Submerged AI,
in one container.
FogPod Imm is an immersion-cooled AI pod. Servers run submerged, sealed from dust and humidity, and hold steady at full performance.
Talk to sales
Where AI compute
runs submerged.
400 kW of IT load in a standard 40 ft HC container. Backed by in-house immersion cooling, field-proven across sites.

8 tanks in two rows, 32 RU / 29 OU each. Servers run fully submerged in dielectric fluid, sealed from air, dust and humidity, and every tank carries two hot-swappable CDUs, so cooling is 2N at the tank.
The roof is the heat rejection. Coils and fans above the tank hall reject the fluid loop's heat straight to ambient, so nothing is installed beside the container.
A backpack air unit at the end of the hall keeps the aisle, the network cabinet and the electrical cabinet at temperature. They are the only parts of the pod that stay in air.
Electrical cabinet and network cabinet at the end of the row. 3ph 415V/480V at 60 Hz in, N+1 on UPS and PDU out to every tank.
Built for
PCIe GPU fleets
NVIDIA RTX PRO 6000
NVIDIA H100
NVIDIA H200
NVIDIA RTX 5090Smooth
performance.
Servers run sealed under fluid in a system Fog designs and builds end to end, from the tank to the container.
Efficient and sustainable
PUE 1.05 with no fans and no air handling. The heat leaves through dry coolers on the roof.
Easy to deploy
Factory-built and tested, designed and made by Fog. Power and network in, and the pod is running.
Smooth operation
Submerged servers see no dust, humidity or thermal cycling, so they fail less often.


