Cisco and NVIDIA Redefine Enterprise AI Infrastructure for an Agentic Future

The AI landscape of 2026 is defined by action. Systems no longer just analyze; they reason and execute in real time. This shift to 'agentic AI' has inverted the traditional workload model, with inference—the process of running live AI—now constituting the overwhelming majority of enterprise AI compute, a figure industry observers place above 80%. This execution happens everywhere: core data centers, specialized clouds, and, critically, at the network edge where data is born and instant decisions are required.

This distribution creates a formidable challenge. Managing performance, security, and policy across a fragmented mosaic of systems stifles scale. Cisco's response, developed in partnership with NVIDIA, is the Secure AI Factory—an architectural blueprint designed from the ground up for this new reality. Its core principle: enterprise AI must be secure and observable to be scalable. By pre-integrating compute, networking, and storage into validated systems, Cisco aims to remove the integration burdens that delay projects.

The company is now extending this framework. New announcements focus on three fronts: performance, with support for NVIDIA's latest Spectrum-6 Ethernet silicon in Cisco switches; security, via embedding Cisco's firewall technology directly into NVIDIA BlueField servers; and edge deployment. For the edge, Cisco is introducing support for NVIDIA's Blackwell GPUs across its UCS servers, enabling AI in settings from retail floors to remote factories. A separate reference architecture, Cisco AI Grid with NVIDIA, helps telecom providers offer managed AI services on their existing infrastructure.

To simplify large-scale rollouts, Cisco is publishing new validated design guides for major storage partners. The company has also integrated Red Hat's AI Factory software into the platform, expanding the open-source tooling available to development teams. The objective is to provide a consistent operational model from core to edge, turning the ambition of pervasive, agentic AI into a manageable engineering reality.

Source: Cisco

Source:Cisco
← Back to News