By Adam Levine
The hot topic during Nvidia's Wednesday earnings call was...segment reporting. It may sound like an obscure topic, but CEO Jensen Huang used a change in the company's reporting style to bat back a bearish narrative. It's no doubt needed. Nvidia delivered another stellar earnings report, and the stock still slid.
The new disclosure separates data center sales among a handful of large customers known as hyperscalers -- like Amazon.com and Microsoft -- from everyone else.
The idea is to show a more diverse revenue base than many had supposed, and to convince investors that Nvidia doesn't need its largest customers to keep their capital expenditures growing at extraordinary rates in order for Nvidia to keep its even more extraordinary momentum. The first question during the company's earnings call was about the new accounting, and it kept popping up later in the Q&A.
If Nvidia was hoping the new segments would jump-start the stock, it isn't working yet. Shares were down nearly 3% in the two days following Nvidia's report. That may be because the nine quarters of data Nvidia released don't back Huang's argument. According to the new reporting, first-quarter revenue from hyperscalers grew 115% from last year, while sales to smaller customers was up 74%.
But Nvidia is pulling from a playbook that has worked before. In May 2024, the company began to use a table in its CFO commentary to get investors to focus on more than just sales of its graphics processing unit chips, the workhorses of the artificial-intelligence revolution in computing. Nvidia split the data center sales line between compute chips, which include GPUs and CPUs, and Nvidia's networking chips. Networking sales started out modest, but by last year they rose by 142%, to $31 billion, which made Nvidia the biggest data center networking-chip company in the world.
It took two years, but Huang made his point: Nvidia not only dominates computing in the AI data center, but in networking as well.
Now Nvidia is shifting its reporting, starting with the latest quarter, to meet a new narrative that has held back the stock: As goes Big Tech, so goes Nvidia.
It's pretty well known that Nvidia has five end customers that make up a significant portion of the downstream demand for the company's data center chips -- Amazon, Microsoft, Alphabet, Meta Platforms, and Oracle. Together they will likely represent about $750 billion in AI data center capex this year, up from around $400 billion the year before. Nvidia remains the primary beneficiary of all that spending.
At some point, investors figure Big Tech will reach its spending limits. These companies have huge operational cash flows, but now it isn't enough to cover the expenses, and they have started to dip into debt markets. Further spending growth in 2027 is likely to require more borrowing. For now, lenders are lined up around the block to back these loans, but that could change.
So, a question has persisted: How much of Nvidia's revenue is coming from just "public clouds and the world's largest consumer internet companies," as the company put it in its quarterly materials? And what happens to Nvidia's business if it all slows down or goes away?
We now know that in the first quarter, hyperscaler customers were responsible for about half of data center sales. But Huang is directing investors' attention over to the half called AI Clouds, Industrial, & Enterprise. This "everyone else" basket includes so-called neoclouds like CoreWeave, large corporations, factories, and, of increasing importance, governments.
"This second category is fairly poorly understood because there are just so many companies, and each one of the installations [is] relatively small compared to, of course, one of the hyperscalers," said Huang in the earnings call.
He told an analyst on the call, "We should be growing faster than hyperscale capex."
Huang is a masterful CEO, but he can't change the facts on the ground. And for now, the overall trend seems to be that hyperscale customers are growing in importance, not shrinking. In fiscal 2025, the hyperscale segment was 47% of data center sales, but that was 55% in the past 12 months.
But Huang may be looking ahead, the same way he did when the company started to report networking chip sales two years ago.
This is part of an overall pattern that has developed since Nvidia began dominating the chip market. Investors began to wonder whether it was all too good to be true, and started poking holes in the narrative. Ever the optimist, Huang has used his presentations -- to investors and developers -- to push back on the bear case. It has worked before.
For Nvidia's longest fans, the new segments carry a more significant message. After all, Nvidia was built originally on sales to gaming platforms. Its chips were always sought out for the most hardcore PC gamers. But by last year, gaming was just 7% of Nvidia sales.
The new reporting lumps Nvidia's gaming chips in with chips for cars, robots, video, and scientific workstations, which represented about 8% of Nvidia's first-quarter sales. It's definitely the end of an era for Nvidia.
Write to Adam Levine at adam.levine@barrons.com
To subscribe to Barron's, visit http://www.barrons.com/subscribe
(END) Dow Jones Newswires
May 22, 2026 21:31 ET (01:31 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments