Nutanix Kubernetes Platform (NKP) is the Kubernetes management product of Nutanix, and Nutanix Enterprise AI (NAI) is its model-serving product for generative AI. Both are licensed under the Nutanix License and Services Agreement as term licences that are not tied to hardware.[1][3] NKP is counted on the physical cores or vCPUs of worker nodes. NAI is counted on GPU memory, or on vCPUs where a cluster has no GPUs. This article covers the editions, packages, SKUs and counting rules for both. The wider Nutanix landscape is summarised in Nutanix licensing.
Editions
NKP
Nutanix sells NKP in three editions. NKP Starter “is not available to purchase separately. It is included with NCI Pro and Ultimate.”[1] Catalog: NKP Starter included with NCI Pro and Ultimate, not sold separately.
| Edition | How it is obtained | Deployment | Selected inclusions |
|---|---|---|---|
| NKP Starter | Included with NCI Pro and Ultimate | NCI (AHV only) | Unlimited NKP clusters, upstream Kubernetes, cluster lifecycle management, Rocky Linux only |
| NKP Pro | Standalone or add-on to any NCI cluster | NCI, vSphere, bare metal, public cloud | Bring your own operating system, observability stack, backup and restore, service mesh |
| NKP Ultimate | Standalone or add-on to any NCI cluster | NCI, vSphere, bare metal, public cloud | Fleet management, multitenancy, GitOps, cost management, management of EKS and AKS clusters |
Sources: NKP datasheet and Software Options.[1][2] The Software Options page states that NKP Pro and Ultimate “can be purchased as add-on licenses to any NCI cluster”.[2]
NKP Full Stack
NKP Full Stack is a bundle “that includes NCI, NKP, NDK, NUS, and NDB for Pro and Ultimate editions”.[1] The tiers pair NCI and NKP editions as follows:
| Component | Full Stack Starter | Full Stack Pro | Full Stack Ultimate |
|---|---|---|---|
| NCI | NCI Starter | NCI Pro | NCI Ultimate |
| NKP | NKP Starter | NKP Pro | NKP Ultimate |
| Nutanix Data Services for Kubernetes (NDK) | Not included | Included | Included |
| Nutanix Unified Storage (NUS) | Not included | NUS Pro 5 TiB per cluster | NUS Pro 5 TiB per cluster |
| Nutanix Database Service (NDB) | Not included | 10 vCPUs per cluster | 10 vCPUs per cluster |
Source: NKP datasheet.[1] NKP Full Stack Ultimate also “includes metro availability and sync replication”. More NUS or NDB capacity is bought as separate NUS mixed-mode or NDB add-on licences.[1]
NKP Cloud Kubernetes
NKP Cloud Kubernetes is an Ultimate-only SKU “explicitly designed to manage the EKS or AKS environments”.[1] The Software Options page explains that a dedicated NKP management cluster is deployed for these environments, and that it “requires and uses the same licenses”.[2]
Nutanix Enterprise AI
NAI “is available in three packages”:[2]
| Package | Tiers | Contents |
|---|---|---|
| Nutanix Enterprise AI Full Stack (NAI-GPT) | Full-Stack Pro, Full-Stack Ultimate | NAI Pro, NCI Pro or Ultimate for all licensed CPU cores, NKP Pro or Ultimate for all licensed vCPUs, NDK, NUS Pro 5 TiB per cluster, NDB 10 vCPUs per cluster |
| Nutanix Enterprise AI and NKP (NAI-NKP) | NAI-NKP Pro, NAI-NKP Ultimate | NAI Pro plus NKP Pro or Ultimate for all licensed vCPUs |
| Nutanix Enterprise AI Standalone | NAI Pro | NAI Pro “for use on any CNCF-certified Kubernetes anywhere, including three-tier and Public Clouds” |
Source: Software Options, Nutanix Enterprise AI section.[2] In every package, NAI itself is licensed as “NAI Pro for all licensed vCPUs or GPU RAM GBs”.[2]
Metrics
Three rows of the Nutanix metric catalog apply:
- Core for NKP on bare metal, and for NKP Full Stack and the NCI part of NAI Full Stack.
- vCPU for NKP on virtual machines, including the public cloud, for NKP Cloud Kubernetes, and for NAI on clusters without GPUs.
- GPU RAM GB for NAI on inference clusters with GPUs.
Counting / floors
NKP standalone
The NKP datasheet (2026-05-21) states that NKP “is priced on either the number of physical CPU cores if running on bare metal or the number of virtual CPU cores (vCPUs) in your deployment if running on VMs (including in the public cloud), for all the worker nodes (including the worker nodes in a management cluster)”.[1] The Software Options page (2026-09-04) adds that licences apply to worker nodes “in both management and workload/attached clusters”, and that “Control plane nodes do not require licenses.”[2] Catalog: NKP: license all worker-node cores or vCPUs; control plane nodes exempt.
In practice, an NKP count is the sum of the worker nodes of every management cluster and every workload cluster it attaches. Each node is counted in physical cores or in vCPUs, depending on where it runs. Control plane nodes are left out.
