drop, ΔT, and flow rate — with concrete figures: rack densities moved from 20–50 kW (air-cooled) to 300–400 kW (liquid-cooled); recommended flow is ~1–1.5 liters per minute per kW at under 3 PSI; Motivair’s engineered loops aim to stabilize ΔT and maintain GPU clock speeds across thousands of racks.