NVIDIA Cements Its Role as the Backbone of AI Infrastructure

"Networking revenue grew 162% year over year to $8.2 billion. That is not a rounding error. It nearly tripled while the compute side of the Data Center business grew 56%. Networking is now growing at nearly three times the rate of the GPU business itself. This matters because networking is not just cables and switches. For NVIDIA, it means NVLink, InfiniBand, and Spectrum-X Ethernet."

"When a hyperscaler builds a gigawatt-scale AI data center, it needs GPUs, yes. But it also needs the high-speed fabric that lets those GPUs talk to each other fast enough to train and run frontier models. That fabric is increasingly NVIDIA's. The majority of AI deployments now include NVIDIA switches, with GPU attach rates roughly on par with InfiniBand."

"CEO Jensen Huang described NVIDIA as the only company with AI scale-up, scale-out, and scale-across platforms. That is not a marketing line. It reflects a real structural shift. Customers buying GB200 and GB300 systems are buying the networking stack alongside the compute. The two are bundled into the same infrastructure purchase."

NVIDIA's Q3 FY2026 results revealed a significant business transformation beyond headline GPU revenue growth. Networking revenue reached $8.2 billion with 162% year-over-year growth, nearly triple the 56% growth rate of compute-focused data center business. This networking segment encompasses NVLink, InfiniBand, and Spectrum-X Ethernet—high-speed interconnect systems essential for connecting thousands of GPUs in large-scale AI data centers. NVIDIA now bundles networking infrastructure with GPU systems, positioning itself as a complete platform provider rather than solely a chip manufacturer. CEO Jensen Huang emphasized NVIDIA's unique capability across scale-up, scale-out, and scale-across platforms, reflecting how hyperscalers purchasing advanced GPU systems simultaneously acquire the networking stack required for frontier AI model training and deployment.

Read at 24/7 Wall St.