Cisco is expanding its existing partnership with Nvidia to capture a larger share of the AI infrastructure build-out underway at hyperscalers and large enterprises, Axios reported Monday in an exclusive. Specific contractual, technical and commercial terms of the expansion remain undisclosed.

The Axios report signals Cisco's intent to convert its enterprise networking installed base into rack-scale AI infrastructure sales, a category Cisco has identified as a faster-growing revenue line than its core campus and branch networking business in 2025.

What is changing?

The expansion extends a partnership that already ties Cisco UCS servers, Nexus switches and Silicon One ASICs into joint reference architectures with Nvidia GPUs, BlueField DPUs, Spectrum-X Ethernet switches and the Nvidia AI Enterprise software stack. Axios did not publish new product SKUs, pricing tiers or service-level changes in its initial report.

Why now?

Hyperscaler capex keeps hitting records. Microsoft, Alphabet, Amazon and Meta together guided roughly $250 billion in capital expenditure for calendar year 2025, the bulk earmarked for AI-optimised data-centre space, silicon, networking and power. Enterprises outside the hyperscaler tier are spending additional tens of billions on internal AI capacity, according to recurring IDC and Goldman Sachs surveys.

That capital pull has translated into record GPU shipments, double-digit year-over-year growth in Ethernet switch revenue and renewed demand for high-radix fabrics suited to large training clusters.

What does Cisco gain?

A deepened Cisco-Nvidia relationship converts Cisco's enterprise relationships into AI infrastructure revenue. Channel partners gain pre-validated reference architectures they can quote to mid-market and large-enterprise buyers lacking hyperscaler-scale engineering teams. Cisco's AI POD and Hyperfabric offerings already bundle compute, switching and orchestration for Nvidia GPU clusters; an expanded deal typically widens validation, support and financing terms.

What does Nvidia gain?

A wider Cisco engagement broadens Nvidia's OEM and integrator footprint and pushes Spectrum-X deeper into enterprise Ethernet environments historically dominated by Catalyst and Nexus gear. Spectrum-X and Nvidia's compute-networking software stack become available to thousands of Cisco-aligned resellers worldwide.

What comes next?

Cisco is likely to publish further details around its fiscal Q3 earnings release in mid-May, where AI infrastructure bookings and partnership milestones have become a recurring topic on the company's earnings calls. Nvidia's GTC conference in late March 2025 has historically produced OEM-side product launches that pull the Cisco partner channel into the spotlight.

For enterprise buyers and channel partners, the practical questions are validated configurations, bundle pricing and support tiers — none of which Cisco or Nvidia clarified in the first 24 hours of the Axios report.

Competitive impact

Dell, HPE, Supermicro and Lenovo all sell Nvidia-based AI servers and will be watching for tighter Cisco-Nvidia integration that could squeeze their enterprise accounts. For Arista and Juniper, a deeper Cisco-Nvidia partnership raises the bar on validated reference designs for AI workloads running over Ethernet rather than InfiniBand, the fabric Nvidia has used in its largest internal deployments.