NVIDIA AI Data Center Project

    HDmann supplied the secondary liquid cooling package for an AI training hall built around NVIDIA accelerator platforms. The scope covered rack level supply and return manifolds, EPDM hose assemblies with UQD quick couplings, stainless isolation and balancing valves, and the fittings needed to complete each circuit.

    Every manifold was fabricated to the rack layout, branch spacing and per-server flow rate agreed with the integrator, pressure tested at 1.5 MPa and delivered labelled branch by branch. Hose assemblies were 100% pressure tested at 25 bar before packing, and the valve packages were kitted per loop with a line list so the site team received one box per circuit.

    The result is a wetted loop that behaves the same at rack 1 and rack 40: balanced flow to each cold plate, dripless service swaps, and a documented pressure test trail from factory to commissioning.