Four building blocks. Together they form a complete cooling chain from chip to plant room.
Coolant distribution unit
The heart of the system. Coverage from 200 kW to 3000 kW, with modular units at 200 / 400 / 800 / 2000 / 3000 kW steps. Controls flow, temperature and pressure, and feeds both cold plate and immersion loops.
· 200–3000 kW full range
· Modular, serviceable design
· For cold plate and immersion
Rack-mount GPU server
Eight GPUs in 6U, liquid-cooled, built for AI training and inference clusters where air cooling has already run out of headroom.
· 6U rack-mount, 8 GPUs
· Liquid-cooled
· For AI training and inference
Containerised compute
A self-contained enclosure carrying racks, cooling and distribution. Used when capacity is needed on site fast, or when the room is not ready yet.
· Compute and cooling in one enclosure
· Fast on-site deployment
· Customisable
Module-level data center
A repeatable liquid-cooled unit you can add row by row, instead of committing an entire hall up front.
· Add capacity incrementally
· Liquid-cooled from day one
· Repeatable build
Tell us your load and room, and we will match the configuration.
Contact KELVINS