vCluster

vCluster is a Kubernetes multi-tenancy and virtual cluster platform, formerly known as Loft, that enables platform engineering teams to create and manage isolated tenant clusters with the elasticity and self-service capabilities of a hyperscale cloud provider. The open-source vCluster project has accumulated over 30 million pulls and powers infrastructure at 50-plus GPU clouds and Fortune 500 enterprises including NVIDIA, JPMorgan Chase, and Adobe. The platform runs 100,000-plus GPU nodes and over 1 million CPU nodes in production worldwide. Each virtual cluster provides a fully isolated Kubernetes API server, controller manager, and scheduler inside a host cluster, delivering control plane isolation with separate namespaces, RBAC, CRDs, and network policies. Workload isolation supports both shared nodes for cost efficiency and private dedicated nodes for strict hardware-level separation on bare metal. The platform deploys on EKS, GKE, AKS, Rancher, Red Hat OpenShift, Tanzu, KubeVirt, standalone bare metal, and private cloud. QumulusAI reduced isolated Kubernetes environment provisioning from hours to 1 minute per cluster. Boost Run achieved production launch in under 45 days. Lintasarta operates 170-plus virtual clusters in production. Polarise enforces 100 percent data residency at the infrastructure layer. vCluster is a CNCF-certified Kubernetes distribution and the only vendor named in ClusterMAX security criteria for multi-tenancy isolation. The company introduced vMetal in 2026, extending virtual cluster capabilities to bare metal GPU racks for AI factory and AI cloud provider use cases.