Nvidia Keeps Its Old Chips Selling Hot

Patience might be the one thing Nvidia can’t afford these days. Luckily for the chip maker, it doesn’t look like it will need to.

Ahead of Nvidia’s fiscal first-quarter results Wednesday afternoon, there was a growing fear among investors that major customers such as Google, Amazon.com and Microsoft would curb their orders for the latest generation of the company’s artificial-intelligence systems to await the launch of a new line later this year.

Nvidia previewed those new systems under the name Blackwell at its developers conference two months ago to great acclaim, but noted that they wouldn’t ship until near the end of this year. Hence, the worries about an “air pocket” in demand that might hurt near-term sales—and a stock price that has more than tripled in value over the past 12 months and made Nvidia worth nearly $2.4 trillion.

Nvidia’s report Wednesday managed to ease those fears. Revenue and operating earnings for the April quarter comfortably beat Wall Street’s estimates, as did the forecast for the period ending in July. Nvidia never projects financial results beyond the current quarter, but Chief Executive Officer Jensen Huang said he expects demand for the current series of chips known as Hopper to outstrip supply “for some time.” That is because the buyers of these systems are competing hard to build up generative AI capabilities and can’t really afford to wait for the next big thing. “Everybody is anxious to get their infrastructure online,” Huang said.

Chief Financial Officer Colette Kress drove the point home further, saying in an interview Wednesday: “One more quarter of you not training your [AI] model is just one more period of time that you are not able to get more efficient.” Nvidia’s largest customers—the aforementioned tech giants—also all notably ratcheted up their near-term capital spending forecasts in their latest quarterly reports, citing the need to invest more in AI infrastructure. Nvidia’s share price picked up 6% in after-hours trading Wednesday—passing the $1,000 mark for the first time.

Nvidia has become a very different company over the past year. Its fiscal first-quarter report a year ago was the first to make clear just how much demand for AI systems was going to remake its fortunes. Since then, Nvidia’s annual revenue has tripled to its current level just under $80 billion through the last quarter, and that is almost entirely on the back of its data center segment that sells those AI chips to cloud giants, enterprises and now even national governments trying to build their own sovereign AI networks. Data-center revenue hit $22.6 billion in the latest quarter—more than five times its size a year ago.

The ride is far from over. Wall Street expects Nvidia’s annual data center revenue to top $100 billion over the next year compared with nearly $66 billion now, according to FactSet estimates. Managing that growth will require Nvidia to keep its product transitions smooth, and deftly handle factors outside its control, such as the need for liquid cooling in data centers running the most advanced of the coming Blackwell lineup.

Mostly though, Nvidia’s biggest challenge may be managing the sky-high expectations that have come with the transformation of its business. That means good financial results might not always be good enough; Susquehanna analyst Christopher Rolland said in a note to clients Monday that Nvidia’s forecast would likely need to beat Wall Street’s target by at least $1.5 billion just to get a “flat reaction” for the stock.

Nvidia’s stock will also likely be sensitive to moves from competitors—including in-house chip efforts at its own largest customers. Those will likely just prove to be mere speed bumps, given the long lead the company has had to establish its position in AI chips and the necessary software. But even small speed bumps jostle when hit at high speeds.

Write to Dan Gallagher at [email protected]