Cisco Expands Secure AI Factory With NVIDIA

Cisco has announced a significant expansion of its Secure AI Factory in collaboration with NVIDIA, unveiling a comprehensive framework designed to help organisations deploy artificial intelligence (AI) across their entire infrastructure, from centralised data centres to distributed edge environments.

The move is aimed at addressing one of the most persistent challenges facing enterprises today: transitioning AI initiatives from experimental pilots to full-scale, secure production. By delivering an integrated architecture, Cisco is enabling enterprises, neocloud providers, sovereign cloud operators, and service providers to deploy AI systems more rapidly, reducing timelines from months to weeks while embedding security from the outset.

According to Chuck Robbins, Chair and CEO of Cisco, organisations are increasingly aware of AI’s transformative potential but remain cautious about scaling deployments securely.

“In partnership with NVIDIA, we are addressing this challenge with an architecture that sets a new benchmark for performance, simplifying how AI infrastructure is deployed, operated, and secured,” he said.

Echoing this sentiment, Jensen Huang, founder and CEO of NVIDIA, emphasised the importance of embedding security across the entire AI stack.

“AI factories are transforming every industry, and security must be built into every layer – from silicon to software – to protect data, applications, and infrastructure,” Huang noted. “Together, we are building a secure foundation that enables organisations to scale intelligence with confidence.”

Driving AI Beyond the Data Centre

A key focus of the expanded Secure AI Factory is enabling AI inference at the edge, where data is generated and real-time decisions are required. This shift reflects growing demand for low-latency processing in sectors such as healthcare, manufacturing, and telecommunications.

Cisco and NVIDIA are advancing this capability through enhanced support for GPU-powered edge deployments. By integrating NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs into Cisco’s UCS and Unified Edge portfolios, enterprises can now run mission-critical AI workloads closer to where data resides, without the cost and complexity of full-scale data centre infrastructure.

At the service provider level, the newly introduced Cisco AI Grid – built on a joint reference design – combines Cisco’s Mobility Services Platform with NVIDIA’s Blackwell GPU architecture. This enables telecom operators to monetise their networks by delivering managed AI services with carrier-grade reliability and data sovereignty.

Scaling Performance for AI Factories

Cisco is also pushing the boundaries of performance for large-scale AI environments. Its latest networking innovations include high-capacity switching systems designed to handle the most demanding AI workloads.

Among the highlights is the introduction of a 102.4Tbps Cisco N9100 switch powered by NVIDIA Spectrum-6 Ethernet silicon, alongside the general availability of 800G switching solutions based on Spectrum-4 technology. These advancements are intended to support the growing data and compute requirements of AI factories.

To simplify deployment, Cisco has integrated its Nexus Hyperfabric solution into the Nexus One platform, enabling organisations to transform complex, multi-vendor environments into streamlined, full-stack deployments. This reduces operational complexity while accelerating time to value.

Customers building AI infrastructure can now choose between two validated deployment models: one aligned with NVIDIA’s Cloud Partner programme and another based on Cisco’s Cloud Reference Architecture powered by Silicon One, both designed to deliver scalability and consistency.

Embedding Security Across the AI Stack

Security remains a central pillar of the Secure AI Factory. As AI models and autonomous agents become increasingly critical and valuable assets, Cisco is embedding protection across every layer of the infrastructure.

The company’s Hybrid Mesh Firewall extends policy enforcement across network switches, workloads, and now NVIDIA BlueField data processing units embedded in GPU servers. This approach allows threats to be mitigated at the server level before reaching sensitive data.

In addition, Cisco AI Defense introduces enhanced capabilities for securing AI models and agents. Through integration with NVIDIA’s NeMo Guardrails, organisations can implement safeguards for AI behaviour, conduct automated vulnerability testing, and secure interactions between distributed AI agents operating across edge and core environments.

Securing the Agentic Enterprise

Cisco is further strengthening its AI security portfolio by extending support for NVIDIA’s OpenShell runtimes, part of the NVIDIA Agent Toolkit. This integration introduces governance controls and continuous monitoring of AI agent actions, ensuring that autonomous systems can operate safely within enterprise environments.

By validating every action performed by AI agents, Cisco aims to bridge the gap between innovation and risk, enabling organisations to deploy intelligent automation without compromising security or trust.

Industry Perspective

Industry stakeholders have welcomed the expanded partnership, highlighting its potential to simplify AI infrastructure deployment while improving scalability and performance.

Technology leaders point to the combination of high-performance networking, flexible architectures, and integrated security as critical enablers for the next phase of AI adoption, particularly as organisations shift focus from experimentation to operationalisation at scale.

Analysts also note that the collaboration addresses a growing need to bring AI processing closer to the edge, where real-time insights are generated, without introducing new infrastructure silos or security vulnerabilities.