Cisco extends its Secure AI Factory with Nvidia

Cisco and Nvidia continue to develop integrated packages aimed at helping enterprises and larger customers build the infrastructure needed to deploy and secure AI at scale.

As Nvidia’s GTC event kicks off this week, the two vendors announced the expansion of their jointly developed Secure AI Factory with Nvidia, which melds Cisco security and networking technology, Nvidia DPUs and AI Enterprise software, and multivendor storage options.

New to the portfolio is the 102.4Tbps Cisco N9100 powered by Nvidia’s Spectrum-6 Ethernet switch silicon. The new switch joins the 800G N9100 powered by Nvidia Spectrum-4 Ethernet switch silicon, which is available now.

** Related:[More Cisco news and insights]**

Cisco’s N9100 Series and its Secure AI Factory are unique in that they can be deployed in Nvidia Cloud Partner (NCP)-compliant reference architectures based on Spectrum-X Ethernet, or in strictly Cisco Cloud Reference Architectures built on Silicon One Nexus switching environments.

Depending on requirements, customers can choose between different network operating systems, including Cisco NX-OS or the open-source SONiC (Software for Open Networking in the Cloud). The intention is to provide a network backbone for a variety of customers, from the enterprise to neocloud and sovereign cloud, Cisco stated.

N9100 customers will be able to use Cisco Nexus Hyperfabric, which is now a part of Cisco Nexus One, to more easily integrate the AI systems, said Kevin Wollenweber, senior vice president and general manager of Cisco data center and internet infrastructure. Specifically, Hyperfabric automates the deployment and operation of AI infrastructure, including GPU clusters, Ethernet networking, telemetry and storage systems.

“Customers can now control and manage this environment and operate it like it was a traditional data center fabric,” Wollenweber said. “The ability to bring it under the same Nexus umbrella is actually a huge selling point for AI customers, because their IT infrastructure folks, their operational people that are running the network, already understand how to use these Nexus tools, and so they can now add AI workloads and kind of accelerated computing technologies like GPUs, but in that same Nexus umbrella,” Wollenweber said.

“As Al becomes operational and distributed, complexity becomes the enemy of scale. Fragmented architectures force customers to manage integration, policy enforcement, observability, and security across silos, increasing cost and slowing innovation,” said Wollenweber. “Architecting silicon, networking, compute, security, and Al software into a cohesive system gives organizations a unified operating model, stronger performance guarantees, and embedded trust.”

Those are the driving ideas around Cisco Secure AI Factory with Nvidia, Wollenweber said.

Introduced a year ago, Secure AI Factory with Nvidia integrates Cisco’s Hypershield and AI Defense packages to help protect the development, deployment, and use of AI models and applications. Hypershield uses AI to dynamically refine security policies based on application identity and behavior. It automates policy creation, optimization, and enforcement across workloads. AI Defense discovers the various models being used in a customer’s AI development and uses four features to help customers enforce AI protection: AI access, AI cloud visibility, AI model and application validation, and AI runtime protection.

Cisco integrates Hybrid Mesh Firewall technology

On the security side, Cisco said it will embed its Hybrid Mesh Firewall technology to allow for security policy enforcement on Nvidia BlueField data processing units (DPU) that are embedded in Nvidia GPU servers connected to Cisco Nexus One fabrics. Cisco Hybrid Mesh Firewall offers a distributed security fabric that features a zero-trust security framework integrated into the network for segmentation, AI application protection and advanced threat protection, Cisco stated.

The idea is to block threats at the server level before they ever reach an organization’s data, Wollenweber said. Customers can protect AI workloads by microsegmenting those systems as well as offer application-level security.

The companies are also extending Cisco AI Defense by adding support for Nvidia’s NeMo Guardrails, which is part of Nvidia’s overall AI Enterprise software package. NeMo Guardrails lets developers control how AI LLMs behave. In this case, Cisco and Nvidia said they are looking to aim the package at helping customers control AI agent behavior.

Cisco AI Defense will support and secure Nvidia’s new open agent development platform, OpenShell, adding controls and guardrails to govern agent and claw actions, Cisco stated.

In addition to the security enhancements, Cisco said it will be adding support for Nvidia’s RTX Pro 4500 Blackwell server in its UCS portfolio and in its Unified Edge package. The RTX PRO 4500 Blackwell server is a data-center optimized GPU for AI workloads. The idea is that UCS customers could add GPUs directly into UCS servers already deployed in data centers.

The Cisco Unified Edge package includes a 25G network uplink and is powered by an Intel Xeon System on a Chip (SOC) central processing unit. The 3U, 19-inch chassis features five front-facing slots for compute nodes, GPUs and future modules that may include a networking node for routing, switching, and secure access service edge (SASE) capabilities. By now supporting the 4500 Blackwell, customers can support AI workloads at the edge, Cisco said.

More Cisco news:

• Cisco extends its Secure AI Factory with Nvidia
• Cisco blends Splunk analytics, security with core data center management
• Cisco: LPO not a panacea but plays strategic role in AI networks
• Cisco issues emergency patches for critical firewall vulnerabilities
• How Cisco’s platform mindset is meeting the AI era
• Cisco extends AgenticOps model across networking, security, observability products
• Cisco amps up Silicon One line, delivers new systems and optics for AI networking
• Takeaways from Cisco’s AI Summit
• Cisco: Infrastructure, trust, model development are key AI challenges
• AI, security tailwinds signal promising 2026 for Cisco
• Cisco adds intelligent policy enforcement to mesh firewall family