Industry news. The move to fanless, fully liquid-cooled rack architectures is widening the scope of liquid cooling inside the rack.
From chips to the whole rack
Historically liquid cooling was applied to the highest-power components – GPUs and CPUs – while the rest of the rack relied on air. Fanless rack designs remove that air path altogether, so liquid cooling is being extended to network interface cards, power busbars and power boards, optical transceivers and other components that previously used airflow.
The effect is that the cooling content per rack rises, and every one of those additional cold plates becomes another point that needs flow, pressure and temperature monitoring.
Why it matters for instrumentation
More liquid-cooled components means more sensors and more control loops per rack. Differential pressure across manifolds, coolant temperature at supply and return, and flow verification per branch all become part of the facility monitoring scope rather than optional extras.

From a real project panel: ZB-IOM-633 I/O modules collecting rack and plant-room points before they are passed to the BMS.
Integration implications
These measurements are only useful if they reach the facility management platform. Protocol conversion between the cooling control layer and the building or DCIM system is therefore part of the design, not an afterthought.
Source: industry reporting on rack-scale AI cooling architectures, 2026.