Nutanix: AI Platform, Bare-Metal Kubernetes, and Neocloud Services

Contents

In addition to strategic partnerships with NetApp and Cisco, Nutanix announced a series of products at its .NEXT conference in Chicago targeting AI infrastructure, Kubernetes extensions, and new offerings for specialized cloud providers. However, a large portion of the new features is only scheduled for the second half of 2026.

With its Agentic AI platform, Nutanix is addressing the next stage of AI system development: applications that not only execute models but also independently make decisions and control processes. Technically, the provider combines infrastructure, platform services, and management functions in a stack where virtual machines, containers, and GPU-accelerated AI workloads can be run in parallel.

The complete platform is only announced for the second half of 2026, remaining a roadmap item for now. However, individual components are already available: Nutanix Unified Storage 5.3 brings Smart Tiering to public clouds and multi-tenant object scaling for large data volumes.

Also new is support for RDMA (Remote Direct Memory Access) in Unified Storage, which is intended to significantly reduce latency in data-intensive AI workflows. Data Lens 2.0 can now be operated entirely in local and air-gapped environments for the first time – relevant for customers in regulated industries who require ransomware detection and data governance without cloud connectivity.

The new Nutanix Cloud Manager 2.0 (NCM) provides multi-site and multi-domain management via a central control plane. Cost control, AIOps functions, and self-service provisioning are integrated. Nutanix emphasizes that cost analysis runs entirely on-premises – a differentiating feature compared to cloud-based FinOps tools, which should be particularly relevant for security-critical environments. In parallel, Nutanix is expanding support for sovereign cloud scenarios, for example with NC2 on AWS GovCloud (available immediately) and the planned NC2 support for the already available AWS European Sovereign Cloud later this year.

Kubernetes without Hypervisor: NKP Metal and Dual-Native Architecture

With NKP Metal, Nutanix is extending its Kubernetes platform to directly utilize physical servers. The goal: containers and virtual machines should run under a unified operating model, without an additional virtualization layer. Nutanix calls this architecture “Dual Native” – organizations can run Kubernetes workloads either on VMs or directly on bare metal, with unified management and security models.

The approach primarily targets scenarios with high performance requirements: AI training with dense GPU configurations, inference workloads, and edge deployments. In these use cases, an additional hypervisor layer creates measurable overhead – both in GPU passthrough and network latency. Bare metal eliminates this layer but typically brings higher complexity in provisioning, firmware updates, and lifecycle management.

Nutanix aims to close this gap by transferring automation, lifecycle management, and data services from its own platform to the physical infrastructure. NKP Metal is currently in early access with selected partners; general availability is planned for the second half of 2026.

New Target Group: Neocloud Providers

Nutanix is addressing specialized cloud providers focusing on AI infrastructure more strongly than before. These so-called Neoclouds – providers that primarily offer GPU-based resources for AI training and inference – are increasingly developing into platform operators who want to offer not just pure computing capacity but also Kubernetes-as-a-Service and managed AI services.

For this, they primarily need multi-tenancy, billing models, and isolation between customer environments. Nutanix provides a multi-tenancy framework for this with granular control of resources, security policies, and network rules per tenant. This is complemented by usage-based billing – for example, based on GPU consumption, token usage for LLM inference, or API calls. Token-based cost control, in particular, addresses a specific problem: companies running LLM workloads are often surprised by unexpectedly high inference costs.

Also aimed at service providers is the new Service Provider Central: a multi-tenant IaaS product based on the Nutanix Cloud Platform, which enables central management of multiple tenants with automated workflows via Nutanix Central. Each tenant receives its own private cloud environment with access to Nutanix Prism and can independently manage compute, storage, network, and identity. Nutanix explicitly aims to attract former VMware Cloud Service Providers looking for new platform options after Broadcom's restructuring. The offering will be available in the second half of 2026.

Ecosystem and Database Integration

In addition to strategic alliances with NetApp and Cisco, Nutanix is expanding its hardware support with a new Foundation Central Appliance. This simplifies the initial installation of Nutanix Cloud Infrastructure and AHV on servers from Cisco, Dell, Fujitsu, HPE, and Lenovo, as well as its own NX platform. Dell PowerStore will be generally available as an external storage option, complemented by extended Dell private cloud automation.

On Google Cloud, Hyperdisk and C3 bare-metal instances will be added for Nutanix Cloud Clusters (NC2) in the second half of the year – allowing storage to be scaled independently of compute in the public cloud. This aligns with the disaggregation approach from the NetApp partnership.

In the database sector, Nutanix is integrating MongoDB more closely. The combination of Nutanix Database Service and MongoDB Ops Manager is intended to automate deployment, operation, and recovery. For example, distributed database clusters can be deployed faster, and point-in-time recovery for individual clusters can be performed.

Context: Lots of Roadmap, Little Available

.NEXT 2026 shows how Nutanix aims to transform from an HCI provider to a universal cloud platform for VMs, containers, and AI workloads. However, a large part of the announcements is still in the future: the Agentic AI platform, NKP Metal, the Neocloud offering, Service Provider Central, and RDMA support are only scheduled for the second half of 2026. Immediately available, on the other hand, are Unified Storage 5.3, Data Lens 2.0, and Cloud Manager 2.0 – solid platform updates, but not revolutionary.

The strategic question is whether Nutanix can deliver the roadmap on time before the post-VMware market consolidates. Bare-metal Kubernetes and agentic AI infrastructure are areas where Red Hat, VMware by Broadcom itself, and the hyperscalers are also active. The practical benefits will only become apparent when the announced components can actually be used in everyday operations.

(vbr)