platform-eng
Bare Metal Kubernetes OS for GPU Clouds
vMetal runs a lightweight Kubernetes binary directly on your hardware, eliminating k3s and kubeadm dependencies so GPU clouds maximize performance without layering complex distributions.
Trusted by the fastest-growing AI cloud providers
Problem
Why Bare Metal K8s Deployments Stall
Traditional Kubernetes distributions add layers that cost performance, time, and revenue.
Overhead from Stacked Distributions
Layering k3s or kubeadm on bare metal hardware introduces unnecessary dependencies that drain GPU performance and slow provisioning cycles.
Slow Path from Rack to Revenue
Building a bare metal Kubernetes OS yourself takes 6 to 10 engineers, 6 to 12 months, and significant capital before a single tenant is live.
Isolation Without a Performance Tax
Standard Kubernetes forces you to choose between tenant isolation and operational efficiency, leaving GPU workloads exposed or underutilized.
Solution
One Binary from Bare Metal to Kubernetes
vCluster Standalone runs as a single binary directly on Linux hardware with no external Kubernetes dependency. vMetal adds zero-touch provisioning, GPU lifecycle management, and kernel-native workload isolation so AI cloud providers go from bare metal racks to production Kubernetes in days.
Built for Bare Metal Kubernetes at GPU Scale
Every layer from hardware provisioning to workload isolation is purpose-built for GPU clouds running directly on bare metal.
Kubernetes Distribution
Bare Metal Kubernetes Without Dependencies
vCluster Standalone runs as a single binary on bare metal Linux, replacing k3s, kubeadm, and k0s as the base layer. No intermediate runtime required, no overhead stacked between your hardware and your Kubernetes control plane.
- No k3s or kubeadm required
- Single binary on bare metal Linux
- Lightweight control plane, zero extra layers
Hardware Lifecycle
Zero-Touch Bare Metal Provisioning
PXE boot, OS installation, machine registration, and network configuration handled automatically. vMetal manages the full GPU server lifecycle from initial rack to decommission, turning hardware into production Kubernetes nodes without manual intervention.
- PXE boot and OS install automated
- Full GPU server lifecycle management
- Network automation via Netris integration
Dynamic Scaling
Auto-Provision GPU Nodes on Demand
Auto Nodes acts as Bare Metal Karpenter, automatically provisioning GPU servers via Terraform when tenants schedule workloads. Physical nodes scale in response to real demand without manual provisioning queues.
- Scales bare metal GPU nodes automatically
- Terraform-driven provisioning per workload
- Eliminates manual node provisioning queues
Workload Security
Kernel-Native Isolation on Bare Metal
vNode (currently in private beta) provides kernel-native workload isolation using seccomp, cgroups, namespaces, and AppArmor per workload. Tenants run securely on the same bare metal Kubernetes OS with no hypervisor tax and no GPU performance penalty.
- seccomp, cgroups, AppArmor per workload
- No hypervisor overhead on GPUs
- Prevents container breakout at kernel level
Network Isolation
Hardware-Enforced Per-Tenant Networking
VLANs, VXLANs, VRFs, ACLs, and DPU policies enforce hard network boundaries between tenants sharing the same bare metal Kubernetes infrastructure. Powered by Netris integration for automated network lifecycle management.
- VLANs and VXLANs per tenant
- DPU policy enforcement on hardware
- Automated network lifecycle via Netris
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 a bare metal Kubernetes OS and how does vMetal differ from standard distributions?
A bare metal Kubernetes OS runs Kubernetes directly on physical hardware without an intermediate operating system layer or dependency like k3s or kubeadm. vMetal delivers vCluster Standalone as a single binary that runs directly on bare metal Linux, eliminating the need for any external Kubernetes base layer. This reduces overhead, simplifies the software stack, and maximizes available GPU performance for AI cloud providers and data center operators.
Does vCluster Standalone require k3s, kubeadm, or any existing Kubernetes cluster?
No. vCluster Standalone is specifically designed to run as a binary directly on bare metal Linux with no external Kubernetes dependency. It replaces k3s, kubeadm, and k0s as the base layer rather than sitting on top of them. This is the core distinction between vCluster Standalone and conventional Kubernetes distributions, which require an existing host cluster or OS-level Kubernetes runtime before tenant workloads can be scheduled.
How does vMetal handle bare metal GPU server provisioning at scale?
vMetal handles the full GPU server lifecycle through zero-touch provisioning. When a new server is racked, vMetal automates PXE boot, OS installation, machine registration, and network configuration without manual steps. Auto Nodes extends this to dynamic scaling, automatically provisioning bare metal GPU nodes via Terraform when tenants schedule workloads, so capacity responds to real demand rather than manual provisioning queues.
How does workload isolation work on a shared bare metal Kubernetes OS?
vNode (currently in private beta) provides kernel-native isolation using seccomp, cgroups, Linux namespaces, and AppArmor per workload, preventing container breakout without adding hypervisor overhead. Network isolation is enforced at the hardware level through VLANs, VXLANs, and DPU policies via Netris. This layered approach lets multiple tenants share the same bare metal Kubernetes infrastructure while maintaining strong security boundaries and preserving full GPU performance.
Is vCluster's Kubernetes distribution CNCF-certified?
Yes. vCluster provides a CNCF-certified Kubernetes distribution, meaning each tenant cluster passes full Kubernetes conformance testing and is 100% API-compatible. GPU cloud providers can offer their customers a standard, conformant Kubernetes experience on bare metal hardware without proprietary API deviations or compatibility limitations.
How quickly can a GPU cloud provider go from bare metal hardware to serving tenants?
Lintasarta launched Indonesia's leading GPU cloud in 90 days using vCluster and vMetal, deploying more than 170 isolated tenant clusters on bare metal infrastructure. Boost Run launched a managed Kubernetes service in less than 45 days with zero new platform engineering hires. The integrated stack from bare metal provisioning through tenant cluster orchestration removes the years of engineering typically required to build equivalent infrastructure from scratch.
Launch Your Bare Metal Kubernetes Platform
See how GPU cloud providers go from bare metal to production Kubernetes in days.
