Cisco and Nvidia unveil new secure AI edge framework

Cisco has expanded its Secure AI Factory with Nvidia into a broader framework for deploying and securing AI across central data centres and edge locations such as hospitals, warehouses and vehicles.

The expanded architecture ties compute, networking and security into reference designs that Cisco positions as a way for enterprises and service providers to move AI from pilots to production without bespoke integrations.

The announcement focuses on two edge tracks: one for enterprise deployments that run AI inference closer to where data is generated, and another for service provider infrastructure that can host managed edge AI services.

Enterprise edge

At the enterprise edge, Cisco will support Nvidia RTX PRO 4500 Blackwell Server Edition GPUs across its Cisco UCS and Cisco Unified Edge portfolios. Cisco says this will let businesses run AI workloads outside the data centre while reducing power use and physical footprint compared with larger-scale systems.

The emphasis on local inference reflects growing demand for low-latency AI processing in operational environments. Cisco pointed to use cases such as analysing video feeds on factory floors and supporting real-time decisions in clinical settings.

Service provider edge

For telecoms operators and other service providers, Cisco introduced a reference design called Cisco AI Grid with Nvidia. It combines Cisco's Mobility Services Platform with Nvidia RTX PRO Blackwell Series GPUs.

Cisco says the design enables service providers to use existing networks to offer managed services for edge AI applications, with reliability and data sovereignty as key considerations in regulated environments.

Switching upgrades

Cisco also detailed new networking hardware for large-scale AI infrastructure, including a 102.4 Tbps Cisco N9100, based on Nvidia Spectrum-6 Ethernet switch silicon.

The line-up also includes an 800G N9100, based on Nvidia Spectrum-4 Ethernet switch silicon, which is now generally available. Cisco also cited its earlier work on Cisco Silicon One G300 for scale-out and P200 for scale-across systems.

In addition, Nexus Hyperfabric is now part of Cisco Nexus One and will support Cisco N9000 Series switches. Cisco positioned Nexus One as a unified management plane for these environments.

Customers building large "AI factories" will have two validated design paths. One uses a reference architecture compliant with the Nvidia Cloud Partner programme; the other is a Cisco Cloud Reference Architecture built on Cisco Silicon One, which Cisco says follows the same design tenets.

Security focus

Security across infrastructure and AI agent deployments is a central theme of the expanded framework. Cisco is extending its Hybrid Mesh Firewall so policy enforcement can run on Nvidia BlueField data processing units in Nvidia GPU servers connected to Cisco Nexus One fabrics.

The goal is to block threats at the server level before they reach data and applications, reflecting a shift towards enforcement points closer to workloads rather than relying solely on perimeter defences.

Cisco is also integrating Cisco AI Defence with Nvidia NeMo Guardrails, part of Nvidia AI Enterprise software. The companies say this will cover model security, automated vulnerability testing and guardrails for AI agents at the edge, and extend to securing agent-to-agent interactions in distributed deployments.

Separately, Cisco says AI Defence will support and secure Nvidia OpenShell runtimes, part of the Nvidia Agent Toolkit. Cisco says the integration adds controls and guardrails to govern "agent and claw actions".

"Most organizations understand the potential for AI to transform their businesses, but they're navigating how to deploy the technology safely and at scale," said Chuck Robbins, Chair and CEO, Cisco. "In partnership with NVIDIA, we're solving that challenge with an architecture that sets a new standard for performance - making it simpler to deploy, operate, and secure AI infrastructure."

Nvidia highlighted security as a requirement for AI infrastructure that spans both core data centres and edge sites.

"AI factories are transforming every industry, and security must be built into every layer-from silicon to software-to protect data, applications, and infrastructure," said Jensen Huang, founder and CEO of NVIDIA. "Together, NVIDIA and Cisco are building the secure foundation for AI infrastructure-core to edge-so companies can scale intelligence with confidence."

Industry partners also cited switching density, management and reference-architecture alignment as key factors in scaling AI infrastructure. Mary Johnston Turner, Global Lead for Digital and Datacenter Infrastructure and Services at IDC, said the market has moved from experimentation to operational execution and highlighted the need to push real-time inference to the edge without creating new silos.