Key Takeaways:
Cisco is betting that rack-scale AI computing, delivered as a validated full-stack package with Supermicro hardware and NVIDIA silicon, will let it win a share of the largest datacenter buildout in history.
Cisco is adding Supermicro's rack-scale, liquid- and air-cooled GPU systems to its Secure AI Factory with NVIDIA, an NCP-compliant architecture aimed at trillion-parameter training for enterprises, neoclouds and sovereign clouds. The expansion, announced Aug. 25, extends the full-stack architecture to any AI use case, from trillion-parameter training to edge inference, and pairs Cisco's networking with Supermicro's high-density compute.
"We are at the beginning of one of the largest datacenter buildouts in history," Jeetu Patel, president and chief product officer at Cisco, said. "Every organization is racing to scale AI — but speed only counts if it comes with control of data, managed token costs, and real ROI."
The NCP-compliant architecture pairs Cisco Silicon One-based switches on the front end with NVIDIA Spectrum-X-based switches on the back end, unified by Cisco Nexus One. Cisco says it is the only NVIDIA technology partner to run its own switches and network operating system in an NCP-compliant solution — a differentiator against Arista Networks and Broadcom, which supply networking gear to hyperscale and AI customers. The systems support NVIDIA Vera Rubin NVL72 and HGX Rubin NVL8 platforms, with availability beginning October 2026.
"AI factories are revenue-generating infrastructure, where compute produces intelligence, and intelligence drives revenue," Justin Boitano, vice president of Enterprise AI at NVIDIA, said. "By expanding the Cisco Secure AI Factory with NVIDIA, Cisco is providing enterprises and neoclouds full-stack infrastructure that can get into production faster and generate more value from every watt."
Customers gain access to Supermicro's liquid- and air-cooled systems, validated and sold as part of Cisco's broader AI infrastructure portfolio, letting them manage high-density AI clusters alongside non-AI workloads. Rack-to-fabric liquid cooling pairs Cisco's liquid-cooled AI networking systems with Supermicro's liquid-cooled servers, unlocking high-throughput inference and training workloads that demand dense GPU clusters.
To cut deployment risk, Cisco is introducing Cisco Validated Infrastructure Services (CVIS), aligned with NVIDIA Infrastructure Services, to certify that infrastructure is built exactly as designed and aligned with reference architectures. Cisco is investing in a dedicated large-scale AI lab to develop tools and test software supporting CVIS. On the operations side, NVIDIA AI Enterprise software and AgenticOps through Cisco Cloud Control let customers correlate job health with compute, NIC, optics and network performance metrics for end-to-end observability.
Cisco and Supermicro each bring global supply chain expertise to mitigate delivery timeline challenges, particularly around GPU and memory access. Early customers include Sharon AI, whose chief executive James Manning said NCP validation gives confidence that infrastructure is optimized from day one, and WWT, whose cloud and AI chief technology officer Neil Anderson called the full-stack approach exactly what customers need now. ePlus and NTT DATA also cited the partnership's end-to-end reach across compute, networking, power and liquid cooling.
The expansion puts Cisco squarely against Arista Networks and Broadcom in the AI networking market, while giving NVIDIA a deeper channel into enterprise, neocloud and sovereign cloud customers. For Cisco, the move ties its networking franchise to the fastest-growing part of data center spending; for NVIDIA, it extends the reach of its reference architectures beyond hyperscalers. Supermicro compute solutions become available as part of the Secure AI Factory in October 2026.
This article is for informational purposes only and does not constitute investment advice.
