Nutanix Is No Longer an HCI Company. And That Changes Everything.
Ask an infrastructure architect what Nutanix does, and there is a reasonable chance the answer will involve hyperconverged infrastructure (HCI). Compute and storage combined into a single cluster. A software-defined alternative to traditional three-tier infrastructure. Perhaps even the familiar Nutanix-branded appliance (Supermicro) that helped establish the company more than a decade ago.
None of these descriptions is historically incorrect. But they no longer adequately describe Nutanix.
The company has evolved from an HCI pioneer into a broader software platform provider addressing enterprise virtualization, Kubernetes, databases, and artificial intelligence. Its technology increasingly operates independently of the traditional hyperconverged deployment model, including on supported third-party infrastructure and across different cloud environments.
Enterprises are reconsidering their virtualization strategies following changes in the VMware ecosystem. Kubernetes adoption continues to accelerate, while organizations struggle to manage the operational complexity of containerized applications. And artificial intelligence (AI) is introducing an entirely new set of challenges around infrastructure economics, governance, security, and data sovereignty.
And I think these developments have something in common. Fundamentally, they are not hardware problems. They are platform problems.
And this is where the discussion around Nutanix needs to change.
HCI is an architecture and Nutanix is a software platform
Nutanix built its reputation by challenging traditional infrastructure architectures. At a time when enterprise data centers were dominated by separate compute, storage, and virtualization layers, Nutanix introduced a distributed software architecture that consolidated these capabilities into a more integrated operating model.
The commercial proposition was initially straightforward. Purchase an appliance, deploy a cluster, and simplify infrastructure operations. It worked. Well, it still does very much so!
HCI became an established enterprise architecture, and Nutanix emerged as one of its most recognizable vendors. But the appliance was never the most important part of the technology. The software was. Over time, Nutanix transitioned toward a software subscription business, expanded support for third-party hardware, and introduced additional deployment architectures.
Today, organizations can deploy supported Nutanix solutions on hardware from manufacturers including Dell, HPE, Lenovo, and Cisco. Since 2025, Nutanix also supports architectural approaches that separate compute and storage resources, including selected external storage integrations (Dell PowerFlex, Dell PowerStore, Everypure, NetApp and more soon).
This does not mean that Nutanix software runs on arbitrary hardware or that every storage system is supported. At least not yet, but they are working on that, and a lot has changed over the past 24 months.
An organization no longer needs to approach Nutanix as a decision to purchase a particular appliance. It can evaluate Nutanix as infrastructure software and select an appropriate supported architecture. HCI remains a valuable deployment option, but it is no longer the defining boundary of the product portfolio.
HCI is an architecture and Nutanix is a software platform. Understanding that difference is essential to understanding the company’s direction.
Talking about virtualization is the starting point
The changes in VMware’s strategy under Broadcom have prompted many enterprises to reconsider infrastructure decisions that had remained largely unchanged for years. For Nutanix and many others, this creates an opportunity. Nutanix Cloud Infrastructure (NCI) provides a virtualization and infrastructure platform built around AHV (Acropolis Hypervisor), distributed storage services (Acropolis OS aka AOS), and Prism management.
For organizations seeking an alternative to their existing VMware environment, NCI can represent a practical starting point. Replacing one hypervisor with another is not necessarily a long-term infrastructure strategy, but for some customers that’s already enough.
A virtualization migration may solve an immediate commercial or “partnership” problem without addressing the broader challenges of application modernization, automation, cloud native operations, and AI adoption.

There is also a risk of viewing every customer requirement through the lens of HCI. An enterprise operating a conventional 3-tier environment may have no immediate interest in replacing its storage infrastructure. Another may be more concerned about Kubernetes governance than hypervisor licensing. A third may be struggling to bring AI applications into production while controlling the cost of inference. These organizations have different problems, and the solutions should reflect that.
Nutanix’s expanding portfolio makes it possible to have conversations beyond infrastructure replacement.
Kubernetes has an open source foundation. What about enterprise-grade operations?
Kubernetes (K8s) has become a foundational technology for modern application development. It provides a standardized way to orchestrate containerized applications, scale workloads, and abstract applications from much of the underlying infrastructure. Its success is closely connected to open source. Organizations can build Kubernetes environments using upstream projects and a broad ecosystem of Cloud Native Computing Foundation (CNCF) technologies.
In principle, this offers substantial freedom.
In practice, operating Kubernetes reliably at enterprise scale is considerably more complicated than deploying a cluster. A production environment requires networking, ingress, observability, logging, security policies, identity management, persistent storage, backup, disaster recovery, and lifecycle management.
Each of these requirements introduces additional components, operational responsibilities, and integration work. Many organizations initially address these requirements by assembling their own Kubernetes platform from open source projects. This approach can be technically successful, particularly when experienced platform engineering teams are available.
But it also creates an important question: How much time should an organization spend building and maintaining its Kubernetes platform instead of delivering applications that create business value?
The software components may be freely available, but operating them is not free. Engineering resources, integration testing, security updates, upgrades, troubleshooting, innovation management and ongoing maintenance all contribute to the total cost of ownership.
Open source gives you freedom. Nutanix makes that freedom actually usable.
Nutanix Kubernetes Platform (NKP) approaches this problem by combining upstream Kubernetes and established open source technologies into an integrated, lifecycle-managed enterprise platform. NKP is not an attempt to replace the Kubernetes ecosystem with a proprietary alternative. Its architecture is based on upstream Kubernetes and largely on established CNCF technologies, allowing organizations to benefit from open standards and the broader open source community.

