Red Hat OpenShift Virtualization 4. 22 has achieved NVIDIA-Certified Hypervisor status for NVIDIA HGX systems, validating near bare-metal GPU performance for representative AI and accelerated computing workloads. You no longer have to choose between the flexibility of VMs and the scale of Kubernetes for your AI pipeline. GPU-accelerated virtual machines and containerized AI workloads can run on the same cluster, under the same API, managed by the same team.

The NVIDIA-Certified Hypervisors program validates that virtualization platforms deliver near bare-metal performance by accurately exposing hardware topology and implementing key performance optimizations. Certification testing covers compute, memory, data-path efficiency, and LLM inference workloads, the performance-critical behaviors that enterprise AI deployments depend on. For Red Hat customers, this certification confirms what many have already experienced: OpenShift Virtualization is ready for demanding GPU workloads in production.

OpenShift Virtualization is one of the inaugural Kubernetes-native hypervisors to achieve this certification. Unlike traditional hypervisors, OpenShift Virtualization runs virtual machines as Kubernetes pods, giving GPU-accelerated VMs the same orchestration, lifecycle management, and operational model as containerized applications. This means enterprises can consolidate GPU workloads whether they run in VMs or containers on a single OpenShift cluster, managed through a unified Kubernetes API.

For organizations modernizing their virtualization infrastructure, this eliminates the need to maintain separate stacks for traditional VM workloads and cloud-native AI applications. OpenShift Virtualization is built on KubeVirt, a CNCF project, running on KVM. Combined with Red Hat's enterprise lifecycle support and certified hardware ecosystem, customers get NVIDIA's performance validation with the operational assurance of Red Hat's platform. Customers running Red Hat OpenShift 4. 22 can enable OpenShift Virtualization through the OperatorHub.