vCluster Labs is strengthening its AI infrastructure capabilities with vMetal, a new solution focused on managing bare metal GPU resources across Neoclouds and AI factories.
Designed to bring cloud-like automation to physical infrastructure, vMetal enables organizations to provision, assign, upgrade, and repurpose GPU servers through a centralized control plane. The platform automates the lifecycle of bare metal machines, helping infrastructure operators improve efficiency and scalability while reducing manual intervention.
The launch comes as organizations face increasing complexity in building and managing GPU-driven environments. From hardware provisioning and lifecycle management to cluster orchestration and AI tooling, fragmentation across infrastructure layers continues to pose operational challenges. vMetal addresses this gap by bridging physical infrastructure with higher-level orchestration frameworks.
According to Torsten Volk, Neocloud providers must balance optimal GPU utilization with strong workload isolation. He noted that improving the interface between physical hardware and Kubernetes environments is critical to enhancing both efficiency and security in GPU allocation.
As part of the broader vCluster Platform, vMetal integrates with tenant orchestration and AI-ready environments to deliver a unified infrastructure stack. This allows organizations to run AI workloads seamlessly—from bare metal servers to production-ready deployments—within a single operational framework.
At the orchestration layer, vCluster enables secure multi-tenant Kubernetes environments, allowing teams to isolate workloads while maximizing resource utilization. Virtual clusters support self-service capabilities and help maintain clear boundaries between tenants on shared infrastructure.
At the application level, vCluster Certified Stacks provide pre-configured AI environments that combine tenancy models, isolation policies, and AI tooling into deployable blueprints. These stacks are designed to reduce setup complexity and accelerate time to deployment.
The initial Certified Stacks include integration with NVIDIA Run:ai, supporting scalable GPU orchestration with performance and fairness controls. Omri Geller highlighted the need for coordinated infrastructure, scheduling, and isolation to support large-scale AI workloads.
Lukas Gentele noted that the platform aims to make infrastructure as dynamic as the workloads it supports, combining machine management, orchestration, and AI environments into a single solution.
