Red Hat initiated KubeVirt in 2016, intending for Kubernetes to manage virtual machines alongside containers, allowing users to run older workloads on the same platform as modernized applications.
This objective persists, yet since version 1.0, external projects have built upon KubeVirt, integrating it into a broader ecosystem. Examples include Open Sovereign AI Cloud (OSAC) utilizing it for VM, cluster, and bare-metal services, and Cluster API Provider KubeVirt (CAPK) using KubeVirt VMs as Kubernetes nodes.
KubeVirt is currently a CNCF incubating project, with participation from entities such as Nvidia, Microsoft, AMD, Intel, and SUSE. The community is progressing toward its graduation as a full CNCF project.
Enhancement trackers for versions v1.8, v1.9, and v1.10 illustrate the practical scope of "next." The queue of items has grown from a small set to over a hundred between v1.6 and v1.10, with v1.9 introducing dozens of new proposals. The KubeVirt Enhancement Proposal (VEP) process maintains core stability while the queue expands, relying on designs preceding code, a single tracker, and ownership by special interest groups (SIGs).
New developments include hardware support and confidential VMs. Devices, DRA, and new hardware GPUs on VMs can now be claimed via Kubernetes Dynamic Resource Allocation, a beta feature in v1.9. This claim model is expanding to generic host devices, network devices like SR-IOV, and CPUs, enabling a VM to request coordinated allocations of GPU, NIC, and CPUs.
Confidential VMs address the need to run sensitive AI and application workloads on shared infrastructure while maintaining data protection during execution. KubeVirt is increasing support for confidential VMs and exploring integration with attestation and secret-release services, including Trustee from the Confidential Containers project. Fan Zhang, co-lead of Kubevirt's AIE and Confidential Compute working groups, stated, "For AI, the trust boundary should extend across the full execution path—CPUs, accelerators, and the interconnects between them."
Support is expanding across Intel Trust Domain Extensions (TDX), AMD Secure Encrypted Virtualization (including SEV-SNP), IBM Secure Execution, and emerging Arm CCA. In v1.9, the feature gate for AMD SEV and SEV-SNP reached beta, Intel TDX support remained alpha, and IBM Secure Execution support achieved general availability (GA).
Regarding extensibility, structured plugins offer a path for features outside the core, allowing users to implement functionality while retaining KubeVirt API guarantees. This is alpha in v1.9. Another area is the virtualization stack itself; while libvirt, QEMU, and KVM remain the default, the goal is to make this stack pluggable, allowing out-of-tree solutions like Cloud Hypervisor or Firecracker to operate without forking KubeVirt.
Harshit Gupta, senior software engineer at Microsoft, commented, "KubeVirt enables customers to modernize VM workflows by bringing Kubernetes orchestration, storage, and networking to virtual machines." He added, "Extending these benefits to virtualization solutions built on alternative stacks adds user flexibility."
The evolution reflects a shift from the initial focus on migrating legacy workloads to empowering teams to deploy modern, next-generation platforms across diverse environments, such as AI inference and hardened multi-tenant setups.
