Coral Pod Platform

Coral AI Factory

A Coral AI Factory is an array of container pods sharing three services: a power bus, a liquid cooling loop and the AI fabric. Twelve C40 pods provide 9.6 MW of thermal capability and 2,016 rack units. The distinction that matters is not the total. It is that the total is reached in 800 kW increments.

Drawing

Twelve pods, one factory

Topology diagram of twelve CoralDC container pods interconnected by a shared power bus, liquid cooling loop and AI fabric
Increments

Why increment size is a commercial argument, not an engineering one

A conventional AI campus is a single decision. One site, one grid connection, one construction programme measured in quarters or years, one moment at which capacity either arrives or does not. The capital is committed before the first token is served.

A Coral factory is a series of decisions. Each pod is independently financeable. Each can frequently be sized to operate inside a building's existing electrical capacity rather than waiting on an interconnection queue. Each earns revenue before the next is ordered. If demand does not materialise, the exposure is one pod rather than one campus.

ScalePodsThermal capabilityRack units
Single deployment1800 kW168U
Cluster64.8 MW1,008U
AI factory129.6 MW2,016U

Shared Services

Shared services

Power bus

A common distribution spine so each pod connects rather than each pod requiring its own service.

Liquid loop

A shared secondary loop and heat rejection plant, sized across the array, which is materially more efficient than per-pod rejection at scale and creates a single point at which heat can be offered to an offtaker.

AI fabric

Ethernet or InfiniBand across the array so pods present as one cluster rather than as independent islands.

Next Step

Request an Architecture Review

A structured technical assessment covering electrical capacity, floor loading, delivery access, plant space and jurisdiction for a specific site.