Mellanox, acquired for about $7 billion in 2020, laid the strategic groundwork for Nvidia’s end-to-end networking ecosystem and bundled AI infrastructure solutions.

Nvidia is shifting from a pure chipmaker to a comprehensive AI infrastructure provider, with networking becoming a core pillar and a focus of investor attention as AI workloads grow.

At GTC 2026, Nvidia unveiled the Rubin AI supercomputer, the Inference Context Memory Storage, and upgraded Spectrum-X Ethernet photonics switches, expanding data-center AI capabilities.

The concept of 'AI factories' underpins Nvidia’s data-center networking, with NVLink, InfiniBand, Spectrum-X, and co-packaged optics forming the core of the AI factory data-center stack.

Nvidia pursues a full-stack, partner-delivered approach, bundling chips, networking systems, and software into integrated solutions rather than selling components in isolation.

The data-center networking division is rapidly becoming a major revenue driver, with quarterly revenue around $11 billion and annual figures surpassing $30 billion, second only to compute.

Executive messaging positions networking as the data center’s new unit of computing and a foundational element of AI infrastructure.

Industry observers note Nvidia’s networking business is exceptionally large and impactful, sometimes outpacing standalone networking peers.

Ongoing and upcoming GTC messaging emphasizes expanding platforms and next-generation AI supercomputing to reinforce leadership across the AI infrastructure stack.