Immersion Cooling

Immersion cooling is a thermal management technique where IT equipment is fully submerged in a thermally conductive but electrically insulating dielectric fluid. The fluid absorbs heat directly from all components, eliminating the need for fans, heatsinks, and cold plates. Single-phase immersion maintains the fluid in liquid state; two-phase systems use the fluid's boiling point to absorb heat through phase transition. While more efficient than air or direct-to-chip cooling, immersion has seen slower adoption in GPU infrastructure than anticipated due to supply chain limitations, maintenance complexity, and operator preference for direct-to-chip solutions that integrate with existing data centre designs.

Immersion systems typically use engineered fluids from 3M (Novec series) or Shell (Immersion Fluid S5 X). The fluid must be compatible with all server components — a non-trivial requirement given the variety of materials in modern GPU servers. Operators report that immersion can handle power densities exceeding 100kW per rack, but practical deployment challenges include fluid management during maintenance, compatibility with server warranties, and the weight of fluid-filled tanks requiring reinforced flooring.

We evaluate cooling strategy as part of data centre technical assessments. Immersion cooling proposals require careful scrutiny of total cost of ownership, fluid procurement logistics, and operational readiness — claims of efficiency gains often omit fluid replacement costs and maintenance overhead.