The product line
The C-Series.
Three containerized, liquid-cooled AI data centre bases. Factory-prefabricated, craned into place, and run by the Mash Compute platform. Specs below are reference configurations, specified per deployment.
Three bases, one architecture.
C10
Edge and cold-climate sites
C20
Flagship training and inference
C40
Large-scale training
| Base | Size | Power | GPUs | Built for |
|---|---|---|---|---|
| C10 | 10 ft | Edge-scale | Multi-GPU | Edge and cold-climate sites |
| C20 | 20 ft | High-density | GPU-dense | Flagship training and inference |
| C40 | 40 ft | Hyperscale | GPU-cluster | Large-scale training |
Shared across the line
Common characteristics.
Cooling
Cold-plate liquid cooling with CDU and dry cooler, plus cold-climate free cooling. Design PUE ≤ 1.15; measured figures available on request.
Density
40–60 kW per rack, purpose-built for high-density AI workloads.
Availability
Target 99.99%, with optional N, N+1, or 2N redundancy.
Network
Spine-Leaf fabric with isolated out-of-band management.
GPU-agnostic
NVIDIA, AMD, or others — decoupled so the base sidesteps export-control chokepoint risk.
Scale
Tile multiple bases horizontally to grow capacity without redesigning the site.
The Mash Compute platform.
Unified management, orchestration, and operations across every base: DCIM, BMS and EMS, multi-tenancy, alerting, billing, and 24/7 remote operations.
Built on
FastAPI / Go microservices · React + TypeScript · PostgreSQL + InfluxDB · Redis + MQTT
Three ways to buy
Own it, configure it, or subscribe.
Buy the base
Purchase C-Series modules outright. Own them and site them on your ground.
GPU-agnostic configuration
We integrate and warranty the GPUs of your choice — NVIDIA, AMD, or others.
Compute as a Service
No hardware spend. Subscribe to compute monthly or annually while we build and operate it.
Want the full spec sheet?
We'll send reference configurations and size a base to your deployment.