Introducing NKP Metal: Unlocking a Dual Native Architecture for Deploying Kubernetes on Bare-Metal and Virtual Servers

Modern IT environments rarely follow a single, monolithic pattern. As platform engineering teams scale their Kubernetes footprints, infrastructure demands vary as widely as the applications themselves. While hypervisors like Nutanix AHV provide flexibility, enterprise density, and dynamic resource management, specific operational scenarios demand a different approach: deploying directly onto physical servers.

First announced at .NEXT in April 2026, Nutanix introduced a new capability for the Nutanix Kubernetes Platform (NKP) solution to meet this exact need: NKP Metal. Extending the benefits of Hyperconverged Infrastructure (HCI) to the Nutanix Kubernetes Platform (NKP), NKP Metal unlocks a true Dual Native Architecture, treating both containers and VMs as core architectural pillars. This makes it possible for platform teams to deploy, secure, and manage both bare-metal and virtualized Kubernetes clusters through a single, unified control plane, gaining direct physical hardware access without sacrificing operational simplicity.

Laying this groundwork, NKP Metal is designed to avoid the need for fragmented, third-party infrastructure scripts by delivering an end-to-end Kubernetes lifecycle, from automating node discovery and OS imaging to declarative cluster scaling and seamless Day-2 upgrades. NKP Metal is currently under development.

The Background: How VMs and Bare Metal Share the Container Stage, and What's Driving the Decision

To understand why NKP Metal matters, it helps to look at how container deployment patterns have evolved over time.

The Promise and Paradox of Containerization

Containers fundamentally transformed modern application development. By packaging code alongside its runtime dependencies, containers enabled software teams to build, ship, and scale applications quickly, making them the de facto standard for building large-scale applications in the public cloud.

Because containers provide an abstraction layer between the application and the host OS, it seems intuitive that containers would serve as a direct replacement for virtual machines (VMs). Why run a guest operating system inside a hypervisor when a container engine can achieve isolation on a shared kernel?

Why Virtualized Containers Became the Standard

Far from being a historical anomaly, running containers on virtual machines was the logical, natural evolution of modern infrastructure economics and operations.

In the public cloud, every major hyperscaler built their managed Kubernetes offerings (such as AWS EKS, Azure AKS, and Google GKE) on top of virtual machines. Virtualization gave cloud providers the multi-tenancy, granular resource slicing, and extreme hardware density needed to deliver cost-effective compute while running highly profitable managed services businesses at scale.

In enterprise data centers, a similar operational logic prevailed. Long before Kubernetes emerged, IT departments had spent over a decade standardizing their tooling, security models, backup routines, and provisioning workflows on virtualized infrastructure. Deploying containers inside virtual machines was the most efficient path forward. It enabled platform teams to adopt cloud-native applications rapidly while leveraging the proven resiliency, dynamic elasticity, and high-density utilization of their existing virtual environments, without having to reinvent their underlying infrastructure operations.

Market Forces Driving Bare Metal

Recently, a confluence of market forces has sparked renewed demand to run containers directly on physical bare-metal hardware.

The rapid rise of generative AI has fundamentally shifted hardware priorities. Organizations making massive investments in modern GPUs want direct, unthrottled access to physical hardware topologies to extract every ounce of processing power and minimize latency across dense compute clusters.

Meanwhile, containerized applications are expanding far beyond core data centers out to remote edge sites, manufacturing floors, retail hubs, and telecommunications towers. These localized, bandwidth-constrained environments often require lean operational footprints, where minimizing cost and hardware footprint is a priority and the operational benefits of virtualization are less impactful.

At the same time, broader shifts across the IT industry are forcing organizations to re-evaluate their long-term infrastructure strategies. Recent disruption to the virtualization landscape are prompting enterprise leaders to re-examine their software stacks and actively seek greater architectural flexibility. This is leading some to consider whether standardizing on an enterprise hypervisor is still the right approach.

Lastly, there are some extremely latency sensitive applications that may have never been virtualized in the first place, in situations where optimizing for lower latency supersedes any efficiency or operational gains from virtualization. The amount of latency added by modern hypervisors is nearly undetectable, but the additional layer may add complexity to troubleshooting efforts.

Bridging the Gap with a Dual Native Architecture

These shifts leave platform teams caught between developer requests for direct bare-metal hardware access and executive demands for operational efficiency and cost management.

A Dual Native Architecture makes this trade-off unnecessary. It enables teams to continue realizing the maximum operational and efficiency benefits of virtualizing containers for standard microservices, while seamlessly granting direct physical access to GPUs and bare-metal nodes when the need arises, all under one operational umbrella.

Caption: Dual Native Architecture enables integrated management of Virtual Machines and Containers within a single platform.

What is NKP Metal? Bringing Enterprise Simplicity to Physical Infrastructure

At a high level, NKP Metal will be a new deployment option for the Nutanix Kubernetes Platform that extends the Nutanix operating model directly onto physical bare-metal servers. Our design objective is to bring the automation, enterprise storage, and lifecycle management that organizations expect from Nutanix directly to bare-metal Kubernetes environments.

The Forgotten Challenges of Bare-Metal Management

Enterprise IT standardized on virtual machines decades ago. It’s likely that many organizations have forgotten just how difficult, brittle, and operationally expensive managing bare-metal infrastructure actually is.

