Compute Cabins
20ft and 40ft cabins housing dense GPU or CPU racks with integrated cooling distribution.
- Envelope
- 20ft / 40ft HC
- Supported load
- 0.8 – 3.0 MW per cabin
- Rack capacity
- 24 – 48 racks
- Cooling
- Air, cold plate or immersion
What sits inside
IT load and rack rows
The compute cabin is the payload module. Rack rows, cable management, liquid manifolds and local power distribution are installed and tested in the factory. Air-cooled and liquid-cooled variants use the same structural envelope, so a site can mix both without changing the pad layout.
Reference configuration
- Envelope
- 20ft / 40ft HC
- Supported load
- 0.8 – 3.0 MW per cabin
- Rack capacity
- 24 – 48 racks
- Cooling
- Air, cold plate or immersion
- PUE target
- Below 1.3
- Monitoring
- DCIM and leak detection
Final configuration subject to project design
Rack rows
Standard 19in racks with hot/cold aisle separation.
Power distribution
Busway or PDUs fed from the power zone.
Liquid manifold
Supply and return headers with isolation valves.
Fire and access
Detection, suppression and door access control.
Cable management
Overhead trays and underfloor routing options.
How it connects
Each cabin exposes a defined set of interfaces. The boundary is agreed before fabrication.
Where it is used
AI training clusters
High-density GPU rows with liquid cooling.
Edge compute
Single or paired cabins near the network edge.
Temporary capacity
Bridge capacity during a permanent build.
Technical documents
Catalogs, solution briefs and reference specifications for both product lines.
Download CatalogRelated systems
Specify this system
Send the load, redundancy level and site conditions. We will confirm the cabin configuration.