Nvidia and Cisco take AI factory to rack scale

The announcement reflects a shift in enterprise AI from research experimentation to production deployment. Cisco's Will Eatherton and Nvidia's Marc Hamilton both emphasized that the value lies not in the hardware alone but in the operational layer — the management plane, observability, and the ability to debug issues across the entire stack when an AI workload slows down in a real environment.

Cisco is folding in its previously developed tools, like the Nexus One management plane and the Cloud Control operations platform, to provide a common way to run day-to-day operations and monitoring. Nvidia, meanwhile, is adopting this reference architecture fully for its customers. The two companies see this as a way to close the deployment gap between acquiring GPUs and running a production-grade AI factory, which they describe as covering not just a rack but the whole facility, including front-end networks and storage connections.