Hardware role · AI and inference

Compute

The AI and inference role: runs your models and your everyday apps.

Close-up of a rack holding six servers with blue lights and blue cabling.

What it does

Close-up of rack servers with rows of drive bays and blue status lights.

Runs your AI in your building

Models run on your own cards, so questions and documents never leave the building.

A wave of tiny blue and violet points of light.

Sized for your models

We size the machine for the models you want to run, on graphics cards you own.

A short stack of servers with blue light along their front edges.

Hosts your everyday apps too

The same machine runs the services your business uses every day.

At a glance

Every fleet is different. We size each machine with you and help you source it from any vendor.

Compute: at a glance
RoleAI and inference, plus your everyday apps
Runs onA machine you own, or one we help you choose and source
NeedsGraphics cards sized for the models you want to run
Works withGateway, Sentinel, Workshop and May
Managed byOnyx, on the same screen as the rest of your fleet

What's included

What's included with Compute
SoftwareOnyx, set up in this role on your machine
HardwareYours. We help you choose and source it from any vendor, and you buy direct.
Managed serviceOptional add-on: we watch and look after the machine for you

Works with

Three server cubes joined by a glowing path, with small padlocks along it.

Gateway

One checked way in. Every change approved, every update built and scanned.

Six servers in a ring, each wired by a line of light into a glowing padlock at the center.

Sentinel

Watches every log and locks out bad actors across your fleet.

Hands fitting a memory module into an open machine on a workbench, with spare modules and a screwdriver nearby.

Workshop

A bench for building and testing, and a spare pair of hands for the fleet.

Questions

What can it run?

The AI models you choose, plus the services your business runs every day.

Do I need Compute to use May?

Yes. May's model runs on the graphics cards of a machine in the Compute role.

What hardware does it need?

That depends on the models you want to run. We size it with you and help you source it from any vendor.

A person at a desk at night studying a monitor of alerts and charts, with a lit server rack behind.

Add Compute to your fleet