24/7 Wall St.: Nvidia's networking arm could be a $60B AI‑data business

Nvidia’s networking business has become large enough to stand on its own in the AI buildout, based on the company’s latest quarterly figures. Nvidia said on May 20 that data-center networking revenue reached a record $14.8 billion in its fiscal first quarter of 2027, up 199% from a year earlier and up 35% from the prior quarter. Annualized, that puts the segment near a $60 billion run rate, the figure highlighted in a May 23 article by 24/7 Wall St. That matters because Nvidia is usually discussed as a GPU company. The latest results show a second business inside the same AI data-center stack: the switches, fabrics and interconnect products that move data among chips and servers at scale. Nvidia’s first-quarter data-center compute revenue was $60.4 billion, while networking added another $14.8 billion. (investor.nvidia.com) ### What exactly counts as Nvidia’s networking business? Nvidia defines the category as data-center networking revenue, not gaming or automotive sales. In prior company commentary, that bucket has included InfiniBand, Ethernet products such as Spectrum-X, and NVLink-related fabric used to connect accelerated computing systems inside AI clusters. The products matter because large AI systems depend on more than processors. (investor.nvidia.com) Training and serving models across thousands of chips requires high-bandwidth, low-latency links between servers, racks and clusters. 24/7 Wall St. said that demand is rising as AI data-center expansion spreads globally. ### Why did the $60 billion figure get attention? The $60 billion number is not a separate company forecast. (investor.nvidia.com) It is an annualized run rate based on the $14.8 billion Nvidia reported for one quarter. Multiplying one quarter by four gives a figure a little above $59 billion, which 24/7 Wall St. rounded to about $60 billion. MarketWatch, citing the same earnings release, called the quarter a record for Nvidia’s networking business and said analysts had expected about $14.0 billion. (247wallst.com) The reported figure came in above that level. ### Why is Broadcom part of this story? Broadcom is one of the clearest comparison points because it also sells AI networking technology into data centers. 24/7 Wall St. framed Nvidia’s networking scale against Broadcom and other established infrastructure vendors as AI clusters get larger. (investor.nvidia.com) Broadcom’s own AI narrative also extends beyond custom chips. Recent coverage of Broadcom’s first-quarter results said AI networking contributed roughly $2.8 billion, with demand tied to Ethernet fabrics and switch silicon such as Tomahawk and Jericho. (marketwatch.com) That makes Broadcom a direct participant in the same buildout, even if the product mix differs from Nvidia’s. ### Why does networking rise with GPU demand? Nvidia’s first-quarter release showed the relationship in the numbers. (247wallst.com) Data-center compute revenue reached $60.4 billion, and networking revenue rose alongside it to $14.8 billion as customers built larger AI systems. The company has been pushing integrated systems rather than standalone chips. That means AI spending can flow not only into GPUs but also into the fabric connecting them, including Ethernet and InfiniBand deployments inside hyperscale and enterprise data centers. (edgen.tech) Prior Nvidia commentary cited Spectrum-X as a contributor to networking growth. ### What should investors watch next? Nvidia said its fiscal second-quarter 2027 revenue is expected to be about $91.0 billion, plus or minus 2%, though the company did not break out a separate networking forecast. (investor.nvidia.com) The next earnings report will show whether networking can sustain anything close to the first quarter’s pace. Broadcom’s next updates on AI revenue, including networking and custom silicon, will offer another benchmark for how much of the AI data-center budget is shifting beyond processors alone. (sdxcentral.com) Nvidia’s next quarterly filing and earnings call will provide the next official read on that comparison. (investor.nvidia.com)