Direct-to-chip cooling
Liquid cooling where metal plates with fluid channels sit right on top of the GPUs and processors.
Also: cold plate
In direct-to-chip cooling, a cold plate is fixed to each hot chip and liquid flows through it, carrying heat to a coolant distribution unit that passes it to the building's water loop. It removes most of the heat, with fans handling what is left. This is the method used in Nvidia's GB200 and GB300 NVL72 racks and most liquid-cooled AI servers. It is less disruptive than immersion, because servers still look and are serviced much like ordinary ones.