Every company and its dog will happily chew your ear off about its AI offerings. Much of this is -- how shall we say? -- junk. But Cisco is a different story. Cisco really is an AI powerhouse, thanks in no small part to its hand-in-glove Nvidia partnership.
Today, at Cisco Live, its annual network conference, the company introduced new solutions designed to help enterprises and service providers modernize their data centers for AI workloads, without adding complexity.
Also: Dell wants to be your one-stop shop for enterprise AI infrastructure
The core of today's announcements is Cisco Silicon One, the company's new unified, high-performance networking chip architecture. Silicon One is designed to power routing and switching across data centers, service providers, AI/machine learning clusters, and enterprise networks. Traditional approaches require different silicon for different network roles, such as routing versus switching, or service provider versus enterprise. According to Cisco, Silicon One delivers a single, flexible architecture that can be deployed across all these environments, simplifying design, deployment, and operations.
You'll find this architecture in Cisco's Nexus 9000 series switches, designed to connect the large clusters of GPUs that are essential for AI and machine learning. With the Nexus 9000s built to scale from tens to thousands of GPU clusters, the modular architecture allows you to expand your network as your AI demands grow, Cisco noted.
In addition, the next generation of switches will include Nvidia Spectrum application-specific integrated circuits (ASICs), which can work with the newest versions of Cisco's data center switch operating system, NX-OS.
This architecture, by the way, supports up to 51.2 terabits per second (Tbps) per device. Companies need this level of bandwidth in their data centers to deliver high-performance AI services.
Cisco also addresses the growing need for high-bandwidth, low-latency, and power-efficient networking by introducing 400 gigabit per second (Gbps) bidirectional optics. This enables you to upgrade your data center networks without replacing existing fiber infrastructure.
Also: AI's biggest threat isn't what you'd think - here's how to protect yourself
Cisco's offerings are flexible in other ways. Cisco's new Unified Fabric Experience with Nexus, for example, allows customers to streamline network operations across environments. It enables network administrators to merge disparate fabrics into a single, unified management dashboard. In particular, this will enable IT teams to manage LAN, storage area networks (SAN), Internet Protocol Fabric for Media (IPFM), and AI/ML networks from a single pane of glass, with the next release coming in July 2025
Cisco is also rolling out Intelligent Packet Flow, a feature that uses real-time telemetry to dynamically steer traffic across AI fabrics, optimizing performance and proactively detecting issues. The upcoming Cisco AI Assistant, integrated into the Nexus Dashboard, will simplify troubleshooting using natural language queries and intelligent recommendations.
Not sure what you'll need from your network for AI workloads? Cisco is also now offering Cisco Validated Designs (CVDs). These are pre-validated, optimized reference architectures for AI workloads, eliminating guesswork and speeding up deployment, Cisco said. These designs cover everything from hardware setup to software tuning, ensuring high performance and reproducibility.
Also: The best AI for coding (including a new winner - and what not to use)
In a related field, Cisco and Nvidia are building out the Secure AI Factory, which is designed to address AI's unique security challenges.
The Secure AI Factory is built on the networking side from Cisco Silicon One and Nvidia Spectrum-X. For the compute side, it relies on Cisco Unified Computing System (UCS) servers. These, in turn, are based on NVIDIA HGX and NVIDIA MGX platforms. Finally, the Cisco/NVIDIA Secure AI Factory uses Cisco AI Defense for the software.
Get the picture? Nvidia and Cisco are working to help businesses scale their AI ambitions, whether they are hyperscale builders, emerging neocloud providers, or enterprises modernizing their infrastructure. If you're serious about running your own AI infrastructure, consider Cisco's offerings.
Get the morning's top stories in your inbox each day with our Tech Today newsletter.