Nutanix NKP 2.19 adds bare-metal Kubernetes and an AI catalog

Dual native infrastructure with NKP Metal

According to Nutanix, platform teams face two pressures at once: getting production-ready AI infrastructure to developers quickly, and preventing virtual machines and containers from becoming separate operational silos. NKP 2.19 is designed to address both.

The central piece is NKP Metal, which Nutanix says is built to remove the operational friction of bare-metal Kubernetes. It uses Foundation Central for automated node discovery, Ubuntu OS imaging and cluster provisioning. Platform teams can then deploy containerized workloads directly on physical hardware, using the same operational workflows as virtualized clusters. Nutanix says this gives the raw performance needed for demanding, low-latency applications, while a single control plane covers both bare metal and enterprise virtualization.

Canonical, the company behind Ubuntu, backs the approach. Ivan Ramos Garza, VP of Alliances and Verticals at Canonical, said that through Foundation Central, "NKP Metal customers on Ubuntu now have an accelerated, automated path to production Kubernetes."

Storage, upgrades and a VM Operator

NKP 2.19 includes several operational additions:

• AOS multi-writer support: apps can use ReadWriteMany file storage delivered directly from a Nutanix AOS cluster through the Nutanix CSI driver. Nutanix says this simplifies sharing data across pods and suits CI/CD caches and logs.
• UI-driven upgrades: cluster upgrades can be run from the interface, with pre-flight checks and post-upgrade validation meant to help identify configuration issues and support maintenance planning.
• VM Operator for Nutanix AHV: delivered through the NKP App catalog, this Kubernetes operator lets teams manage supported virtual-machine lifecycle operations through Kubernetes APIs, so VMs and containers share the same workflows.

AI catalog, AMD GPUs and CNCF conformance

Nutanix argues that moving from AI experimentation to production often means months spent stitching together drivers, operators and MLOps frameworks. NKP 2.19 aims to reduce that effort for supported tooling.

The NKP AI Catalog offers a curated set of AI and machine-learning tools deployable through the NKP user interface. It includes lifecycle management for supported components and guided UI workflows for configuring AI and machine-learning workspaces.

On hardware, NKP now has validated support for AMD Instinct MI350 GPUs and the AMD GPU Operator, available through the NKP Catalog and subject to documented prerequisites. Nutanix presents this as a way to widen hardware choice beyond a single GPU ecosystem.

The company also notes that, starting with version 2.18, NKP is certified under the CNCF Certified Kubernetes AI Conformance Program. According to Nutanix, the program validates support for dynamic GPU resource allocation, gang scheduling and tenant workload isolation.

What it means for platform teams

For organizations running both virtual machines and containers, the release is aimed at reducing the number of separate tools and processes needed to manage them. Teams that want bare-metal performance for Kubernetes no longer have to give up a shared control plane, according to Nutanix.

Nutanix says NKP 2.19 is designed to provide a consistent operating model for supported modern-application deployments. Interested customers can try NKP through the Nutanix Test Drive, consult the documentation for supported APIs, configurations and environments, read the release notes on the Nutanix Portal, or contact their Nutanix team about upgrade options.