Each customer gets a self-service cloud on dedicated hardware, where they provision machines, connect over SSH, and launch Kubernetes, Slurm, or inference clusters
vCluster Labs, the platform operators use to build and run their own AI cloud like a hyperscaler, today announced vCluster Platform v5.0, now in limited availability. Platform v5.0 enables AI cloud providers to turn bare metal offtake into private customer clouds. Each customer gets a self-service environment where they provision their own machines, choose OS images, connect over SSH, and launch their own clusters, on hardware the provider owns and controls.
Platform v5.0 treats the tenant as a first-class object. Administrators see every machine, network, and image across the datacenter in a single inventory, then grant each customer the resources and cluster templates they are entitled to, with quotas governing how much capacity they can hold. From there, customers provision, connect, and launch clusters without filing a ticket.
As bare metal capacity becomes more widely available, providers face a growing margin problem. Bare metal earns roughly 14% gross margin after depreciation, and more than half of a typical provider's revenue comes from one or two customers, according to McKinsey. Selling compute and managed services returns roughly 2.5 times the revenue per megawatt that colocation does, per company filings and Bernstein analysis. Adding cloud services alongside bare metal is a natural move for providers looking to improve unit economics, and Platform v5.0 runs both businesses from one stack: bare metal offtake and managed Kubernetes, Slurm, and inference clusters.
“With 70% of organizations prioritizing AI investments and 76% already running GPU workloads, demand for dedicated AI infrastructure keeps climbing, and buyers expect it to behave like a cloud,” said Paul Nashawaty, practice lead and chief analyst, ECI Research. “AI cloud providers that can give each customer a self-service environment on dedicated hardware, and run bare metal and managed services from one platform, are better positioned to turn that capacity into durable revenue.”
What vCluster Platform v5.0 Adds to Bare Metal Offtake
With vCluster Platform v5.0, a provider defines a customer once, and every machine, network, image, and cluster beneath that customer inherits the boundary. That works for a provider selling bare metal alone, and the managed layers above it sit on the same platform whenever the provider wants them. "Bare metal is a real business, and v5.0 makes it faster to run," said Lukas Gentele, Co-Founder and CEO of vCluster Labs.
vMetal, the infrastructure orchestrator beneath vCluster Platform, presents every bare metal machine in the datacenter as a single inventory. Providers choose their provisioning driver, NVIDIA Infra Controller (NICo) or Metal3, and their network automation layer, NICo or Netris. Machines can be assigned as reserved capacity for committed customers or left on-demand, and moving them between customers is a reassignment, not a teardown.
Customers manage themselves down to bare metal. They provision machines from their assigned capacity with the provider's certified golden images, launch Kubernetes, Slurm, or inference clusters from the provider's templates, and add their own SSH keys. Each customer runs on dedicated bare metal with their own network environment, following the Private Nodes model vCluster Labs recommends for customer-facing offerings.
"Customers who commit to dedicated capacity want to run it themselves," said Seth Demsey, Co-Founder and Co-CEO of Corvex. "Their platform teams want to divide capacity among internal groups and give developers a fast path to compute. vCluster Platform v5.0 gives our customers a self-service environment of their own on dedicated hardware."
Providers get the vCluster console on day one. Every tenant, machine, image, and network is an API object, so a provider building a branded portal uses the same API the console runs on.
"Providers already have the metal, the network, and the scheduler," Gentele added. "Turning that into something a customer can buy on day one shouldn't take a second platform."
Availability
vCluster Platform v5.0 is in limited availability starting today. AI cloud providers can request more information on private customer clouds.
Supporting resources
- Solution page for private customer clouds on vCluster
- Documentation for Machines, Network Environments, and OS Images
About vCluster Labs
vCluster Labs is the platform operators use to build and run their own AI cloud, like a hyperscaler. vCluster turns raw GPUs into cluster products they can sell, from bare metal provisioning up through managed Kubernetes, Slurm, and inference clusters, with tenant isolation at the infrastructure layer. The result is more margin per GPU, less dependence on a few big customers, and new managed services live in days instead of months. Trusted by fast-growing AI cloud providers including Nebius, Groq, Deutsche Telekom and Corvex, and by enterprises including Adobe, Samsung, and Deloitte, across 100K+ GPUs.
View source version on businesswire.com: https://www.businesswire.com/news/home/20261119596356/en/
Contacts
Media contact
Heather Fitzsimmons
Mindshare PR
heather@mindsharepr.com
650-279-4360