Nutanix has announced new AI infrastructure initiatives with NVIDIA, Supermicro and Netris, targeting the high costs and technical challenges companies face as they move AI applications into production.

The initiatives focus on three areas: lowering the cost of AI inference, improving the CPU performance needed for agentic AI, and simplifying the management of AI infrastructure shared by multiple customers.

The announcements expand Nutanix’s strategy of providing a unified software platform for enterprise AI deployments across GPUs, CPUs, VMs and containers. The company is positioning its platform as a way to reduce the complexity of assembling AI infrastructure from multiple vendors, a growing concern among enterprise managers.

A core element of the strategy is a jointly engineered system with Supermicro built to support private AI inference, which allows enterprises to run AI models on infrastructure they control rather than relying exclusively on public cloud services.

The system will use Supermicro servers equipped with eight NVIDIA B300 Blackwell Ultra GPUs per node, supported by NVIDIA Spectrum-X Ethernet networking. Nutanix will provide its Cloud Platform, Kubernetes Platform and Enterprise AI software, enabling companies to manage workloads from early infrastructure provisioning through model deployment.

To boost reliability, the system includes a dedicated NX-8170-G10 management cluster and a separate NX-8150-G10 storage cluster. The storage system will handle demanding AI tasks like data ingestion and model serving.

The networking infrastructure includes NVIDIA Spectrum-4 Ethernet switches, ConnectX-7 and ConnectX-8 SuperNICs, and BlueField-3 DPUs. This mix of gear is designed to support high-speed communication among GPUs while lowering networking overhead.

Nutanix touts the solution as a simpler path to deploying private AI applications without the expense and delays of building customized infrastructure.

However, neither company has released pricing or performance benchmarks demonstrating the system’s cost-per-token savings. The offering is scheduled to become available in early 2027 through Supermicro and its channel partners.

NVIDIA and Netris Collaborations

Beyond GPU infrastructure, Nutanix is collaborating with NVIDIA to support its Vera CPUs, which are geared to handle the data retrieval and workflow orchestration tasks that surround AI model inference. They are essential for agentic AI applications, which perform multistep workflows with limited human intervention.

Nutanix plans to offer Day Zero Vera support initially through its Kubernetes Platform running on bare metal, followed by support for virtualized environments. The companies also plan to use NVIDIA OpenShell to provide secure execution environments for AI agents. No release date or performance benchmarks have been announced for the Vera integration.

Nutanix’s third initiative, developed with Netris, targets cloud providers operating GPU infrastructure for multiple customers.

The collaboration combines Nutanix’s compute and storage management software with Netris network automation technology. It will allow providers to assign networking resources to individual customers while maintaining isolation between their workloads. The system supports both bare-metal and virtualized environments, with networking capabilities that include virtual private clouds and automated configuration.

A technology preview of the combined platform is planned for early 2027, although general availability remains unannounced.