Direct liquid cooling is, on paper, one of the gentlest duties a stainless hose can be given. Water with glycol in it, a few bar, forty or fifty degrees, no vibration to speak of. What makes it demanding is the consequence of a leak: the fluid is running within centimetres of live electronics whose replacement cost dwarfs the entire cooling loop, in a facility that is not allowed to stop.
What this service does to a hose
Very little, mechanically. The failure modes that matter here are the ones that come from installation and from chemistry rather than from load.
Chlorides. Cooling water is not neutral. Make-up water, inhibitor packages and whatever the facility dosed it with last quarter all contribute, and chloride content plus warm temperature is the classic stress corrosion cracking recipe for austenitic stainless. It is slow, it is invisible until it leaks, and it is the single thing worth getting right at the specification stage. Our media compatibility page covers what the limits are.
Galvanic pairing. A cooling loop is rarely all one metal. Stainless bolted to aluminium cold plates through a conductive fluid is a battery, and the aluminium loses. This is a loop design question, not a hose question, but it is the hose that gets blamed.
Routing. Racks are dense and the connection has to be made in a space that was designed around the server, not around the pipework. A hose bent tighter than its radius to fit behind a rack is the commonest thing we see on this duty.
What we would fit
| Element | What we would supply | Why |
|---|---|---|
| Hose | 316L annular corrugated, small bore | Adequate for the pressure with large margin; 316L for chloride tolerance over 304L |
| Braid | 316L single braid | More than enough at these pressures, and it protects the hose in a crowded rack |
| Liner | Not required | Velocities on a cooling loop are far below the limit |
| Ends | To the loop’s quick-disconnect standard, or flanged at the CDU | Tell us the coupling family — this is the part that has to interoperate |
| Length | Sized so the installed radius is above the static minimum with the connector engaged | The connector body eats into the space you thought you had |
| Testing | Hydrostatic on every assembly, with a test certificate issued with it | Standard on every pressure assembly we despatch, on this duty and any other |
What we would not fit, and what happens
304L on a loop whose water chemistry is not controlled. It will be fine until it is not, and the failure will be a weep at the worst possible place.
A hose cut to the shortest length that reaches. It has to be long enough to sit above its bend radius in the installed position, with the connector on. Measure with the coupling fitted, not without.
An assembly with no isolation between dissimilar metals, where the loop mixes stainless and aluminium. Raise it at design stage.
What we need to size it
| We need | Because |
|---|---|
| Water chemistry — chloride content and inhibitor package | This is the one that decides the grade, and it is almost never sent |
| Operating temperature and pressure, and the relief setting | The relief setting is the real design pressure |
| Coupling family and size at each end | It has to mate with what is already in the rack |
| Installed geometry, or a sketch with the space you have | Bend radius in a rack is the practical constraint |
| Whether the connection is made and broken in service | That turns a static assembly into a handling one |
| Quantity and whether you want per-assembly certification | It changes the test and documentation route |