The commercial value is not primarily in access to Kubernetes itself. It is in the engineering, integration, validation, lifecycle management, enterprise support, and operational capabilities surrounding it. Rather than independently selecting, integrating, and maintaining every component required for production Kubernetes, organizations can adopt a platform in which much of that work has already been performed.
Depending on the NKP edition, this includes capabilities such as centralized cluster management, observability, security integration, backup, and management across multiple environments.
NKP can also operate beyond Nutanix Cloud Infrastructure, including supported deployments on other virtualization platforms, baremetal infrastructure, and public clouds.
This makes it relevant even to organizations that have no immediate plans to replace their existing virtualization infrastructure. And that is perhaps one of the clearest examples of why Nutanix can no longer be understood exclusively as an HCI company.
The economic value of an enterprise Kubernetes platform
The financial argument for enterprise Kubernetes is often misunderstood. An opensource Kubernetes platform may appear less expensive because its underlying components do not require commercial license fees. But software licensing is only one part of the cost.
Consider an enterprise operating dozens of Kubernetes clusters across development, testing, and production environments. Its platform engineering team must maintain cluster configurations, implement security controls, coordinate upgrades, manage monitoring, and ensure that individual application teams follow consistent operational standards.
Without centralized governance, clusters may evolve differently over time. Some teams adopt new components independently and others postpone upgrades. Security policies become inconsistent, and operational knowledge remains concentrated among a small number of specialists.
The result is not necessarily a failure of Kubernetes. It is a failure to industrialize its operation.
An integrated enterprise platform can help reduce this fragmentation through standardized deployments, centralized lifecycle management, and consistent policies. The potential financial benefit comes from reduced operational effort, fewer duplicated tools, more efficient resource utilization, and faster delivery of application environments.
Whether NKP actually produces a lower TCO depends on the existing environment, licensing, staffing, and operational maturity. A well-managed open source platform with a highly experienced internal team may already be economically efficient. But for organizations spending significant resources assembling and maintaining their own Kubernetes distribution, the economics deserve serious examination.
The objective should be to minimize the cost and complexity of delivering reliable Kubernetes services.
Enterprise AI introduces a new infrastructure economy
Public AI services provide rapid access to capable models without requiring organizations to operate their own GPU infrastructure. But enterprises need to understand where sensitive information is processed, how models are accessed, who can use them, and how much each application consumes.
They also need to decide whether particular workloads should use public AI services, privately hosted models, or a combination of both. This introduces an economic model that is unfamiliar to many traditional infrastructure teams.
Instead of primarily calculating costs in virtual machines, CPU cores, memory, and storage capacity, organizations increasingly need to understand the economics of AI tokens.
Tokenomics
Large language models (LLM) process information in tokens. Tokens represent units of text or other model input and output, and they are commonly used to measure consumption when AI services are billed through APIs. A simple request may consume relatively few tokens and a complex AI application can consume substantially more.
Consider an AI assistant that retrieves internal documents, processes large amounts of context, performs several reasoning steps, and generates a detailed response. Now consider an AI agent that performs similar operations repeatedly, calls external tools, and interacts with other agents. The number of model requests and processed tokens can grow rapidly.
This is where token economics, often referred to as tokenomics in the enterprise AI context, becomes relevant. Organizations need visibility into how many tokens are consumed, which applications generate that consumption, which models are being used, and what those interactions cost.
Without centralized oversight, individual development teams may integrate different AI providers independently. One application uses a premium commercial model and another uses an open-weight model hosted in the cloud. A third runs a private model on internal GPU infrastructure.
Each approach may be technically justified. But without a consistent governance model, the organization can lose visibility into its overall AI consumption.
Nutanix Enterprise AI – Making AI consumption governable
Nutanix Enterprise AI (NAI) addresses this challenge by providing software capabilities for deploying, managing, and governing enterprise AI inference. Rather than requiring every application team to independently integrate models, configure inference infrastructure, and implement its own access controls, NAI introduces a common operational layer.

