Validated Reference Architecture

Cisco AI POD

A plug-and-play AI POD reference architecture with pre-validated servers, network, and GPUs.

An integrated stack pre-validating UCS C885A M8 servers and Nexus 9364E-SG2 switches against the NVIDIA Enterprise Reference Architecture (2-8-9-400). It reduces design risk and shortens deployment time.

What Is Cisco AI POD

An AI POD is a reference architecture (RA) in which servers, network, and GPUs are pre-combined and validated. Instead of validating individual components' compatibility and performance themselves, customers can adopt a pre-configured stack aligned to the NVIDIA Enterprise Reference Architecture (2-8-9-400) as-is. It is a plug-and-play AI infrastructure building block that reduces design risk and deployment time.

Key Capabilities

Pre-validated configuration

Validated combination of UCS C885A M8 + Nexus 9364E-SG2 800G switches

NVIDIA RA compliance

Standard configuration aligned to the Enterprise Reference Architecture (2-8-9-400)

Built-in lossless fabric

RoCEv2 lossless Ethernet AI fabric, pre-configured

Scalable topology

Leaf-spine topology makes it easy to scale

Reference Configuration

ItemConfiguration
ComputeUCS C885A M8 (8-GPU HGX)
NetworkNexus 9364E-SG2 (800G)
ReferenceNVIDIA Enterprise RA 2-8-9-400
FabricRoCEv2 lossless Ethernet
TopologyScalable leaf-spine

What It's Used For

Validated AI infrastructure adoption

Adopt a standard stack quickly, without design-validation burden

Enterprise standardization

Repeat the same validated configuration across multiple sites

Phased expansion

Start with a POD and expand the cluster as needed

Contact Sales