Key Highlights

  • Nvidia (NASDAQ:NVDA) said networking Revenue exceeded USD 21 billion in the fiscal third quarter ended October 25, 2026.
  • Co-packaged optics switches are in Volume production, with one deployment connecting more than 10,000 GB300 GPUs.
  • Management expects total CPU revenue to more than double in fiscal 2028 as Vera CPU Demand outstrips supply.
  • Edge Computing revenue rose 42% to USD 8.2 billion, and physical AI is on track to exceed USD 25 billion this fiscal year.

Networking Becomes a Business in Its Own Right

Graphics processors dominate the headlines, but Nvidia (NASDAQ:NVDA) is building a large business in the hardware that links them together. Networking revenue exceeded USD 21 billion in the fiscal third quarter, a figure that on its own would rank among the largest semiconductor businesses in the world. Every large AI cluster needs high-speed connections inside each rack and across thousands of racks, and Nvidia sells both the switches and the interconnect technology. The company said its co-packaged optics switches, which place optical components directly beside the switch chip, are now in volume production. In one of the first production deployments, these switches connect more than 10,000 GB300 GPUs. Compared with conventional pluggable optics, the company claims about five times better power efficiency, roughly fifteen times better reliability and about 10% lower latency. Those gains matter because networking can consume a meaningful share of a data center's power budget, and link failures can stall large Training runs. By selling the full system, including compute, networking and software, Nvidia captures more revenue per Data Center and makes its platform harder for customers to replace with components from different suppliers.

CPUs Join the Growth Story

The second growth line outside GPUs is the central processor. Nvidia's Vera CPU is sold inside its Rubin rack systems and also as a standalone product. Management said demand for Vera is far outstripping supply, with adopters named on the call including CoreWeave, Meta, Nscale and OpenAI, adding to cloud customers announced the previous quarter. Total CPU revenue is expected to more than double in fiscal 2028. Agentic AI workloads help explain the demand. AI agents that call tools, search databases and run code generate a large volume of general-purpose computing tasks around each model query, and those tasks run on CPUs rather than accelerators. If agentic usage continues to grow, the ratio of CPUs to GPUs in AI data centers could rise, which would open a new market for Nvidia against established server processor makers. The company is also expanding its software reach through its pending Acquisition of Hugging Face, which serves more than 18 million developers and 200,000 companies. That deal would place Nvidia closer to the developers choosing which models and hardware to deploy, strengthening the link between its open models and its hardware.

Edge, Graphics and Physical AI

Nvidia now reports its business by market platform. Edge Computing, which includes workstation, robotics and automotive products, generated USD 8.2 billion in the quarter, up 42% from a year earlier and 14% from the prior quarter. Blackwell workstation sales were a key driver. The broader Graphics reporting segment, which includes gaming and professional visualisation products, produced USD 9.8 billion, up 62% from a year earlier. Physical AI, covering robotics, autonomous vehicles and industrial automation, has nearly tripled from a year earlier and is on track to exceed USD 25 billion for fiscal 2027. These businesses are small next to USD 107.6 billion of data center revenue, but they diversify Nvidia away from a handful of hyperscale buyers. For investors, the key question is whether networking, CPUs and edge products can keep growing if spending on large AI clusters slows. A broader mix would cushion the company in a downturn, and it also offers more ways to grow if AI adoption spreads into factories, vehicles and personal computers. Management's confidence in doubling CPU revenue suggests it sees that broadening already under way.

Where Things Stand

Networking above USD 21 billion, expected doubling of CPU revenue and a growing physical AI business show Nvidia expanding beyond accelerators. These lines remain smaller than data center GPUs but reduce reliance on a single product and a small group of buyers.