One of its most relevant capabilities is the Nutanix Agent Gateway, which includes AI Gateway functionality. The principle is relatively straightforward. Instead of allowing every application to communicate directly with individual model providers, organizations can provide a centrally governed access point. Through this gateway, applications can access supported privately hosted models and external AI providers using a common interface.
Authentication, observability, and token-based rate limiting can be managed centrally. This creates a foundation for controlling AI consumption across different applications and teams.
Importantly, the gateway does not require every model to run on Nutanix infrastructure. Its purpose is to provide a common control layer for supported AI services, regardless of whether the underlying model is privately hosted or accessed through an external provider.
The Nutanix Agent Gateway
To understand the value of an agent gateway, consider how organizations manage traditional network traffic. Applications do not generally connect to every external system without security controls. Enterprises use gateways, identity services, firewalls, proxies, and monitoring systems to enforce policies and maintain visibility.
AI introduces similar requirements, but with additional considerations. Organizations need to govern not only access to models, but also consumption, model selection, and increasingly the actions performed by AI agents.
Nutanix’s Agent Gateway provides a common inference access layer through which organizations can implement supported authentication, observability, and token-based controls.
For example, an enterprise could provide different AI services to different departments:
- A development team might be authorized to use a high-performance external model for certain workloads.
- An internal document processing application might be required to use a privately hosted model.
- Another application might operate under stricter consumption limits because its workload is highly repetitive.
Centralized token monitoring helps administrators understand consumption patterns, while rate limits can prevent individual users or applications from consuming resources without restriction. This does not automatically make every AI interaction compliant or secure. Data classification, application authorization, model behavior, and provider agreements remain separate responsibilities. But it provides an important operational capability that is often missing when AI adoption grows organically.

Recent NAI capabilities also extend governance toward agentic AI through Model Context Protocol (MCP) access management. MCP enables AI agents to interact with external tools and enterprise data sources through standardized interfaces. That introduces additional security requirements because an agent may do more than generate text. It may retrieve information, invoke business systems, or initiate actions. A governed gateway can help organizations control and audit those interactions.

Private Inference & Sovereignty
NAI provides capabilities for deploying and operating private model inference, including supported GPU-based configurations. The objective is not necessarily to replace every external AI service. In many cases, a hybrid approach makes more sense. Less sensitive workloads may continue using external models, while applications involving confidential information can use privately hosted inference services.
The AI Gateway can provide a common access layer across these different environments. This creates architectural flexibility without requiring application developers to manage every inference platform independently.
However, private inference should not be confused with an automatic guarantee of confidentiality or sovereignty. The complete solution must still account for logging, model supply chains, access controls, data retention, encryption, and the location of supporting services.
Kubernetes and AI are closely connected
The relationship between NKP and NAI is important (but not mandatory). Modern AI applications increasingly depend on containerized infrastructure. Inference services, model-serving components, application APIs, data processing services, and supporting tools can all benefit from Kubernetes-based deployment and orchestration. But operating these components reliably requires more than basic container scheduling. GPU resource management, scaling, observability, security, storage, and application lifecycle management become important operational considerations.
- NKP provides a Kubernetes platform built around open source technologies and enterprise lifecycle management.
- NAI adds AI-specific capabilities, including model serving, inference access, and governance.
Together, they can form complementary layers of an enterprise AI architecture. The two products are not inseparable.
Please note: NAI can also operate on supported CNCF-certified Kubernetes environments outside NKP.
Organizations should be able to select the infrastructure and Kubernetes environment that best meets their requirements. The broader opportunity is to avoid creating another isolated technology stack solely for AI. Instead, organizations can extend established platform engineering practices into AI operations. This can reduce duplicated operational processes and help development teams consume AI capabilities through standardized internal services.
Databases are part of the platform story
While Kubernetes and AI receive much of the attention, databases remain central to enterprise IT. Database operations can involve substantial manual effort, particularly when organizations maintain multiple database technologies across different environments. Provisioning, patching, cloning, backup, and lifecycle management are often handled through a combination of scripts, specialized tools, and operational processes.
Nutanix Database Service (NDB) addresses parts of this challenge by providing database lifecycle management and automation for supported database engines and configurations. The objective is similar to that of an enterprise Kubernetes platform. Reduce repetitive operational work, standardize service delivery, and provide internal consumers with faster access to managed services.
NDB is not a universal replacement for every database management tool, and its suitability depends on supported technologies and operational requirements.
But it illustrates another aspect of Nutanix’s broader direction. The company is extending its software beyond infrastructure deployment toward the operational services consumed by applications and development teams.
Conclusion
Nutanix will always have an important place in the history of hyperconverged infrastructure. The company helped change how enterprises deploy and operate infrastructure, challenging the traditional separation of compute and storage. But its evolution has continued well beyond that original market.
Today, as you can see, Nutanix addresses virtualization, infrastructure management, Kubernetes, database operations, and enterprise AI through a growing software portfolio. Together, these solutions represent a shift from selling infrastructure architectures toward providing platforms that support modern applications and digital services.
The financial benefits will vary by organization, and no single platform eliminates the complexity of enterprise IT. But the direction is clear.
Nutanix may have started by changing how infrastructure was built. Its future depends on how effectively it helps organizations build, operate, and govern the applications and AI services running on top of it.
And that is why describing Nutanix simply as an HCI company no longer tells the full story.






































