Techzine reported on Aug. 25 that Cisco will add Supermicro servers to the "Secure AI Factory" it operates with Nvidia.

The first shipment is scheduled for October.

The report said the added system uses a high-density GPU configuration. Companies can run AI clusters and non-AI workloads together in the same environment. Cisco also introduced rack-and-fabric liquid cooling. It combines Cisco's liquid-cooled AI networking system with Supermicro liquid-cooled servers. The solution supports training models with 1 trillion parameters and can also handle high-throughput inference on platforms such as Nvidia Vera Rubin NVL72 and HGX Rubin NVL8.

Cisco provides Nvidia Cloud Partner (NCP)-certified solutions to neocloud and sovereign cloud customers. Cisco Silicon One-based switches handle the front end, and Spectrum-X-based switches handle the back end. Cisco Nexus One connects the two layers.

Cisco said it is the only Nvidia technology partner that applies its own switches and network operating system together to NCP-certified solutions.

Cisco is also introducing Cisco Validated Infrastructure Services (CVIS) tailored to Nvidia Infrastructure Services.

Jeetu Patel (지투 파텔), Cisco president and chief product officer, said, "This is the moment when the largest data center buildout in history is starting." He said, "Every company is joining the race to scale AI, but speed only matters when it is backed by data control, management of token costs, and tangible return on investment."