Technology

Nvidia says it will sell more of its next-generation Blackwell chips than previously anticipated

In this article

NVIDIA founder, President and CEO Jensen Huang speaks about the future of artificial intelligence and its effect on energy consumption and production at the Bipartisan Policy Center on September 27, 2024 in Washington, DC.
Chip Somodevilla | Getty Images

After a quarter where Nvidia’s sales nearly doubled, investors and analysts are wondering how long the chipmaker can keep this kind of growth going now that it has a $140 billion annual revenue run rate.

Those hopes fall on Blackwell, which is Nvidia’s name for a family of server products based around its next-generation AI chip.

CEO Jensen Huang and CFO Colette Kress gave investors several new data points on how Blackwell’s launch is shaping up on a call with analysts on Wednesday. The duo emphasized that the rollout is on track, and they signaled that Blackwell sales over the next few quarters will be limited by how many chips and systems Nvidia can make, not how much it can sell.

“Blackwell production is in full steam,” Huang said. “We will deliver this quarter more Blackwells than we had previously estimated.”

The company’s positive comments on Blackwell are one reason why the stock is only down 1%, despite the company missing elevated expectations from bullish investors who anticipated Nvidia would significantly exceed its own forecasts.

Huang and Kress’s comments also addressed fears about shipment delays that were spurred by reports that said Nvidia was making ongoing engineering changes to its systems to address problems.

Some of Nvidia’s most important end-customers have already received some Blackwell chips, the company confirmed on Wednesday. Microsoft, Oracle and OpenAI have posted pictures of Blackwell-based server racks on their social media accounts, and on Wednesday, the company said 13,000 Blackwell chips have already been shipped to customers.

“There’s still a lot of a lot of engineering that happens at this point,” Huang said. “But as you see from all of the systems that are being stood up, Blackwell is in great shape.”

Those sample chips aren’t the bulk of the shipments that the company is expecting to make. They’re early versions intended to allow customers to start testing and get their systems and software ready for the volume shipments, which will start in Nvidia’s current quarter.

“We will we’ll ship more Blackwells next quarter than this [quarter], and we’ll ship more Blackwells the quarter after that than than our first quarter,” Huang said.

In July, Nvidia said it expected “several billion dollars” of Blackwell revenue in its current quarter, and on Wednesday, the company said it expects the amount of Blackwell sales for this quarter to be higher than its original forecast. Huang also said that Microsoft will soon start to preview its Blackwell-based systems to cloud customers.

A limiting factor to producing more Blackwell systems is the amount of components that Nvidia’s suppliers can provide, Huang said. Additionally, it takes time to ramp up the velocity of a manufacturing process that has gone from zero shipments to billions of dollars of shipments in a few months.

“It is the case that demand exceeds our supply, and that’s expected as we’re in the beginnings of this generative AI revolution,” Huang said.

He also named some of Nvidia’s “great partners,” including TSMC, Amphenol, Vertiv, SK Hynix and Micron.

“Almost every company in the world seems to be involved in our supply chain,” Huang said.

Nvidia said that Blackwell’s gross margins will be lower in the coming months than the 73.5% it reported in the third quarter, but the company said that margin will increase as the product matures. Huang pointed out that Blackwell comes as just the chip itself or in configurations that include an entire rack and other components.

Nvidia’s overall message on Wednesday was that its new Blackwell chip is in short supply because companies like OpenAI need the fastest GPUs available as quickly as possible to develop next-generation AI models. As Blackwell rolls out, Nvidia’s current AI chips, which it calls Hopper, will be relegated to serving AI models, not creating new ones. Nvidia said that Blackwell sales will eventually exceed those of Hopper.

“You see now that at the tail end of the last generation of foundation models, we’re at at about 100,000 Hoppers,” Huang said. “The next generation starts at 100,000 Blackwells.”

WATCH: Nvidia can still grow even with Amazon and Microsoft entering the space: Susquehanna’s Chris Rolland

Articles You May Like

Starmer to announce new green energy deal to protect from ‘whims of dictators like Putin’
‘Sean Combs is not an anomaly’: Former hip-hop dancer speaks out about the music industry
The areas where children are living longest in temporary accommodation
Strictly Come Dancing crowns its first blind winner
Politicians and royals need to be ‘extremely vigilant’ of Chinese spies