From your first cluster to your own campus

Request a capacity plan
  • First cluster10–12 weeks
  • GPUs150K+
  • Powered landUp to 3 GW
  • First gigawattWithin 24 months
01The path

One cluster now. A gigawatt in two years

Start on dedicated racks in our halls, and grow into a campus of your own on our powered land, without changing partners.

~2 MW

Your first cluster, in production

The campus plinth from the same camera at its first stage: one data hall with its roof off and its racks inside at the back corner, and the outlines of the halls to come engraved on the bare aluminium.
  1. Dedicated AI clusters
  2. Build-to-suit
  3. Powered land
02Size your run

How big is your next run?

Pick a system and a GPU count. See the racks, the power and where it runs.

Start from a run

System

10,000GPUs

Racks
139
Power, IT load
19MW

Each square, 2 racks

Where it runs

  1. Dedicated AI clusters10–12 weeksSee the systems
  2. A campus of your ownFirst 250 MW within 12 months
  3. A gigawatt campusWithin 24 months

Estimates. GB300 NVL72 at 72 GPUs and about 135 kW a rack; B300, B200 and H200 at four 8-GPU servers a rack, about 60, 58 and 41 kW. The example runs are sizes for scale, not quotes. Your site’s design sets the real figures.

A close-up of a silver rack front: rows of compute nodes with their vents and ports, a fabric switch full of optical ports across the middle, management modules with status lights, and the rack rails.
03Inside the cluster

Built for runs that finish

Bare metal, our own optics and liquid cooling, watched live, so a long run gets to its last step.

04Beyond the cluster

Serve it, finance it, keep it yours

Request a capacity plan

Tell us about your next run

System
Workload
Region *

We use these details only to answer you. See our Privacy Policy.