Nvidia is retaining its title as the darling of the A.I. revolution—at least for the time being. The Santa Clara, Calif.-based chipmaker reported record revenue yesterday (Nov. 20) for the July-September quarter that surpassed Wall Street’s already lofty expectations. Demand for Nvidia’s A.I. chips continues soaring, according to the company’s CEO Jensen Huang. “The age of A.I. is upon us and it’s large and diverse,” said Huang during an earnings call.
Nvidia’s quarterly revenue jumped 94 percent to $35 billion while net income more than doubled to $19.3 billion. Before the A.I. boom, Nvidia’s graphics processing units (GPUs) were most used by the gaming industry. Today, the company’s data center business, which serves mostly A.I. clients, accounts for nearly 90 percent of its total revenue, bringing in $30.7 billion for the quarter compared with $14.5 billion last year. Gaming revenue, meanwhile, totaled at $3.2 billion. The rest of the company’s sales came from its automotive and professional visualization departments.
Huang said demand for Nvidia’s new Blackwell GPUs, which began shipping this month, “is very strong” and production is “in full steam.” In recent months, Nvidia has struggled to meet demand for its Hopper chips.
Despite a blowout quarter, Nvidia’s share price slipped nearly 2 percent after the earnings release. Some investors were spooked by the chipmaker’s $37.5 billion revenue forecast for the fourth quarter, which would represent a slowdown in the rapid growth witnessed during past quarters. The “conservative revenue forecast” is “a touch disappointing,” said Kathleen Brooks, an analyst at broker XTB, in an investor note.
The company’s customer base is also increasingly concentrated. Cloud service providers, which include the likes of Amazon Web Services, accounted for around half of Nvidia’s data center revenue, compared with 45 percent during the same period a year ago. Nvidia’s reliance on this sector “could be a mild cause of concern” going forward, according to Brooks.
While some players in the tech industry have also raised concerns that A.I. model improvements are hitting a plateau, Huang remains optimistic about their scaling potential. “Pre-training scaling is intact and its continuing,” said the CEO, adding that test-time compute research has also emerged as an additional and promising method of scaling. “As a result of that, the demand for our infrastructure is really great.”
Nvidia expects the A.I. hype will only continue as society embraces the technology. Besides predicting that machine learning will eventually replace coding, Huang noted that data centers around the world are being modernized into “A.I. factories” that will generate A.I. “just like we’ve generated electricity.” The early foundations of an A.I.-driven world are already evident in developments like the adoption of A.I. agents across workplaces and breakthroughs across physical A.I., said Huang. “We expect this growth, this modernization, and the creation of a new industry to go on for several more years.”
