ai-cloud
Bare Metal Provisioning for AI Cloud Providers
vMetal delivers zero-touch bare metal provisioning for GPU fleets, eliminating virtualization overhead so AI cloud providers can scale from rack to production without delays.
Trusted by the fastest-growing AI cloud providers
Problem
Why GPU Deployments Stall at Scale
Manual provisioning slows revenue, wastes capacity, and exposes tenants to unnecessary risk.
Slow Server Setup Kills Revenue
Based on publicly reported GPU cloud pricing, provisioning delays of several weeks can represent substantial lost revenue — the actual impact varies by fleet size and utilization.
Manual Processes Introduce Errors
Hand-configured PXE boots, OS installs, and network configs create inconsistencies that are difficult to debug across large GPU fleets.
No Path From Hardware to Kubernetes
Most provisioning tools stop at the OS layer, leaving platform teams to stitch together k3s, kubeadm, and cluster orchestration themselves.
Solution
Zero-Touch Bare Metal Provisioning Built for GPU Clouds
vMetal handles PXE boot, OS install, machine registration, network automation, and full server lifecycle in one integrated stack. vCluster Standalone runs directly on bare metal as a single binary, eliminating intermediate dependencies like k3s or kubeadm and delivering a complete path from GPU racks to isolated tenant clusters.
From Bare Metal to Tenant Clusters in Days
A complete, production-proven stack that takes GPU servers from rack to managed Kubernetes without stitching together third-party tools.
Bare Metal
Zero-Touch GPU Server Provisioning
vMetal handles PXE boot, OS installation, machine registration, and network configuration automatically. Every bare metal GPU server is provisioned consistently, from rack mount to production-ready, without manual intervention.
- PXE boot and OS install automated
- Full machine lifecycle management
- Zero-touch from rack to production
Kubernetes Distribution
Kubernetes Directly on Bare Metal
vCluster Standalone runs as a single binary on bare metal Linux with no external Kubernetes dependency. No k3s, no kubeadm, no extra layers between your GPU hardware and your tenant clusters.
- Single binary, no external K8s dependency
- Runs directly on bare metal Linux
- Replaces k3s and kubeadm entirely
Dynamic Scaling
Automatic GPU Node Provisioning
Auto Nodes acts as bare metal Karpenter, automatically provisioning GPU nodes via Terraform when tenants schedule workloads. GPU capacity scales to demand without manual intervention from your operations team.
- Terraform-driven node provisioning
- Scales to tenant workload demand
- Eliminates manual capacity operations
Tenant Isolation
Isolated Tenant Clusters on Shared Hardware
Every tenant gets a fully isolated Kubernetes control plane with its own API server, etcd, and RBAC, running as a lightweight pod. Spin up hundreds of isolated tenant environments on your GPU fleet without provisioning separate physical clusters.
- Own API server and etcd per tenant
- Spins up in seconds, not hours
- No separate physical clusters required
Network Isolation
Hardware-Enforced Per-Tenant Networking
Network isolation is enforced at the hardware level using VLANs, VXLANs, VRFs, ACLs, and DPU policies via Netris integration, significantly limiting tenant-to-tenant network reachability.
- VLANs and VXLANs per tenant
- DPU policy enforcement via Netris
- Full L2/L3 network boundaries
Why vCluster
This isn’t a side project. Behind every vCluster deployment is 5+ years of deep K8s engineering, security hardening, and battle-tested infrastructure work at massive scale.
100K+
GPU Nodes Powered
50+
GPU Clouds & F500s
<45
Days to Launch
30K
GitHub Stars
Get Started in 3 Steps
1
Schedule a Demo
Talk to our team about your stack
2
Deploy vCluster
Deploy vCluster on your infra in minutes
3
Onboard Your Tenants
Go live with a hyperscaler-grade tenant experience in days
FAQs
What is zero-touch bare metal provisioning and how does vMetal deliver it?
Zero-touch bare metal provisioning means a GPU server goes from rack mount to production-ready without manual steps. vMetal automates PXE boot, OS installation, machine registration, and network configuration, so every server in your fleet is provisioned consistently. Operations teams define the desired state once, and vMetal handles execution across all physical nodes, reducing setup errors and compressing time-to-production significantly.
Do I need an existing Kubernetes cluster to use vMetal?
No. vCluster Standalone runs as a single binary directly on bare metal Linux with no external Kubernetes dependency. You do not need k3s, kubeadm, or k0s as a base layer. This means you can go from raw GPU servers to a fully operational Kubernetes distribution in a single step, without managing the complexity of intermediate infrastructure components.
How does bare metal provisioning connect to tenant cluster orchestration?
vMetal handles the hardware layer, including PXE boot, OS install, and network automation. vCluster Standalone then runs directly on those provisioned servers as a lightweight Kubernetes binary. From there, the vCluster Platform creates fully isolated tenant clusters on top. The result is a single integrated path from bare metal GPU servers to managed, isolated Kubernetes environments for your end customers, without stitching together separate vendor tools.
What GPU fleets and infrastructure scales has this been proven at?
The platform powers 100K+ GPU nodes in production across 50+ GPU clouds and Fortune 500 customers. It is named in the NVIDIA DGX SuperPOD reference architecture. Lintasarta launched Indonesia's leading GPU cloud in 90 days using this stack. vMetal and the vCluster platform are designed for the scale and reliability requirements of AI cloud providers deploying large GPU fleets.
How does Auto Nodes handle GPU capacity scaling on bare metal?
Auto Nodes works like Karpenter but for physical servers. When tenants schedule workloads that require additional GPU capacity, Auto Nodes automatically provisions new bare metal nodes via Terraform to meet that demand. This eliminates the need for operations teams to manually intervene every time a tenant needs more GPU resources, allowing GPU cloud providers to scale capacity dynamically without proportional headcount growth.
Can this stack run in air-gapped or compliance-restricted environments?
Yes. The vCluster Platform supports air-gapped and FIPS deployment for environments with strict compliance requirements. The full stack, including bare metal provisioning, Kubernetes distribution, and tenant cluster orchestration, can be deployed in disconnected environments without requiring outbound internet access. This makes it suitable for government, defense, and regulated enterprise deployments that require isolated infrastructure.
Launch Your GPU Cloud on Bare Metal
See how vMetal takes GPU servers from rack to production-ready tenant clusters.
