Analysis: Nvidia's networking unit could be a hidden ~$60B business, 24/7 Wall St.

24/7 Wall St. said on May 23 that Nvidia’s networking unit may already be operating at a roughly $60 billion annualized pace, based on the company’s latest quarterly figures. The estimate rests on Nvidia’s disclosure that first-quarter fiscal 2027 networking revenue reached $15 billion, triple the level of a year earlier. Nvidia reported total first-quarter revenue of $81.6 billion and data-center revenue of $75.2 billion on May 20, according to its earnings release. That framing matters because Nvidia is still usually discussed as a GPU company, even as its data-center systems increasingly include the networking fabric that links clusters of AI chips. 24/7 Wall St. said the comparison now extends beyond traditional networking peers such as Arista and Cisco and toward Broadcom, whose AI business also includes networking exposure. Broadcom said on March 4 that first-quarter AI revenue was $8.4 billion, up 106% from a year earlier, driven by demand for custom AI accelerators and AI networking. (247wallst.com) ### Where does the $60 billion figure come from? Nvidia supplied the core math in its May 20 earnings materials when it disclosed $15 billion in first-quarter networking revenue. Annualizing that single quarter produces a run rate of about $60 billion, which is the figure highlighted by 24/7 Wall St. in its May 23 analysis. Nvidia did not separately guide full-year networking revenue in the materials surfaced here, so the $60 billion figure is an annualized run rate, not company guidance. (247wallst.com) May 20 was also the date Nvidia reported that overall revenue rose 85% from a year earlier to $81.6 billion. The same release said data-center revenue rose 92% to $75.2 billion, underscoring how much of Nvidia’s current business is tied to AI infrastructure rather than gaming or other legacy segments. ### What exactly sits inside Nvidia’s networking business? (247wallst.com) Nvidia has described networking as part of the infrastructure needed to connect AI systems at scale. Earlier company commentary tied networking growth to products including Spectrum-X Ethernet, InfiniBand and NVLink. In a prior quarter, CFO Colette Kress said Spectrum-X had delivered annualized revenue above $10 billion, according to reporting that cited Nvidia’s investor remarks. (investor.nvidia.com) Nvidia also said in October 2025 that Meta and Oracle were standardizing on Spectrum-X Ethernet switches for AI data-center networks. That announcement showed Nvidia pitching networking not as an accessory to GPUs, but as part of a full-stack AI data-center architecture. ### Why is Broadcom the comparison investors are making? (sdxcentral.com) Broadcom is one of the few large semiconductor companies with meaningful exposure to AI interconnects as well as AI compute silicon. Hock Tan, Broadcom’s chief executive, said first-quarter AI revenue of $8.4 billion was driven by “custom AI accelerators and AI networking,” and he said the company expected AI semiconductor revenue of $10.7 billion in the second quarter. (investor.nvidia.com) 24/7 Wall St. argued that Nvidia’s networking growth is large enough that investor comparisons may increasingly shift toward Broadcom rather than older networking incumbents. That is an inference about market framing from 24/7 Wall St., not a company statement from Nvidia or Broadcom. ### What should readers watch next in the filings? (investors.broadcom.com) Nvidia’s next hard datapoints will come from its fiscal 2027 quarterly materials, including the earnings release, presentation and any segment commentary posted to its investor relations site. Broadcom’s next update on AI networking and custom-chip demand is expected in its subsequent earnings materials on its investor site, after Hock Tan’s March 4 forecast for second-quarter AI semiconductor revenue of $10.7 billion. (247wallst.com) (investor.nvidia.com)