AI workloads are pushing the limits of traditional data center infrastructure, demanding new levels of performance, scale, security, and operational rigor. To meet these needs, Cisco and NVIDIA are delivering validated, enterprise-ready AI infrastructure-bringing together Cisco switching, compute, security, and Silicon One innovations with NVIDIA accelerated computing, NVIDIA AI Enterprise software, NVIDIA Spectrum-X, and NVIDIA BlueField technologies.

At NVIDIA GTC in March, leaders from both companies laid out the results of this collaboration and how it's shaping the future of AI infrastructure across accelerated processing, networking, security, and operations.

Spectrum-X + Cisco Silicon One: A fabric for AI at scale

At the core of the initiative is the integration of NVIDIA Spectrum-X capabilities into Cisco Silicon One networking infrastructure. In addition to selling NVIDIA Spectrum-based switches, this provides customers with flexibility in a unified operational model.

Key takeaways:

• Cisco has developed new reference architectures (based on NVIDIA Enterprise Reference Architectures), including Cisco Silicon-based switches with NVIDIA SuperNICs as well as Cisco Spectrum-based switches with NVIDIA SuperNICs.
• Integration with NVIDIA Ethernet SuperNICs (NVIDIA ConnectX and BlueField) enhances fabric-wide awareness, optimizing traffic flows, managing congestion, and improving buffer utilization.
• Future updates will make telemetry-level integrations possible, enabling deep NIC-to-switch insights for optimization and fault isolation.

As announced at Cisco Live 2025 San Diego, the key software-level integration between a Cisco NX-OS switch and BlueField-3 was demonstrated, and a Spectrum-X license will soon be available to add to NX-OS switches. To learn more about this exciting announcement, check out this video.

Cisco Nexus Hyperfabric AI: End-to-end full-stack AI infrastructure solution

Nexus Hyperfabric AI is the Cisco platform for AI cluster design, deployment, and lifecycle operations. It goes far beyond switch-level configuration to deliver a full-stack, cloud-managed deployment and ops model, which includes Cisco and NVIDIA networking, accelerated computing, and NVIDIA AI Enterprise software.

Core capabilities include:

• Full-stack blueprinting:Design your topology before you buy (including processing, networking, and storage) without provisioning hardware. Hyperfabric.cisco.com supports early modeling with reference architectures for Cisco customers.
• Cable plan + optics validation:Auto-generate exact physical topologies, including optics, cables, reach, and SKU-level BOMs, to prevent costly procurement errors.
• Helping Hands Deployment Assist feature:Validate physical install step-by-step at rack level; flag miswires in real time using digital twin validation.
• Assertion-based monitoring:Identify the root cause of any detected issue by leveraging fabric and resource monitoring that continuously verifies availability and reliability.
• Agent-based fabric intelligence:NIC-level agents generate synthetic traffic, run end-to-end health checks, and surface anomalies via an integrated telemetry stack.

Nexus Hyperfabric AI ships in summer 2025 with built-in support for BlueField-3 SuperNICs, AI-specific fabric telemetry, and automated validation.

Cisco Secure AI Factory also offers a modular approach to the NVIDIA reference architecture built on the Cisco Nexus Dashboard and Cisco Nexus 9000 Series Switches.

Security at scale: Cisco Secure AI Factory with NVIDIA

As enterprises scale AI workloads, infrastructure security becomes a primary concern. NVIDIA envisions AI factories as integrated, purpose-built systems with unified compute, networking, and software to industrialize the AI lifecycle-from data ingestion to inference-and has created reference architectures for partners to implement. The Cisco Secure AI Factory builds on the NVIDIA AI Factory vision, adding native observability and enforcement across compute, networking, software, and data planes.

Robust security for the AI infrastructure is delivered through two core Cisco solutions:

• Cisco AI Defense:Provides runtime large language model (LLM) observability and threat detection across application programming interfaces (APIs), web-based inference endpoints, and hybrid deployments. Currently in public beta.
• Cisco Hypershield:Offers data processing unit (DPU)-based enforcement using NVIDIA BlueField to create distributed, inline policy enforcement without adding hardware; integrates with Extended Berkeley Packet Filter (eBPF) and supports asymmetric perimeter policies.

The Cisco Secure AI Factory architecture avoids costly hairpinning or hardware retooling and leverages existing BlueField SuperNIC investments for inline enforcement and introspection.

Real engineering depth

The collaboration will enforce the pillars of the Secure AI Factory and expand on them. This includes:

• Dual silicon strategy:Cisco can deploy either Silicon One-based or NVIDIA Spectrum-based switches (running NX-OS) in back-end fabrics. This gives customers optionality with a unified operational model.
• Fabric + NIC synchronization:Co-engineering efforts ensure that Silicon One switches are fully supported by NVIDIA SuperNICs, thereby optimizing end-to-end behavior.
• Expanded BlueField integration:Cisco will strengthen its integration with NVIDIA BlueField to support accelerated cybersecurity use cases, including zero-trust security and real-time threat detection.

AI infrastructure is no longer about piecing together isolated components. It's about building integrated systems that scale linearly, operate intelligently, and secure dynamically. With Cisco and NVIDIA, enterprise customers get an architecture that does all three-across accelerated computing, software, networking fabric, and security.

Whether customers are just starting to build an AI factory with dozens of accelerated computing clusters or planning for a large-scale deployment, this initiative sets the stage for scalable, secure, and simple-to-operate AI clusters.

Find out more about Cisco Secure AI Factory with NVIDIA.