Cisco is expanding its Secure AI Factory with NVIDIA to include Supermicro servers. This gives companies access to liquid- and air-cooled rack-scale systems, validated and sold as part of the Cisco portfolio. The first shipments are scheduled for October 2026.
These are high-density GPU systems. They let users manage AI clusters and non-AI workloads in the same environment. Another new feature is rack-to-fabric liquid cooling. To achieve this, Cisco combines its liquid-cooled AI networking systems with Supermicro’s liquid-cooled servers. The solution is designed to train models with one trillion parameters and support high-throughput inference on platforms such as the Nvidia Vera Rubin NVL72 and HGX Rubin NVL8.
For neocloud and sovereign cloud customers, Cisco delivers Nvidia Cloud Partner-compliant solutions. Switches based on Cisco Silicon One handle the front end, while Spectrum-X-based switches handle the back end. Cisco Nexus One ties both layers together. According to Cisco, it is the only Nvidia technology partner to deploy its own switches and its own network operating system in an NCP-compliant solution.
Certification and supply chain
Cisco is introducing Cisco Validated Infrastructure Services (CVIS), tailored to the Nvidia Infrastructure Services offering. This service certifies that infrastructure has actually been built according to the reference architecture. To develop the associated tools and test software, Cisco is setting up its own large-scale AI Lab.
Both companies highlight their supply chain capabilities in a market where access to GPUs and memory remains limited. For day-to-day operations, companies rely on NVIDIA AI Enterprise and AgenticOps via Cisco Cloud Control, correlating job health with the performance of compute, NICs, optics, and the network.
“We are at the beginning of one of the largest datacenter buildouts in history,” says Jeetu Patel, President and Chief Product Officer at Cisco. “Every organization is racing to scale AI — but speed only counts if it comes with control of data, managed token costs, and real ROI.”
Tip: Cisco and NVIDIA lower the barrier to secure, full-stack AI infrastructure