Nutanix has announced collaborations with NVIDIA, Supermicro, and Netris to help enterprises and service providers deploy and scale production artificial intelligence (AI) workloads. The initiatives focus on reducing infrastructure deployment time, controlling AI operating costs, and simplifying the management of increasingly demanding AI applications.
As businesses move AI projects from experimentation into production, infrastructure requirements are expanding beyond graphics processing units (GPUs). AI agents must execute code, retrieve information, interact with enterprise applications, and process results between model calls. These tasks place additional demands on central processing units (CPUs), storage, networking, and security.
Nutanix plans to combine its infrastructure software with partner technologies to address these requirements across bare-metal servers and virtualised environments.
“Organisations need more than just advanced infrastructure to move agentic AI into production,” said Thomas Cornely, executive vice president of product management at Nutanix. He said the collaborations aim to help customers deploy AI agents and services through a unified platform with security, performance, and governance capabilities.
NVIDIA Vera support targets agentic AI workloads
Nutanix plans to work toward Day Zero support for NVIDIA Vera CPUs, beginning with the Nutanix Kubernetes Platform (NKP) on bare-metal infrastructure. Support is expected to expand to virtualised Nutanix environments.
The collaboration addresses the growing role of CPUs in agentic AI systems. While GPUs handle many computationally intensive AI operations, CPUs support tasks such as code execution, tool calls, data retrieval, and workflow orchestration.
Nutanix also plans to combine its Kubernetes platform with NVIDIA Vera and NVIDIA OpenShell, which provide runtime environments for AI agents. The companies expect this approach to provide a consistent operating model for deploying agentic applications.
Hannah Coutand, director of product marketing for Vera CPU at NVIDIA, said the processor was developed for workloads involving code execution, tool use, and data processing between model calls. She said the collaboration would help enterprises bring agentic AI applications into production.
The planned support remains subject to development and availability.
Nutanix and Supermicro plan private AI inference systems
Nutanix and Supermicro are planning a jointly engineered infrastructure system for private AI inference. The proposed configuration combines Supermicro servers, NVIDIA GPUs, Nutanix Agentic AI software, and NVIDIA Spectrum-X Ethernet networking.
The companies aim to reduce the time and effort required to configure and integrate infrastructure for on-premises AI deployments. Private inference can also give organisations greater control over where AI workloads and data are processed.
The proposed system is expected to scale GPU capacity as demand grows. Nutanix will provide infrastructure management and AI storage intended to support high-throughput data workflows and resilient operations.
The companies say the approach could help customers avoid assembling and validating multiple infrastructure components independently. However, the announcement does not provide pricing, performance benchmarks, or a release date for the jointly engineered system.
Netris collaboration focuses on multitenant AI clouds
The third initiative targets AI cloud providers and neocloud operators that offer GPU infrastructure to multiple customers.
Nutanix and Netris plan to combine the Nutanix Cloud Platform with Netris Network Automation, Abstraction, and Multitenancy (NAAM). The integrated approach is intended to automate network operations and maintain tenant isolation across physical networks, bare-metal servers, and virtualised infrastructure.
Nutanix will provide compute, storage, security, Kubernetes, and AI platform services. Netris will handle network automation and isolation across the physical network fabric.
The companies aim to let providers onboard customers and allocate infrastructure through a self-service model similar to public cloud services. Hardware-enforced network isolation is intended to keep tenants separated as workloads and GPU allocations change.
Need for a common AI infrastructure model
Don Gentile, an analyst at HyperFRAME Research, said Nutanix was extending its infrastructure expertise into bare-metal AI systems, agent execution, private inference, and multitenant cloud services. He highlighted consistent management, automation, and infrastructure choice as important requirements for enterprises and service providers.
In Asia Pacific and Japan (APJ), Nutanix executives also pointed to the need to balance AI adoption with financial discipline and data control.
Faiz Shakir, vice president of sales for APJ at Nutanix, said businesses need measurable returns from AI while managing costs and governance risks. Helder Queiros, senior director of original equipment manufacturer (OEM) and alliances for APJ at Nutanix, said the collaborations were intended to reduce integration work and shorten deployment cycles.
The Ministry of Environment, Water and Agriculture of Saudi Arabia also highlighted its interest in exploring how Nutanix’s hybrid cloud platform could support public-sector AI adoption and the development of national cloud and AI capabilities.
Production AI remains the central objective
Nutanix’s three collaborations address different parts of the AI infrastructure stack: CPU performance for agentic workloads, integrated systems for private inference, and network automation for multitenant services.
The initiatives reflect the growing need to manage AI workloads as operational systems rather than isolated experiments. Their eventual impact will depend on implementation, availability, performance, and the ability to deliver the expected cost and deployment benefits.
Nutanix said the offerings remain at varying stages of development. Some capabilities and integrations are subject to change, and the company has not committed to delivering every planned feature on a fixed schedule.