
Cisco has announced new solutions to accelerate the construction of secure and scalable AI infrastructure for various market segments — from neocloud providers and sovereign clouds to corporate customers and telecom operators.
The announcement was made at GTC in Washington, D.C., on October 28, 2025. The centerpiece was the Cisco N9100 — the first data center switch developed by an NVIDIA partner and based on NVIDIA Spectrum-X Ethernet switching silicon.
Key announcements include:
- The launch of the Cisco N9100 switch series with support for Cisco NX-OS or SONiC operating systems;
- An offer for a reference architecture that meets NVIDIA Cloud Partner requirements for neocloud and sovereign cloud deployments;
- The development of Cisco Secure AI Factory with NVIDIA for enterprises through updates in computing, security, networking, and observability;
- The presentation of the first AI-native wireless stack for 6G in collaboration with NVIDIA and telecom partners.
According to Jeetu Patel, Cisco's President and Chief Product Officer, the industry is at the "beginning of the largest data center capacity buildup in history," and future agentic AI applications will require new architectures that account for energy, computing, and network performance constraints.
New Switches and Reference Architecture for Clouds
Cisco stated that N9100 switches will be available for order by the end of the year and will offer a choice between NX-OS and SONiC. The company positions this as a step toward greater flexibility in building Ethernet networks for AI, as well as unified operation through the Cisco Nexus Dashboard.
Separately for neocloud and sovereign cloud deployments, Cisco introduced the Cisco Cloud Reference Architecture. It is based on the principles of the NVIDIA Cloud Partner reference architecture and utilizes switches based on Cisco Silicon One and Cloud-scale ASIC, including the recently introduced Cisco 8223 based on Silicon One P200 for scale-across networks. The package also includes NVIDIA BlueField-4 DPUs and NVIDIA ConnectX-9 SuperNICs.
NVIDIA emphasized that Spectrum-X Ethernet is designed for "accelerated Ethernet-based networks," and the combination of NVIDIA Cloud Partner principles and Cisco's Cloud Reference Architectures allows customers to choose between deployment on the latest Cisco N9100 switches or Cisco solutions based on Silicon One.
Secure AI Factory and the Step Toward AI-native Networks for 6G
For corporate customers, Cisco is evolving the Cisco Secure AI Factory with NVIDIA architecture, which was introduced at GTC in March 2025. As part of the updates, the company announced expanded integrations in security and observability, as well as new capabilities for core AI infrastructure and the ecosystem.
Specifically, Cisco AI Defense is integrating with NVIDIA NeMo Guardrails. It was also announced that Cisco AI Defense is available for on-premises deployment at the data plane level, allowing security and AI teams to protect models and applications while limiting the amount of sensitive data leaving the organization's data centers. For AI stack monitoring, Splunk Observability Cloud and Splunk Enterprise Security were mentioned — including health observability for AI infrastructure with Cisco AI PODs.
Regarding infrastructure, Cisco announced the validation of Cisco Isovalent for inference workloads on AI PODs, as well as the availability for order of Cisco Nexus Hyperfabric AI with the cloud-managed Cisco G200 Silicon One switch, providing high-density 800G Ethernet as a deployment option in AI PODs. Within AI PODs, Cisco UCS 880A M8 servers with NVIDIA HGX B300 and modular Cisco UCS X-Series servers with NVIDIA RTX PRO 6000 Blackwell Server Edition have also become available for order.
Furthermore, Cisco announced an ecosystem expansion: NVIDIA Run:ai is becoming available through Cisco and partners for AI workload and GPU orchestration. The Nutanix Kubernetes Platform (NKP) was named a supported Kubernetes platform, Nutanix Unified Storage (NUS) as a supported storage option, and Nutanix Enterprise AI (NAI) as a software component to simplify the construction and operation of containerized inference services.
For the public sector, the company noted its work on alignment with the NVIDIA AI Factory for Government — a reference design for AI workloads in highly regulated environments.
In the telecom sector, Cisco, NVIDIA, and other industry partners introduced the first AI-native wireless stack for 6G. It is positioned as a transition path for operators into the "AI era" — starting with implementation via 5G Advanced and laying the groundwork for 6G. The stack description indicates it combines Cisco's user plane function and 5G core software with the NVIDIA AI Aerial platform to support physical AI and integrated sensing with a focus on efficiency and security.
Cisco also cited reactions from several industry representatives — from cloud providers to integrators — who emphasized the importance of a unified operating model (NX-OS/SONiC and Nexus Dashboard), infrastructure openness, and reduced complexity when scaling GPU clusters.