Turnkey AI factory infrastructure,
designed, built and run.
An AI factory is a GPU cluster built for one organization's training and inference, with the site, power, cooling and network around it. We take it from an empty site or a requirement to a cluster running to an SLA, and you choose who owns it.
- Stages
- Design, source, build, operate
- Platforms
- HGX B300, GB300 NVL72 and newer
- Fabric
- RoCE v2 Ethernet
- Owner
- You, your investors or us
One team from the site to the running cluster.
An AI factory usually means a site developer, a hardware reseller, an integrator and an operator, each handing off to the next. We carry all four stages, so the people who design the cluster are the people who answer for it once it runs.
- Site and power firstWe size the cluster against the site: a Tier III colocation hall, your own data center, a retrofit or a modular build.
- Hardware against the designSystems, switches, optics and spares are sourced through our relationships with NVIDIA and server manufacturers.
- Run on our softwareOnce it's accepted, the factory is monitored, repaired and kept current by our team, on our own stack.
Turnkey, compared with assembling it yourself.
Building one in-house means contracting each stage separately and owning the gaps between them.
| Contracting each stage | Turnkey with us | |
|---|---|---|
| Counterparties | Site, hardware, integration and operations, signed one by one | One team across design, sourcing, build and operations |
| Who sizes power and cooling | Whoever is engaged first, before the hardware is chosen | Design, with the hardware order matched to the site |
| Acceptance | Defined per supplier, often after delivery | Acceptance tests agreed at signing, with results for every node |
| After handover | Your team, or a new operator learning the cluster | The team that built it, to an SLA, or your team if you prefer |
Compared by category, not by provider.
A factory grows wave by wave.
Large AI factories are built in monthly waves of 32 nodes. Each wave is racked, burned in and accepted on its own, on a fabric laid out for the size the factory is planned to reach.
- Network
- spine-leaf
- Typical size
- 32 to 128 nodes
- Fabric for the final sizeRoCE v2 Ethernet, spine-leaf, designed once so later waves plug in without re-cabling the first.
- Every node under loadSustained burn-in on each node and collective tests across the fabric, before any wave is accepted.
- Dates from signingThe proposal dates each wave in weeks from signing, so capacity arrives on a schedule you can plan against.
Four ways to own an AI factory.
The stages stay the same. Ownership and who runs the factory after handover change with the way you work with us.
| Way to work with us | Owner | Design | Source | Build | Operate | Software |
|---|---|---|---|---|---|---|
| Cluster deploymentWe design, source and build it, then hand it over burned in. Your team runs it. | You | |||||
| Managed clusterYou own it, we run it. We can build it first, or take over a cluster that's already live. | You | |||||
| SoftwareYour team runs every stage, on our stack. | You | |||||
| Reserved clusterWe own and run it. You reserve it for 60 months, with rollover into the next generation. | Lightyard |
Where an AI factory can go.
Any of the four, picked at the design stage for the power, region and timing the factory needs.
The NVIDIA platforms a factory is built on.
Air-cooled HGX nodes or liquid-cooled NVL72 racks, chosen against the workload and what the site can take.
- Rack scaleNVIDIA GB300 NVL72
- Next generationNVIDIA Vera Rubin
Teams that need a whole factory, not a few GPUs.
01Neoclouds
Capacity built to your standard, ready to sell.
You bring customers and a team. We build the cluster.
GPU clusters for neoclouds03Enterprises & sovereign
An AI factory in your region.
Your own cluster in the jurisdiction you need, on your site or ours.
Private and sovereign AI05Asset owners
A running cluster, operated to an SLA.
You bring hardware, a hall or power. We run it.
Clusters for asset owners06Investors
An operated GPU asset, without building an operator.
You bring capital. We design, build and run it.
GPU infrastructure for investors
Questions we get on the first call.
Something missing? Ask us directly. We usually reply within a few hours.
A GPU cluster built for one organization's AI work, training and inference, together with the site, power, cooling and network it runs on. The term is NVIDIA's; in practice it means a dedicated cluster sized as one system.
We take every stage: site selection if you need it, design, sourcing, build, burn-in and operations. You can also take some stages yourself, such as running the factory after handover.
Your choice. You or your investors own it and we build and run it, or we build and hand it over, or we hold it and you reserve it for the term.
Most clusters are 32 to 128 nodes; larger ones are scoped case by case.
Current NVIDIA platforms, from HGX B300 nodes to GB300 NVL72 racks and Vera Rubin as it ships. The design stage picks one against the workload and the site.
The platform, the size and the site set the date. The proposal dates every wave in weeks from signing, and large factories come up a wave at a time.
In a Tier III colocation facility in the US, Europe, Asia and South America, in your own data center, in a retrofitted building or mining site, or in a modular build on land with power.
Quoted per project, once the platform, the size, the site and who operates it are settled.
Solutions, sites and who we build for.
- AI factoryEvery stage, from the site to a running cluster.
- Cluster deploymentDesigned, sourced, built and handed over burned in.
- Managed clustersYou own the cluster. We build it or take it over, and run it.
- Reserved clustersWe own and run it. You reserve it for the term.
- SoftwareRun every stage yourself, on our stack.
- White-label GPU cloudYour GPU cloud under your brand, built and run by us.
- ColocationA Tier III facility in the region you choose.
- On-premYour data center, with the cluster designed around it.
- RetrofitA building or mining site upgraded for GPU density.
- ModularA prefabricated site on land with power.
- NeocloudsCapacity built to your standard, ready to sell.
- Venture capitalOne cluster, shared by your portfolio companies.
- Enterprises & sovereignYour own cluster in your region.
- Private equityGPU capacity on portfolio sites, or a neocloud as a platform.
- Asset ownersHardware, a hall or a power site, turned into a running cluster.
- InvestorsAn operated GPU asset, without building an operator.