Cisco Adds Supermicro Rack-Scale Systems to Its NVIDIA AI Factory for October Rollout

Cisco will begin selling Supermicro rack-scale AI compute through its Secure AI Factory with NVIDIA in October, giving enterprise, neocloud and sovereign-cloud customers a single portfolio that spans dense GPU servers, networking and cooling. The move brings the compute layer more directly into Cisco’s AI infrastructure offer as customers confront the power and cooling demands of larger clusters.

Cisco announced the partnership with Supermicro on August 25. Its expanded portfolio will include Supermicro liquid- and air-cooled rack-scale systems and dense GPU systems, validated and sold as part of Cisco’s AI infrastructure portfolio. Cisco says the package is intended to support deployments ranging from trillion-parameter model training to high-throughput inference.

A rack-to-fabric design

The key technical addition is not simply a new server supplier. Cisco says customers will be able to combine its liquid-cooled AI networking systems with Supermicro liquid-cooled servers in a rack-to-fabric cooling design. The infrastructure is set to support NVIDIA Vera Rubin NVL72 and NVIDIA HGX Rubin NVL8 platforms, pairing Cisco AI networking with Supermicro servers.

Validation is part of the product

Cisco says the architecture will provide NVIDIA Cloud Partner-compliant solutions for neocloud and sovereign-cloud customers. Its network design uses Cisco Silicon One-based switches at the front end and NVIDIA Spectrum-X-based switches at the back end, unified through Cisco Nexus One. That makes the partnership a full-stack reference architecture, rather than a standalone server resale arrangement.

What Cisco says customers receive

• Cisco Validated Infrastructure Services, aligned with NVIDIA Infrastructure Services, to certify that infrastructure is built as designed and aligned with reference architectures.
• A dedicated large-scale AI Lab that Cisco says it is funding to develop tools and test software for those validation services.
• Operational tooling through NVIDIA AI Enterprise software and AgenticOps in Cisco Cloud Control, including the ability to correlate job health with compute, network-interface, optics and network-performance metrics.

The immediate test is execution in October: Cisco will then need to turn its validated design and support model into available systems for the customers now planning rack-scale AI deployments. For those buyers, the practical change is a route to procure compute, networking and cooling through the same Cisco AI infrastructure portfolio.