AI Factory · IaaS

An AI factory in your own data centre.

On your hardware or ours, on your domain. We design it, bring it live and run it. For regulated, sovereign and air-gapped environments, where nothing leaves your walls.

Talk to us See Day 2 operations

01 — An AI factory, layer by layer

From one GPU to a factory.

Watch a factory assemble: one accelerator, then nodes, racks, a repeatable unit, the networks that join them, and the services you run on top.

  1. GPUEverything starts with one accelerator.
  2. NodeGPUs inside a node talk over a scale-up fabric, as if they were one device.
  3. RackNodes stack into racks, with power, cooling and a management switch.
  4. Scalable unitRacks form a scalable unit: the repeatable building block of the factory.
  5. Compute fabricEach GPU's network port lands on its own leaf switch: a rail-optimised compute fabric.
  6. SpineA spine layer joins every leaf, so any GPU reaches any other in two hops.
  7. PlanesTo grow further, the fabric is repeated in parallel planes, multiplying bandwidth and scale.
  8. The factoryStorage and management run on their own networks. On Day 2 we run it all, on your domain.

02 — Day 0, Day 1, Day 2

Designed, built and run by one team.

  1. Day 0

    Design

    Workload sizing, hardware (yours or ours), power, cooling and network, model selection, and the security and compliance requirements the factory has to meet.

  2. Day 1

    Build and go live

    Installation, the serving stack, the gateway and keys on your domain, and acceptance testing before anyone depends on it.

  3. Day 2

    Run it

    Operated with the MIRASTACK operations stack: hardware-to-business observability, explainable root cause analysis, human-approved remediation, release risk checks, and supply-chain and cryptography scans. Air-gapped first.

    See Day 2 operations

03 — What you get

Capacity you own the walls around.

  • On your premises

    The factory sits in your data centre. Requests, data and models stay inside it.

  • Your hardware or ours

    Bring your own GPUs, or have us supply them. Either way we design around the site.

  • Your domain

    The gateway and keys run on your domain, under your control.

  • Open models

    Open-weight models selected with you, served on your infrastructure.

  • Operated for you

    The team that built it runs it, with the same stack that watches the hardware.

  • Air-gapped first

    Fully operational with no internet connection.

04 — How it works

From first conversation to steady running.

  1. You set the requirementThe workload, the people or systems it serves, and where it must run.
  2. We design itOn Day 0 we produce the plan for hardware, site, models and controls.
  3. We bring it liveOn Day 1 we install, configure and test the factory against your acceptance criteria.
  4. We run itOn Day 2 we keep it healthy, observable and governed for you.

05 — Built for

Where cloud APIs are not an option.

  • Regulated financial institutions.
  • Defence and public sector.
  • Critical infrastructure operators.
  • Sovereign data centre operators.

06 — Commercials

Agreed per contract.

An AI factory is scoped to the site and the workload. Hardware arrangement, models, term, and what we operate for you are set per agreement.

Talk to us

Tell us what you need to run.

Every agreement is shaped to the workload: seats, sessions, capacity, term and location.

Talk to us hello@mirastacklabs.ai