The compact AI factory
deploys fast.
A fully integrated AI data center for a single pod-style GPU cluster, prefabricated and tested as one block. It arrives complete, with minimal onsite work afterward.
Talk to sales
Power, cooling and compute,
with minimal footprint.
Switchboards, UPS and distribution, wired and tested in the factory. Takes 3ph 415V/480V at 60 Hz and carries 5 minutes of ride-through on the way to the racks.
13 racks in a single row on overhead busway and piping. 8 are direct liquid-cooled compute at up to 150 kW each; the rest carry network, storage and management.
CDUs for the liquid loop and CRAH units for the air side, both in one container and both fed from the same cooling source. One plant instead of two, which is less to build and less to run.
Chiller, dry cooler and buffer tank on one skid. The dry cooler carries the heat whenever ambient allows and the chiller only trims, and the tank holds temperature steady through load swings and short interruptions.
1,240 kW of IT load, in 13 racks in a single row. 8 are direct liquid-cooled compute at up to 150 kW each, on system-level N+1 or 2N.
Manufactured,
not built.
Assembled and tested in the factory while the site is prepared: 90% less field work, 60% lower cost.
Compact enough for a colo's first AI tenant, a new site, or a miner moving into AI. Delivered in 75 to 90 days.
Design, manufacture, deliver and commission on one contract. Ten vendors, no owner.
Built for
Frontier AI Engines
NVIDIA GB300 NVL72
NVIDIA GB200 NVL72
NVIDIA HGX B300
HGX Platform