We have reached an inflection point in AI: the rise of agentic systems that reason, decide, and act in real time. As organizations operationalize AI, the workload mix has shifted.
Today, roughly 75-85% of AI workloads are inference rather than training, and the percentage continues to climb. Multi-agent systems are emerging rapidly, orchestrating real-time insights and automation across applications, data domains, and environments.
Inference is what brings AI to life; and inference doesn’t live in one place. It spans hyperscale data centers, sovereign and regional clouds, enterprise environments, service provider networks, and increasingly at the edge – wherever data is generated, and decisions must be made instantly. That reality fundamentally reshapes infrastructure strategy.
As AI 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.
An architectural approach changes the equation. Architecting silicon, networking, compute, security, and AI software into a cohesive system gives organizations a unified operating model, stronger performance guarantees, and embedded trust. That alignment is what turns AI ambition into real-world execution. This cohesion becomes a competitive advantage.
Cisco is built for this moment. Our deep expertise and innovative portfolio across silicon, networking, compute, optics, and software, all deeply integrated with security and observability – creates a trusted foundation enabling customers to move from experimentation to production with confidence.
Cisco Secure AI Factory with NVIDIA
The Cisco Secure AI Factory with NVIDIA is our blueprint for accelerating enterprise AI adoption securely at scale. While many AI factory architectures focus primarily on compute performance, Cisco starts with a different premise: enterprise AI must be secure, observable, and governable before it can scale. By pre-validating compute, networking, storage, and software into a unified architecture, we eliminate integration complexity and reduce deployment risk. Security is embedded across the entire stack—from infrastructure and networking to data movement, workload isolation, and policy enforcement—helping organizations protect sensitive data and operate AI systems with confidence. The result is a flexible deployment model that supports on-premises, edge, and cloud environments while ensuring AI infrastructure is secure, governed, and production-ready from day one.
The outcomes are clear:
- Deployment and operations simplicity through a validated, unified operating model
- High-performance AI infrastructure at scale, optimized for demanding training and inference workloads
- Integrated security, observability, and resiliency across the factory, ensures consistent performance, compliance, and uptime
To help our customers navigate the shift to agentic AI and move from experimentation to production, we are introducing new innovations that expand the capabilities of the Cisco Secure AI Factory with NVIDIA. These innovations provide the performance, security, and architectural flexibility required to deploy AI anywhere, from the core to the edge.
Next generation performance
Support for NVIDIA Spectrum-6 Ethernet Switch Silicon on Cisco’s new N9100 102.4T scale-out switch. This expands our NVIDIA Cloud Partner Reference Architecture (NCP RA) compliant portfolio. Cisco’s Spectrum-4-based N9100 for 800G scale-out are generally available now. Our Silicon One G300 based 102.4T N9300 switches, and deep-buffer P200-based scale-out switches advance AI initiatives with the Cisco Cloud Reference Architecture (CRA). We deliver the flexibility and innovation to meet customers wherever they are on their AI journey.
Hardware-accelerated security
Integration of Cisco’s cutting-edge Hybrid Mesh Firewall technology into the NVIDIA BlueField platform on AI servers connected to Nexus One fabrics. This brings hardware-accelerated, robust protection directly to AI servers, enabling developers to work in high-performance, secure environments.
AI at the Edge
AI doesn’t stop at the data center; it runs inside the network. From RAN optimization to enterprise edge services and sovereign AI offerings, inference must execute across metro and far-edge sites where latency, compliance, and resiliency are mission critical. Extending the Secure AI Factory with NVIDIA to the edge ensures consistent security, lifecycle management, and operational control across thousands of distributed locations.
- Enterprise Edge
Support for the new NVIDIA RTX PRO™ 4500 Blackwell Server Edition across Cisco’s UCS portfolios, including our latest innovation to market, Cisco Unified Edge. By delivering an optimized balance of price, performance, and power efficiency, we enable enterprises to run mission-critical AI workloads, from retail analytics to manufacturing automation and healthcare imaging – even in the most constrained edge environments. - Service Provider Edge
Cisco AI Grid with NVIDIA, a reference architecture that brings together the power of Cisco’s Mobility Services Platform and NVIDIA’s RTX Pro Blackwell Series GPUs in Cisco’s compute platforms to enable service providers to leverage their existing networks to offer managed services for physical AI applications with carrier-grade reliability, sovereignty, and compliance. This unified, modular architecture connects core data centers with edge locations, empowering organizations to deploy AI at scale across factories, cities, vehicles, and more.
Ecosystem expansion
To continue offering customers choice and flexibility, Cisco Secure AI Factory with NVIDIA now supports Red Hat AI Factory software, providing practitioners with an expanded toolkit to accelerate AI. This new offering brings together the best of Red Hat and NVIDIA software with optimized microservices, frameworks and libraries for AI, development and deployment of AI workloads based on fully supported open-source AI software tools. Together, they transform the ad hoc creation, customization, and deployment of AI models into a repeatable, scalable, and safeguarded factory process.
Simplifying deployment at scale
Deploying AI infrastructure at scale should be simple, not a process of piecing together fragmented solutions. Cisco AI PODs and Cisco Validated Designs (CVDs) streamline deployment by providing modular, scalable blueprints that reduce operational complexity and enable IT teams to deploy and manage AI infrastructure as a standardized, repeatable service. Today we announced three new CVDs focused on AI model training use cases on FlashStack, FlexPod, and VAST Data.
Leading the agentic AI era
Cisco is committed to empowering customers and partners to lead in this new era. We are investing, innovating, and partnering to deliver the infrastructure that will define the future of AI from the core to the edge.
This isn’t just the next phase of networking. It’s the foundation for the agentic, AI-driven world ahead.
For more detail on these innovations, read:
Cisco gives its Secure AI Factory with NVIDIA a Secure Multi-Agent Edge Up!
