Nutanix Embraces Neocloud: Multitenant Platform for AI and Kubernetes on Bare Metal
Enterprise AI has taken a leap forward. After an initial phase dominated by the training of large models—an activity that is highly computationally intensive and concentrated in the hands of a few operators with almost exclusive access to GPU clusters—we are now in a phase focused on large-scale production execution of agentic applications for a much broader range of enterprise clients. This change in scale requires different platforms on both the provider and customer sides: not just raw computing power, but security, governance, tenant isolation, and cost control per token.
This is the premise at the .NEXT 2026 in Chicago, which Edge9 is attending, where Nutanix presents the extension of its Nutanix Agentic AI solution towards a category of operators known in the industry as neocloud. The term refers to next-generation cloud providers specifically created to offer rapid and flexible access to GPUs through on-demand services. Unlike traditional hyperscalers, neoclouds are built around AI from the ground up: they do not adapt general-purpose infrastructure to AI workloads, but design their infrastructure based on the requirements of GPU, high-bandwidth networking, and high-performance storage typical of these workloads. The transition to inference and agentic applications, however, requires the ability to serve many more companies simultaneously, with requirements and skills clearly marked as enterprise.
A Platform for Neocloud AI Services
The Nutanix Agentic AI solution, announced at NVIDIA GTC 2026 and currently in early access, is a complete software stack designed to build and execute agentic AI applications on the Nutanix Cloud Platform. It integrates high-performance virtualization, computing resources, storage, networking, and Kubernetes services into a single platform, with the declared aim of reducing operational complexity and making token costs more predictable.
The new features announced at .NEXT, expected in the second half of 2026, extend this platform to neocloud providers with a multitenant AI management portal built on the Service Provider Central framework. The portal introduces strong isolation between tenants and granular resource management, allowing providers to host multiple companies on the same physical GPU infrastructure with security and predictable performance for each client. The catalog of services that neoclouds can deliver through the platform includes GPU-as-a-Service, Kubernetes-as-a-Service, VM-as-a-Service, Notebooks-as-a-Service, VectorDB-as-a-Service, and Models-as-a-Service.
Complementing the multitenant features, Nutanix Cloud Manager (NCM) adds usage-based consumption measurement, which allows providers to track and bill based on actual usage of GPUs, API calls, or model consumption. The combination of these features aims to enable neoclouds to manage capacity, monitor tenant usage, and control distributed AI infrastructure from a unified interface. The positioning also includes sovereign AI installations, a relevant aspect for providers serving clients with data residency constraints.
NKP Metal: Kubernetes Without Hypervisor
Alongside the offering for neoclouds, Nutanix announces NKP Metal, an extension of the Nutanix Kubernetes Platform (NKP) that brings Kubernetes directly to bare metal infrastructure, bypassing a hypervisor. The solution targets edge environments and AI training workloads that require direct access to high-density GPU infrastructure, where the overhead of virtualization can represent a performance bottleneck.
Nutanix's approach is what the company defines as a "dual-native" architecture: containers and virtual machines coexist as first-class components under a unified operational model. Unlike solutions that are exclusively hypervisor-based or exclusively Kubernetes-based, NKP Metal allows management of both types of workloads with the same tools and operational criteria.
NKP Metal maintains the same automation, lifecycle management, and data services capabilities available in Nutanix's virtualized environments. Automated node configuration is done through Nutanix Foundation, the operating system and firmware lifecycle management is handled by Lifecycle Manager, and storage options include the standard CSI interface or Cloud Native AOS, a storage solution built for Kubernetes-native bare metal environments, paired with Nutanix Data Services for Kubernetes-native data services.
NKP Metal is currently in early access for customers with NKP PRO and NKP ULT licenses; general availability is expected in the second half of 2026, the same timeline announced for the multitenant and AI usage measurement features for neoclouds.