Close

Cisco and Nvidia Strengthen Collaboration to Enhance Secure AI Infrastructure

Cisco and Nvidia have taken their partnership to the next level with the launch of an advanced AI infrastructure solution aimed at strengthening enterprise AI security. The newly introduced Cisco Secure AI Factory with Nvidia integrates Cisco’s networking and security technologies, Nvidia’s DPUs, and multiple storage solutions from industry-leading vendors.

Unveiled at Nvidia’s GTC AI Conference in San Jose, California, the Secure AI Factory is designed to protect AI model development and deployment. The solution incorporates Cisco’s Hypershield and AI Defense technologies, which work in tandem to safeguard AI workflows and applications.

Hypershield leverages AI-driven automation to create, refine, and enforce security policies dynamically. It also enables organizations to segment networks in response to potential threats, mitigate vulnerabilities without traditional patching, and ensure seamless software upgrades without service disruption.

AI Defense, on the other hand, provides visibility into AI assets and enhances security across four key areas: AI access control, cloud AI visibility, AI model validation, and runtime protection. It helps organizations manage access to AI resources, monitor unauthorized AI workloads, and provide a consolidated view of AI assets within an enterprise.

In a recent blog post, Cisco’s executive vice president and chief product officer, Jeetu Patel, highlighted the complexity of securing AI infrastructure and emphasized the importance of an integrated security approach. The architecture employs Cisco’s existing security solutions—including Isovalent, Secure Firewall, and AI Defense—along with Nvidia’s BlueField-3 DPUs and DOCA AppShield to provide real-time AI workload threat detection.

The Secure AI Factory is available in two configurations:

  1. Cisco Nexus Hyperfabric AI-based Model:
    • Cisco 6000 Series data center switches managed by the Cisco Nexus Hyperfabric AI controller.
    • Cisco UCS C885A M8, an 8U rack server built on Nvidia’s HGX platform, optimized for AI tasks such as large language model (LLM) training and inference.
    • Nvidia BlueField-3 DPUs and SuperNICs, designed to support high-speed AI workload communication at up to 400Gb/s using RDMA over converged Ethernet (RoCE).
    • Nvidia AI Enterprise software suite, which includes pre-trained models and AI development tools.
    • Storage solutions from VAST Data.
  2. Cisco Nexus 9000 Series Model:
    • Cisco Nexus 9000 Series data center switches with advanced congestion management and flow-control capabilities.
    • Cisco UCS C885A M8 and C845A M8 servers.
    • Nvidia BlueField-3 DPUs and SuperNICs.
    • Future compatibility with Nvidia’s Spectrum-X Ethernet networking platform, managed via the Cisco Nexus Dashboard.
    • Nvidia AI Enterprise software suite.
    • High-performance storage from certified partners, including Pure Storage, Hitachi Vantara, NetApp, and VAST Data.
    • Red Hat’s open-source containerization and automation solutions for optimizing AI workloads.

Cisco and Nvidia also plan to provide flexible deployment models and reference architectures to assist customers in integrating AI infrastructure into their operations.

“The Cisco Secure AI Factory with Nvidia empowers enterprises to build, secure, and scale AI solutions efficiently,” Patel stated. “From data engineering to AI training, inference, and regulatory compliance, this infrastructure is designed to support AI-driven innovation across industries such as public sector, finance, healthcare, manufacturing, and retail.”