Skip to content
ELEMENT 31

Czar Grid

A Czar Grid is several Czar nodes joined into one inference pool for a whole facility. It is the same appliance multiplied, not a second product.

· 2 min read

IN SHORT

A Czar Grid is several Czar nodes joined into one inference pool that serves every robot in a whole facility. It is the same appliance multiplied rather than a second product: robots ask the pool, not a particular node, and each node must attest what it is running before the Grid accepts it. The Grid is also the unit a site subscribes to.

A Czar Grid is several Czar nodes joined into one inference pool for a whole facility. It is not a second product; it is the same appliance, multiplied, sharing a larger workload across one building.

Why it matters

A small floor can live on one Czar. A full distribution center, an assembly plant, or a port terminal has more robots, more cameras, and more concurrent decisions than one node should carry. The Grid is how the pool grows with the building rather than being replaced.

The operator still sees one pool. A robot does not pick a node; it asks the pool, and the pool answers within that robot's latency budget. Adding a node is capacity, not a new system to integrate.

The Grid is also the unit the site buys. The subscription covers the configured Grid, hardware, platform license, and sustainment in one line, and more robots on the pool do not move it.

How it works at Element 31

Each node must attest what it is running before it joins the Grid. The Grid will not accept a node whose software image or sealing does not match the record. That rule holds whether the node was installed on the first day or added later.

Substrate runs across the Grid as one serving system. It spreads inference across nodes, holds each robot's latency budget, and keeps memory, retrieval, adaptation, and evaluation on the pool. Adapted weights and task records stay inside the building.

Posture is set per site. One Grid can run air-gapped, with signed update bundles carried in by media. Another can connect only to the fleet software, the warehouse system, and the PLCs the site already secures. A mixed estate runs both postures under the same appliance.

Safety is unchanged by the size of the Grid. Stops, speed limits, and motion limits stay on the robot. If the Grid cannot answer in time, the robot falls back to its on-board model, and if it cannot proceed, it stops.

The Grid reaches a site by one of two routes: Facility Direct, where the building subscribes, or Czar Chassis, where a robot vendor or integrator ships it under their own name. The Czar Grid page describes how a Grid is sized to a floor.

Questions this raises

When does a site need a Grid rather than a single Czar?
A single Czar is a pool for a small floor. A full distribution center, assembly plant, or port terminal has more robots and more concurrent decisions than one node should carry, so it joins several Czars into a Grid.
Can part of a Grid be air-gapped and part connected?
Posture is set per site. One site can run its Grid air-gapped while another site in the same estate runs connected only to the systems it already secures. Both use the same appliance and the same serving system.
Does adding robots change what the site pays?
No. The subscription is for the configured Grid, with hardware, platform license, and sustainment in one line, fixed for the term. More robots on the pool do not move it.

KEEP READING

Related.