Nvidia Kubernetes Engine (NVKE)

NVIDIA Kubernetes Engine (NVKE) is a management plane for platform teams that want to offer managed Kubernetes clusters on their own NVIDIA GPU infrastructure. Operators install NVKE on a management cluster, then use its API or the nvkectl command-line tool to create and operate separate workload clusters. Each workload cluster has its own control plane, etcd, and node pools, and customers access it through Kubernetes tools such as kubectl. NVKE provisions machines through NVIDIA Infrastructure Controller (NICo), supports cluster and node-pool lifecycle tasks including scaling and upgrades, and uses TPM-based SPIFFE/SPIRE attestation to establish node identity. Etcd data encryption is available per cluster. Its companion project, NVIDIA Kubernetes Recipes, builds and signs the artifacts that NVKE deploys.

NVKE is a preview, not recommended for production, and its APIs and behavior may change. NICo is currently the only supported infrastructure provider. Real installations require operators to supply and configure machines, networking, node images, registries, and access to NICo. Security and operational gaps are significant: API authentication is off by default and access uses plain HTTP unless operators secure it; the management plane cannot upgrade, scale, or replace itself, and there is no management-plane upgrade or backup-and-restore procedure. Evaluation is possible without GPU hardware through a local quickstart using simulated machines, subject to Linux-host resource and tooling requirements; macOS users can use a Colima-based guide. The Apache-2.0-licensed Go project also provides build instructions, deployment and operations guides, and contribution procedures.