Cisco and Nvidia take AI factories from rack to runtime

AI factories are moving from concept to reality as AI systems are increasingly being brought into production. Will Eatherton, Cisco's senior vice president, noted that enterprises are spending heavily on tokens and are eager to offload applications from the cloud to local inference. Nvidia's Gilad Shainer and Marc Hamilton, vice president of solutions architecture and engineering, joined Eatherton to discuss the advancements in AI factories and rack-scale compute at the "Cisco Secure AI Factory With Nvidia Expands to Rack Scale" event.

Cisco is expanding its Secure AI Factory beyond networking to include full rack-scale compute, partnering with Supermicro to bring liquid-cooled systems to market. The solution is designed to be compatible with Nvidia's reference architecture and Cisco Validated Designs, ensuring seamless integration and reducing late-stage integration issues.

Both companies are focusing on monitoring, availability, software upgrades, and lifecycle management for customers operating these systems over time. The partnership also allows for flexibility within the networking stack, enabling customers to customize the technology without compromising on familiar operating tools.