NextGenergy — Make every watt countTalk to our team
Products

Rack manifolds and quick disconnects

The highest-count removable boundary in the system, and the one disturbed most often.

Every server adds two quick disconnects. Every server swap disturbs them again. A disconnected QD is not automatically a sealed one, and one particle on the poppet seat may be enough. We specify manifolds and QDs together, with the QD standard, the cleanliness class at first fill and the re-test procedure after any disturbance written into the operator's procedures, not left to the rack integrator.

Interoperability is not yet settled across the industry. CPC, a quick-disconnect maker, has called for a common qualification and interoperability framework across QD specifications, and Microsoft and Nexthop AI list quick disconnects among the mechanical designs that still lack standardisation. Until that exists, the QD standard and the mating combinations actually tested are a project decision, and we write them down.

Headers and branches are sized to a stated design velocity. nVent's published design guidance targets about 3 m/s with 6 m/s as the ceiling, to limit erosion; whatever the project chooses, the velocity and the transient check belong in the design basis.

Numbers on this page, with their conditions and sources (1)
3 m/sLiquid-cooling distribution pipework is sized to a target velocity of about 3 m/s, with 6 m/s as the upper limit.Condition: Design guidance for CDU-to-rack distribution piping, to limit erosion; a design target, not a measured operating value. Water-hammer and transient checks are separate and depend on pipe size and valve closure times. Source: OCP APAC Summit 2026, Cooling Environments: nVent, "From Rack to Facility: Designing Scalable CDU-Based Liquid Cooling System" (M. Tien, M. Archibald).