Updated
Updated · Cisco Newsroom · Aug 25
Cisco Adds Rack-Scale AI Systems to Secure AI Factory, Starting October 2026
Updated
Updated · Cisco Newsroom · Aug 25

Cisco Adds Rack-Scale AI Systems to Secure AI Factory, Starting October 2026

3 articles · Updated · Cisco Newsroom · Aug 25

Summary

  • Cisco said it will start offering Supermicro rack-scale, liquid- and air-cooled AI compute in October 2026, extending its Secure AI Factory with NVIDIA to support workloads from edge inference to trillion-parameter training.
  • The expansion targets surging neocloud and sovereign cloud demand for secure, validated AI infrastructure that can handle higher power, cooling and networking requirements while reducing deployment risk.
  • Cisco said the new stack is NVIDIA Cloud Partner-compliant and combines Cisco Silicon One front-end switching, NVIDIA Spectrum-X back-end switching and Nexus One management into a unified networking architecture.
  • Supermicro’s dense GPU servers and rack-to-fabric liquid cooling are meant to speed deployment and operations, with Cisco adding validated infrastructure services and a large-scale AI lab to certify builds and improve observability.
  • The move broadens Cisco’s push to sell full-stack AI infrastructure as datacenter buildouts accelerate and customers seek tighter control over data, token costs and return on AI spending.

Insights

Does integrating Cisco, NVIDIA, and Supermicro hardware actually simplify AI factory deployment, or just create a more complex vendor lock-in trap?
Will Cisco's new three-way AI alliance finally solve the massive power and cooling bottlenecks threatening to stall trillion-parameter AI models?
Can this advanced rack-scale infrastructure truly guarantee sovereign cloud compliance against strict new global regulations like the EU AI Act?