Nutanix partners with NVIDIA and Supermicro for AI agent infrastructure

The rise of agentic AI

As organizations move beyond basic generative AI, they are increasingly looking toward 'agentic AI.' Unlike standard models that simply generate text, agents are designed to perform actions. They combine inference—the process of a model generating an output—with tool calls, code execution, and data retrieval.

Because these agents must perform many tasks between model calls, the hardware they run on requires more than just high-end graphics processing units (GPUs). Nutanix argues that CPU performance is becoming a critical factor in these workflows. To address this, the company is forming a series of strategic partnerships to create a unified infrastructure for these complex workloads.

Hardware and networking synergy

A major component of this strategy involves a collaboration with NVIDIA and Supermicro. Nutanix plans to work toward 'Day Zero' support for NVIDIA's Vera CPUs. This support will begin with the Nutanix Kubernetes Platform (NKP) on bare metal, which is software running directly on physical hardware without a middle layer, before expanding to virtualized environments.

To create private runtime environments for these agents, Nutanix will pair the Vera CPUs with NVIDIA OpenShell. Additionally, Nutanix and Supermicro are planning a jointly engineered system. This system will combine Supermicro servers and NVIDIA GPUs with Nutanix Agentic AI (NAI) and NVIDIA Spectrum-X Ethernet networking. The goal is to reduce the time spent on custom configurations and integration while lowering token costs for companies running inference on-premises.

Networking and cloud control

The infrastructure plan also extends to how data moves across networks. Nutanix Cloud Platform will be integrated with the network automation and multitenancy platform from Netris. In this arrangement, Netris will manage the physical network fabric, which is the underlying hardware that connects all devices. They will be responsible for enforcing tenant isolation on that hardware.

Nutanix will provide the overarching AI cloud control plane and the necessary AI storage. This specific combination of services is designed for AI cloud providers who rent out GPU infrastructure to other clients. By providing a unified platform, the partners hope to give these providers a way to offer built-in security, performance, and governance to their customers.

Overcoming deployment hurdles

Currently, many organizations face pressure to turn AI investments into actual business value. However, Nutanix notes that capacity constraints and the sheer complexity of managing the required infrastructure can slow down progress. The stated aim of these new partnerships is to cut these costs and speed up the time it takes for a company to see results from their AI deployments.

While the announcements are significant, Nutanix included a disclaimer that some features will be offered on a 'when-and-if-available' basis. The company noted that these are plans and do not commit to specific delivery dates or functionality. As Analytics India Magazine reported, the information should not be used as a basis for immediate purchasing decisions, as the projects remain subject to ongoing development and commercial arrangements.

The role of the CPU

The technical shift here centers on the balance between the GPU and the CPU. While GPUs are the primary engines for training and running large models, the CPU acts as the coordinator. In an agentic workflow, the CPU must handle the logic of deciding which tool to use, executing the resulting code, and processing the data returned by those tools.

By focusing on Vera CPUs, Nutanix is positioning itself to support the 'connective tissue' of AI. If the CPU is too slow, the agent will experience latency between every action it takes. By optimizing this layer, Nutanix hopes to make the transition from experimental AI to production-ready agents more seamless for enterprise users who need reliable, high-speed execution.