NKP Full Stack
“NKP Full Stack is priced per physical CPU core, and licensing must cover the entire number of cores for the full capacity of the cluster being licensed.”[1] Full Stack licences need “no additional NKP licenses required for worker nodes”. The NUS and NDB entitlements in the bundle “are provided strictly on a per cluster basis and cannot be pooled or reallocated across clusters”.[2] Catalog: NKP Full Stack: all physical cores; included NUS and NDB are per cluster and cannot be pooled.
As a result, three Full Stack Pro clusters carry three separate 5 TiB NUS Pro entitlements. Unused capacity on one cluster cannot cover a shortfall on another.
NKP Cloud Kubernetes
NKP Cloud Kubernetes “is priced per virtual CPU cores (vCPUs), and licensing must cover the entire number of vCPUs for all the worker nodes (including the worker nodes in a management cluster)”.[1]
Nutanix Enterprise AI
“NAI is metered by either aggregate GPU RAM for inference clusters with GPUs installed*, or by vCPUs used by worker nodes running inferencing for clusters with no GPU accelerators.” The footnote reads: “GPU RAM GB license is provided in 1 GB entitlement increments that are quantified across all GPUs in a cluster.”[2] Catalog: Nutanix Enterprise AI metered by GPU RAM GB, or by vCPU where no GPUs.
The GPU metric counts the memory of the GPUs installed in an inference cluster, added up across the cluster. It does not count the number of GPU cards or the share of GPU memory a model actually uses.
Mixing editions and management clusters
The NKP datasheet sets three deployment rules:[1]
- “Mixing different NKP edition licenses in the same NKP management cluster isn’t supported.”
- “If the same NKP management cluster must manage NKP and EKS or AKS, the only possible licensing option is the NKP Ultimate edition.”
- NKP Cloud Kubernetes “is available when only managing EKS or AKS. A licensed management cluster running NKP and licensed with NKP Cloud Kubernetes is required.”
Catalog: NKP: one edition per management cluster; managing EKS or AKS alongside NKP requires Ultimate.
SKUs and terms
Nutanix publishes the NKP SKU formats in the datasheet:[1]
| Type | SKU format | Example |
|---|---|---|
| New NKP licence | SW-NKP-[EDITION]-[SUPPORTLEVEL] | SW-NKP-PRO-PR |
| New NKP Full Stack licence | SW-NKP-FS-[EDITION]-[SUPPORTLEVEL] | SW-NKP-FS-ULT-PR |
| New NKP Cloud Kubernetes licence | SW-NKP-C8-ULT-[SUPPORTLEVEL] | SW-NKP-C8-ULT-PR |
| Renewal | RSW-NKP-[FS:C8]-[EDITION]-[SUPPORTLEVEL] | RSW-NKP-PRO-PR |
| Upgrade | USW-NKP-[FS]-[FROM:EDITION]-[FROM:SUPPORTLEVEL]-2-[TO:EDITION]-[TO:SUPPORTLEVEL] | USW-NKP-PRO-PR-2-ULT-PR |
Edition codes are STR (Starter, Full Stack only), PRO and ULT. Support level codes are PR (Production), MC (Mission Critical), FP (Federal Production), FM (Federal Mission Critical), AP (Authorized Partner, Production) and AM (Authorized Partner, Mission Critical).[1]
“All SKUs deliver term licenses and are configurable for any duration between 12-60 months. Licenses must be renewed for continued access to software and support.” The licences “are portable across hardware platforms”, and the Nutanix end-user licensing agreement “is applicable for all NKP software”.[1] NKP Full Stack licences are described as “available in 1-5-year term options”.[1]
Virtualization & partitioning
NKP has no partitioning rule of its own. Licensing follows where the worker nodes run. On bare metal, the physical cores of the worker node count. On a hypervisor, whether NCI with AHV or vSphere, the vCPUs of the worker node VM count. NKP Starter runs only on NCI with AHV, while Pro and Ultimate also run on vSphere, bare metal and public clouds.[1][2]
NKP Full Stack is the exception. Because it is licensed per physical core of the whole cluster, the number of worker-node vCPUs on that cluster does not change the count.[2]
Cloud / BYOL
NKP Pro and Ultimate can run on AWS, Azure and GCP, and vCPUs are counted there as on any VM.[1] NKP Cloud Kubernetes is the offer for managing Amazon EKS or Azure AKS clusters.[2] NAI Pro supports AWS EKS, Azure AKS and Google Cloud GKE, and the Standalone package runs on any CNCF-certified Kubernetes.[2] Under the LSA, all Nutanix software may be used only on Certified Hardware and “Nutanix authorized public cloud-based service offerings”.[3]
Programs
NKP and NAI have no program of their own. They are bought through the same partner channel and under the same LSA terms (automatic renewal, audit, Usage Capacity) as other Nutanix software. Those terms are described in Nutanix licensing.[3]
Out of scope
- NCI, NCM, NUS and NDB licensed on their own: see Nutanix Cloud Platform licensing.
- The NCI Cloud Native (NKS) add-on, which the cited NKP documents do not cover.
- Third-party AI models and runtimes served through NAI, such as NVIDIA NIM or Hugging Face models, which have their own licence terms.
- List prices: Nutanix does not publish NKP or NAI prices.