Nutanix has launched new capabilities designed to help enterprises and service providers operate AI agents across hybrid cloud environments, the company announced Wednesday. The San Jose, Calif.-based firm unveiled Nutanix Enterprise AI 2.8, which provides centralized oversight for AI inference and agentic AI, alongside Nutanix Kubernetes Platform 2.19, designed to simplify container management for bare-metal and virtualized setups. The releases mark Nutanix's effort to combine infrastructure, compute, and intelligence layers into a single platform for building, running, and securing AI agents that operate on-premises or in public clouds.
Nutanix Enterprise AI 2.8 delivers a centralized agent gateway that enables customers to monitor token consumption, track which models are in use, and see spending in real time, according to Thomas Cornely, the company's executive vice president of product management. IT teams can set access policies, cap usage levels, and route workloads to appropriate models depending on cost and performance requirements. The release also includes private inferencing and a centralized Model Context Protocol layer to govern how agents access applications, aiming to prevent fragmented MCP configurations and provide a single management point for developers and AI builders. Nutanix Kubernetes Platform 2.19 adds Cloud Native Computing Foundation AI conformance, GPU optimization, and an AI catalog with open-source components plus Nutanix's AI Gateway, private inferencing, and MCP capabilities. The platform's "dual-native" architecture can run on Nutanix's AHV hypervisor, bare metal, or public clouds, letting customers test AI services in the cloud or on existing on-premises infrastructure before moving to production. Service Provider Central extends the management plane with service provider-governed multitenancy across virtual machines, data, networks, and containers, helping managed service providers modernize VMware-based environments while adding AI services.
Cornely said Nutanix is merging the core elements enterprises need to build, run, secure, and govern agentic AI across hybrid environments, giving IT a consistent method to control how agents access cloud models and enterprise applications and data on-premises or in public clouds. "Very few companies, and I would argue actually no other companies, have a platform that allows you to support all of this and give you a consistent way to govern, monitor, operate and just build these end-to-end solutions," Cornely told CRN. He added that not all tasks should consume tokens from the most expensive, highest-performance model, recommending that organizations use high-end models for advanced requests and open-weight models for more common requests. Anthony Jackman, chief innovation officer at Pittsburgh-based solution provider Expedient, said the company's entire AI product line is cloud-native by design, running on Kubernetes, with NKP as the default platform, and that Expedient has been collaborating with Nutanix on multitenancy for several years as the first customer to use it.
The enhancements arrive as agents run on CPUs and GPUs while containers operate on both legacy and virtual machine infrastructure, creating a need for platforms that can support all three tiers with consistent governance, monitoring, and operational capabilities, according to the release. The work builds on 16 years of Nutanix platform development and applies it to an AI model in which agents run on containers, consume applications on virtual machines, and connect to GPU-based intelligence in the cloud or on-premises. Cornely noted that customers are already seeing token consumption and costs climb as AI use expands, making visibility and controls critical. Jackman emphasized that NAI 2.8 and its new MCP gateway is very important because AI without controlled access to applications and data doesn't deliver value to enterprises, and centralizing control makes it easier for clients to adopt AI securely. The releases position Nutanix to serve both direct enterprise customers and service providers, including neoclouds that currently serve a small number of large tenants but will need more agile multitenancy as they target enterprise customers. For channel partners and managed service providers, the combination of easier operation at scale and greater consistency across clients translates to lower pricing and improved customer experiences.