Cisco Secure AI Factory with NVIDIA FAQ

Overview

Cisco Secure AI Factory with NVIDIA, developed in collaboration with NVIDIA and strategic ecosystem partners, is a modular reference design that accelerates secure and rapid AI adoption from core to edge. It combines advanced AI infrastructure, including rack-scale systems and dense GPU systems (with liquid- and air-cooled options), with integrated security and observability from Cisco. By leveraging enterprise data and AI models as foundational inputs, it empowers organizations to operationalize AI pipelines with agility and purpose, enabling faster delivery of trusted agentic and physical AI applications and unlocking business value through AI.

This security-first solution embeds security and observability at every layer of the stack, continuously monitoring and analyzing security posture to provide advanced threat detection, investigation, and response. Unlike other AI factories, Cisco Secure AI Factory with NVIDIA integrates security and observability throughout the entire AI infrastructure, helping enterprises securely develop and deliver trusted AI tokens and applications.

●     Cisco provides high-performance, enterprise-proven networking, liquid-cooled rack-scale infrastructure, dense GPU servers (both liquid-cooled and air-cooled options), and scalable storage from technology partners that, working together, form the foundation of a secure and efficient AI infrastructure. Together with AI orchestration and application software, these capabilities accelerate every phase of the AI pipeline, enabling faster development and delivery of trusted AI outcomes. For more than 40 years, enterprises have relied on Cisco’s market-leading^® networking, which is now even more important to the success of AI initiatives. Cisco’s high-performance Ethernet networking offers options to deploy switches with Cisco Silicon One^® or NVIDIA Spectrum-X silicon for AI scale-out fabric.

●     Flexible, modular deployment options help operationalize the solution at your own pace and make sure infrastructure limitations and security concerns aren’t barriers to AI adoption. Enterprises have the flexibility to choose between a modular, pre-validated AI infrastructure stack, backed by Cisco Validated Designs (CVDs); a turnkey cloud-managed AI infrastructure stack; or a build-your-own option to purchase and deploy individual products as needed by buying Cisco and partner products separately and integrating them on their (the enterprise’s) own or with help from Cisco or one of our partners. For large-scale deployments, Cisco is the first NVIDIA technology partner to deliver an NVIDIA Cloud Partner (NCP)–compliant reference architecture built on partner-developed networking systems, spanning both Cisco Silicon One and NVIDIA Spectrum-X Ethernet switch silicon, unified by the Cisco Nexus® One architecture, with a choice of Cisco NX-OS (the industry’s most deployed data center OS) or SONiC. To accelerate these deployments, Cisco Validated Infrastructure Services (CVIS) enables an end-to-end solution design validation, aligned with NVIDIA’s NVIS methodology, to validate the full-stack design against the reference architecture. To learn more about the expansion of Cisco Secure AI Factory with NVIDIA to rack-scale architecture in partnership with Supermicro, please see the FAQ here.

●     Professional services from Cisco and our partners minimize technical, deployment, and financial risk — helping ensure a fast time to market and a low cost per token.

What sets Cisco Secure AI Factory with NVIDIA apart from other AI infrastructure solutions is its foundational emphasis on security and observability at all layers of the stack, and the Cisco high-performance Ethernet networking that enterprises have trusted for 40 years. This enables enterprises with exceptional flexibility, enhanced security, and superior performance, equipping them to navigate the evolving landscape of AI with confidence.

Cisco Secure AI Factory with NVIDIA

●     Complex AI infrastructure deployments: It simplifies the deployment of secure, scalable, and well-architected AI infrastructure from core to edge by offering a modular reference design that combines AI software, compute, networking, storage, security, observability, and a Kubernetes platform into a secure AI infrastructure, helping organizations avoid stalled AI projects and reduce operational complexity.

●     AI security vulnerabilities: It provides integrated security across the entire AI pipeline — protecting AI models, frameworks, applications, agents, and infrastructure from emerging cyber threats such as prompt injection, adversarial attacks, model poisoning, data leaks, and unauthorized GPU access.

●     AI infrastructure performance bottlenecks: The solution delivers enterprise-grade, rack-scale, high-performance AI infrastructure to maximize GPU utilization and optimize performance for AI pipeline phases including training, optimization, and inferencing. This includes handling both east/west traffic (between GPU servers) and north/south traffic (clients to GPU servers to storage), which are the dominant traffic patterns in AI workloads.

●     AI development and delivery software

●     AI compute (GPU-accelerated servers, with air- and liquid-cooled options, and rack-scale systems)

●     High-performance Cisco Nexus One based AI networking (with air- and liquid-cooled options) and Cisco optics

●     Storage certified by Cisco and NVIDIA

●     AI security (model, application, agents, workload, and infrastructure protection, detection, and response)

●     Observability for AI (infrastructure and AI agent monitoring)

●     AI orchestration and GPU management software

●     Kubernetes platform

Key capabilities of Cisco Secure AI Factory with NVIDIA

Key products in Cisco Secure AI Factory with NVIDIA

●     Security at every layer: Unlike other AI factories in the market, Cisco Secure AI Factory with NVIDIA embeds security at every layer of the stack (AI models, agents, and associated software components, applications, workloads, and infrastructure) to help securely develop and deliver trusted AI tokens and applications. Cisco AI Defense integrated with NVIDIA AI, Cisco Hybrid Mesh Firewall that includes Isovalent, Cisco Secure Firewall, and Splunk Enterprise Security enables end-to-end security for the full stack. Hybrid Mesh Firewall serves as a single enforcement point for security policies, including enforcement on NVIDIA BlueField DPUs on AI servers, preserving CPU and GPU resources for AI processing.