Without a hypervisor or cloud abstraction layer, platform teams running bare metal need to contend with considerable operational challenges:

• Hardware Sprawl and Resource Waste: To maintain strict workload isolation, avoid security cross-contamination, and prevent noisy-neighbor issues, organizations deploying bare metal often have to dedicate entire physical servers to specific tasks. Running separate physical nodes for development, testing, staging, and production (alongside dedicated control plane nodes) frequently leads to low physical server utilization, massive hardware sprawl, and wasted power and cooling.
• Complex Day 0 to Day 2 Operations: Managing physical servers requires dealing with out-of-band management controllers, manual PXE-boot workflows, brittle custom deployment scripts, device driver management, and complex physical network wiring.
• Firmware and OS Drift: Unlike VMs, which can be reprovisioned in seconds, physical servers require individual OS image patching, BIOS updates, and firmware upgrades. Managing these lifecycle tasks manually across hundreds of physical nodes creates operational silos and severe security vulnerabilities.

How Virtualization and HCI Solved These Problems

The original promise of virtualization was straightforward: maximize hardware utilization by consolidating multiple virtual machines onto a single physical server to minimize idle compute. Over time, however, IT organizations discovered that the most profound savings and strategic value didn't just come from server density, but from the operational simplicity that virtualization made possible. In fact, this software-defined abstraction layer is the foundational technology that made modern public clouds possible, turning physical infrastructure into API-driven, on-demand compute.

Virtualization eased operational complexity by decoupling the software stack from the physical machine. Suddenly, servers became software-defined objects that could be provisioned in minutes rather than weeks, moved across physical hosts without downtime, and managed through unified, automated control planes. By insulating day-to-day operations from physical hardware friction, virtualization established the gold standard for enterprise resiliency, dynamic resource allocation, and operational efficiency.

Hyperconverged Infrastructure (HCI) built directly on these principles by expanding the benefits of virtualization beyond compute to include storage and networking. By collapsing these traditionally separate tiers into a single software-defined fabric, HCI enabled IT teams to manage their entire infrastructure holistically. Compute, storage, and networking resources could now be scaled effortlessly, patched and upgraded seamlessly without application downtime, and operated with unmatched resiliency and built-in redundancy.

Bringing the Nutanix Operating Model to Bare Metal

NKP Metal brings the operational benefits of HCI directly to bare-metal Kubernetes. It is designed to give organizations the best of both worlds: the direct, unthrottled hardware access required for GPU workloads or lean edge sites, paired with the automated lifecycle management, resiliency, and enterprise data services of Nutanix.

It’s important to note that NKP has long supported pre-provisioned bare-metal nodes. While these pre-provisioned nodes can be successfully leveraged as part of NKP's broader fleet management capabilities, they differ from NKP Metal with respect to how the physical nodes are managed from a lifecycle perspective. With pre-provisioned nodes, customers are responsible for provisioning the operating system themselves and manually managing ongoing patching and firmware updates. In contrast, NKP Metal delivers an automated, production-grade stack, integrating automated server provisioning, OS patching, and infrastructure automation capabilities driven by Foundation Central.

Defining NKP Metal and Its Core Capabilities

Specifically,our design for NKP Metal will introduce four foundational capabilities to the Nutanix cloud-native stack:

• Automated Server Provisioning & Lifecycle Management: NKP Metal will automate the bare-metal host lifecycle. It will leverage Nutanix Foundation for automated bare-metal node discovery and deployment and Nutanix Lifecycle Manager (LCM) for automatic provisioning of bare-metal servers, patch base operating systems, and update physical hardware firmware—bringing cloud-like operational consistency to bare-metal servers.
• Integrated Storage Running in Kubernetes Pods: Rather than requiring external storage arrays or complex manual storage provisioning, NKP Metal will include containerized storage running natively within Kubernetes pods (Cloud Native AOS data services). Our design objective is to provide high-performance, resilient persistent block and file storage directly to containers, tightly integrated with Nutanix Data Services for Kubernetes (NDK) for application-consistent snapshots, disaster recovery, and data protection.
• Unified Dual Native Coexistence: NKP Metal will make it possible to run physical bare-metal Kubernetes clusters side-by-side with virtualized clusters within the same enterprise environment. Our design objective is to allow platform teams to manage both substrates under a single, unified operational framework, breaking down infrastructure silos.
• Flexible Control Plane Virtualization for Bare-Metal Clusters: To maximize efficiency and resiliency, NKP Metal is designed to enable  platform engineers to virtualize the Kubernetes control plane—running control plane nodes as highly resilient, elastic VMs on Nutanix AHV—while running the actual worker nodes directly on physical bare-metal servers. This capability will avoid the need to dedicate multiple expensive physical servers just to host a bare-metal cluster's control plane, protecting high-availability governance while dedicating physical bare metal purely to compute-intensive worker workloads.

Future-Ready Enterprise Infrastructure

By combining the operational strengths of virtualization with the direct hardware access of physical servers, Nutanix aims to deliver true architectural freedom. NKP Metal is designed to offer platform teams the operational flexibility to run virtualized Kubernetes where density and flexibility matter, and bare-metal Kubernetes where hardware access and stack reduction demand it—all within a single, enterprise-ready platform.