Backbone of a beast – charting Nvidia's rise to networking titan

Nvidia's networking hardware now ships with a default configuration that is easier and faster to deploy than previous options.

Nvidia has emerged as a dominant force in the networking industry, surpassing Cisco in datacenter Ethernet sales. In the second quarter, Nvidia's networking revenue reached $17.5 billion, exceeding Cisco's total revenue of $17.3 billion for the same period. This rise is largely attributed to Nvidia's $6.9 billion acquisition of Mellanox in 2019. The acquisition enabled Nvidia to transition from a GPU manufacturer to a system builder, offering tightly integrated compute and networking solutions optimized for AI workloads. This integrated approach simplifies deployments and speeds time to market for customers like Recursion, a company focused on AI-driven drug discovery. Nvidia's networking revenue experienced significant growth inflection points in 2023, coinciding with the AI boom, and again in 2024-2025 with the rollout of Blackwell-generation GPUs and NVLink switches. The company's NVLink Fusion licensing program is further expanding its reach by allowing third-party chip designers to adopt its interconnect technology.

Nvidia's strategic integration of compute and networking, exemplified by the Mellanox acquisition and subsequent co-design efforts, has positioned it as a leader in AI infrastructure. This approach, focused on optimizing entire systems for specific workloads rather than individual components, addresses the growing demand for high-performance data movement in fields like AI-driven drug discovery. Companies are increasingly seeking end-to-end solutions that simplify complex deployments and reduce time-to-market, a gap Nvidia is effectively filling.

The expansion of Nvidia's NVLink Fusion technology to third-party chip designers signifies a broader ecosystem play. By licensing its interconnect technology, Nvidia is creating a new revenue stream and further solidifying its influence across the AI hardware landscape. This move could lead to more widespread adoption of its interconnect standards, potentially shaping the future architecture of AI clusters and high-performance computing environments.