Nutanix expands AI ecosystem with NVIDIA, Supermicro and Netris
Nutanix has unveiled a trio of ecosystem partnerships at NVIDIA GTC Berlin, collaborating with NVIDIA, Supermicro, and networking automation vendor Netris to simplify the path from AI investment to production workloads. The announcements centre on agentic AI infrastructure, private inference cost control, and multitenant GPU-as-a-service delivery for neoclouds and enterprise service providers.
The company framed all three moves as responses to a common customer pressure: organisations are accumulating GPU capacity but struggling to operationalise it quickly. Nutanix's stated goal is to provide a consistent cloud operating model across bare-metal, virtualised, and containerised environments.
The three deals
The most technically specific announcement involves NVIDIA's Vera CPU. Nutanix said it plans to work towards Day Zero software support for Vera on its Nutanix Kubernetes Platform (NKP) in bare-metal configurations, with virtualised environments to follow. The rationale is that agentic AI workloads combine inference with tool calls, code execution, and retrieval steps that sit between GPU operations; Nutanix argues that CPU performance has become a bottleneck in AI factory pipelines, not just the accelerator layer. NVIDIA's Hannah Coutand, Director of Product Marketing for Vera CPU, said the collaboration is aimed at giving enterprises a consistent cloud operating model for high-performance agent execution.
The second partnership involves Supermicro. Nutanix and Supermicro are co-engineering a validated server system that combines Supermicro hardware with NVIDIA GPUs, NVIDIA Spectrum-X Ethernet networking, and Nutanix's Agentic AI (NAI) software stack. The intent is to reduce multivendor integration complexity and bring down per-token inference costs for private, on-premises deployments. No pricing, benchmark figures, or delivery timeline was disclosed in the release.
The third announcement pairs Nutanix with Netris, whose Network Automation, Abstraction, and Multitenancy (NAAM) platform will handle physical network-layer multitenancy for AI cloud providers. The integrated solution targets neocloud operators who need to onboard tenants rapidly, enforce hard isolation at the hardware level, and provision GPU capacity as a self-service cloud service. Nutanix will supply compute, storage, Kubernetes, and its Agent Gateway and model-as-a-service layer; Netris provides the network fabric automation.
Thomas Cornely, Executive Vice President of Product Management at Nutanix, said organisations need "an AI cloud that simply connects intelligence, agents, enterprise applications, and data across CPUs, GPUs, virtual machines, containers, on-premises environments, and cloud."
Market context
Nutanix built its business on hyperconverged infrastructure and has been steadily repositioning as an AI platform company as the HCI market matures. The three announcements follow a familiar playbook: validated reference architectures with tier-one hardware partners reduce customer integration risk and give Nutanix a route to new workloads without requiring changes to its software licensing model.
The private inference market is increasingly competitive. Dell, HPE, and Lenovo all offer NVIDIA-validated AI factory configurations, and hyperscalers continue to extend managed inference services with on-premises extensions such as AWS Outposts and Google Distributed Cloud. Nutanix's differentiator is its unified operating model across hybrid environments rather than any single hardware specification, but that positioning is not unique: VMware by Broadcom and Red Hat make closely related claims.
The Vera CPU collaboration is worth watching separately. NVIDIA has historically dominated the accelerator layer; its move into server CPUs with Vera signals an ambition to own more of the AI factory stack. Day Zero software support from Nutanix would give early Vera adopters a validated Kubernetes substrate, which could be a meaningful sales lever as the chip ramps to general availability.
Most of the announcements are stated as plans and intentions rather than shipping products. Buyers should treat them as roadmap signals. The next near-term milestone will be customer wins and published benchmark data for the Supermicro jointly engineered system, which will be the clearest test of whether the cost-control claims hold in production.