●     Cisco AI Networking: With 40 years of enterprise networking experience, Cisco delivers high-performance Ethernet options for AI infrastructure. Customers can deploy Cisco Silicon One–powered N9300 Series switches and N9100 Series switches powered by NVIDIA Spectrum-X Ethernet silicon to support scale-out front-end and back-end networking, from enterprise environments to rack-scale deployments. For NVIDIA Cloud Partner Reference Architecture (NCP RA)–compliant designs, Cisco Validated Infrastructure Services (CVIS) helps customers deploy and scale infrastructure as their business needs grow.

●     Cisco Unified Edge: Purpose-built for AI inferencing at the edge, Cisco Unified Edge consolidates compute, networking, and security into a single modular chassis, enabling real-time AI inferencing and agentic workloads without data-center latency.

●     Observability for AI: Splunk delivers end-to-end visibility across Cisco Secure AI Factory with NVIDIA, enabling teams to monitor the performance, quality, security, and cost of their AI infrastructure stack. Specifically, Splunk AI Infrastructure Monitoring helps ensure that the AI Infrastructure stack remains performant, resilient, and secure. This includes human-guided AI assistants such as the Splunk AI Assistant for SPL, the AI Assistant in Splunk Observability, and Cisco AI Canvas, as well as autonomous AI agents such as the troubleshooting agent for Splunk Observability. Splunk Observability can also monitor agentic AI applications (Splunk AI Agent Monitoring).

●     Workload PODs: run AI workloads such as model training, optimization, and inferencing using Cisco’s T-shirt sized, full-stack Cisco Validated Designs (CVDs) optimized for AI workloads.

●     Services PODs: deliver essential shared capabilities—including security, observability, and data services, serving multiple workload PODs within a Cisco Secure AI Factory with NVIDIA deployment.

●     Securing the AI models, agents, and applications: Cisco AI Defense, integrated with NVIDIA AI, empowers security and AI practitioner teams with comprehensive tools for robust testing and runtime security of LLMs and generative AI applications and agents. Utilizing algorithmic red-teaming techniques, AI Defense evaluates generative AI models against diverse security (data privacy, prompt injections, etc.) and safety (for example, toxic behavior) risks without requiring application modifications. Additionally, AI Defense applies runtime controls to help ensure applications comply with leading frameworks, including OPSWAT LLM and MITRE ATLAS.

●     Securing the workloads and infrastructure: Cisco Hybrid Mesh Firewall delivers unified security management with consistent, pervasive policy enforcement across multiple control points. Here are the key products that make up Hybrid Mesh Firewall:

◦    Isovalent: provides enhanced visibility into containerized workloads, protecting against lateral movement and proactively mitigating vulnerabilities

◦    Cisco Secure Firewall: enables advanced threat protection across perimeters at scale without compromising performance

◦    Cisco Live Protect: enables critical infrastructure hardening with vulnerability shields for Cisco AI networking devices with Cisco Nexus One

●     Security operations: Splunk Enterprise Security, a threat detection and incident response platform, enables real-time detection, investigation, and response through powerful analytics, automation, and risk-based insights.

●     Supermicro’s NVIDIA NVL72 rack-scale AI systems (liquid-cooled).

●     Supermicro’s NVIDIA HGX and NVIDIA MGX based dense GPU servers (with liquid- and air-cooled options) with support for NVIDIA B300, Rubin, and RTX PRO Blackwell Server Edition GPUs.

●     Cisco’s rack and modular servers with support for NVIDIA RTX PRO Blackwell Server Edition GPUs

●     Cisco Unified Edge Computing Platform (with support for NVIDIA RTX PRO Blackwell Server Edition, L40S, and L4 GPUs; air-cooled)

●     Observability for AI, powered by Splunk, delivers end-to-end visibility across Cisco Secure AI Factory with NVIDIA, enabling teams to monitor the performance, quality, security, and cost of their AI application stack. Specifically, Splunk AI Infrastructure Monitoring helps ensure that the AI infrastructure stack remains performant, resilient, and secure.

●     With Splunk Observability Cloud, teams gain real-time insights into AI infrastructure health, availability, and resource utilization (for example, GPU, power, network, nodes, and token costs), empowering proactive root-cause analysis, rapid issue resolution, and alerts that improve efficiency and reliability. With AI Agent Monitoring in Splunk Observability Cloud, teams can monitor the performance, quality, security, and cost of LLM and agentic applications, including tracking prompts and responses for hallucinations, bias, and semantic quality.

●     Splunk Enterprise Security extends this visibility to protect AI workloads, correlating security events from Cisco technologies (Cisco AI Defense, Cisco Hybrid Mesh Firewall, Isovalent, Nexus Hyperfabric) with operational data to detect and mitigate threats such as data leaks, prompt injections, and unauthorized access.