One thermal path, six boundaries
Liquid cooling is bought as components and fails as a system. Our approach starts from that gap: we design and commission across the whole path from the facility water interface to the silicon, and we write down what each boundary was tested against.
The thermal path
From facility water interface to chip: six boundaries, who makes each one, and why the record has to travel with the heat.
Read →02Return-water grade
The first number that decides whether waste heat has a buyer. Why we design to it before we design to PUE.
Read →03Three clocks
Manufacturing, commissioning, service life. Three leak risks with three owners and three kinds of evidence.
Read →Component vendors optimise their box. We optimise the line.
A cold-plate maker will give you thermal resistance at a stated flow. A CDU maker will give you capacity at a stated approach. A facility designer will give you a supply temperature. Each number is honest inside its own boundary. The failures happen where the boundaries meet: the approach that eats your free-cooling hours, the quick disconnect that was never leak-tested after the last server swap, the 55 °C loop that was mixed down to 30 °C before it reached the fence and became worthless to the greenhouse next door.
We select cold plates, CDUs and manifolds against a written specification, integrate them into factory-built skids where that is possible, commission the whole path against written acceptance criteria, and stay through operation. The deliverable is a system that meets its acceptance criteria and a turnover package that shows it.