Nvidia CEO Jensen Huang said he expects the artificial intelligence boom will last well into next year and made what could be the largest single bet yet in the tech sector to back up his optimism.

The company’s sales forecast on Wednesday blew past Wall Street’s expectations and it said it would buy back another US$25-billion of its shares, a move most companies make when their leadership thinks the company is undervalued. Nvidia’s stock price, though, has more than tripled this year and was set to hit an all-time high after Wednesday’s results.

Nvidia said it plans to ramp up production of its hardware into next year, quashing doubts that a few analysts had raised about how long the AI craze could last. The company has a near-monopoly on the computing systems used to power services like ChatGPT, OpenAI’s blockbuster generative AI chatbot.

“We have excellent visibility through the year and into next year, and we’re already planning the next-generation infrastructure with leading [cloud computing firms] and data centre builders,” Huang told investors on a conference call.

In an interview, Huang said two things are driving that demand: a switch from traditional data centres that were built around central processors to ones built around Nvidia’s powerful chips, and the rising use of content generated by AI systems in everything from legal contracts to marketing materials.

“These two fundamental trends are what’s behind everything that we’re seeing, and we’re about a quarter into it,” he said. “It’s hard to say how many quarters are ahead of us, but this fundamental shift is not going to end. This is not a one-quarter thing.”

Huang’s move to buy back stock when it is more expensive than it has ever been tops the bets that even other large tech companies are making on AI, but comes as its price-to-earnings multiple fell to about 43 from 60 after analysts upgraded their earnings estimates in May.

Nvidia’s big spenders

Microsoft said the $10.7-billion in capital spending it made in its fiscal fourth quarter — a large portion of which went toward Nvidia hardware — is a figure that would continue to rise. It has also invested $10-billion in OpenAI.

Meta Platforms, Amazon.com’s cloud computing unit AWS and others have also bet tens of billions of dollars collectively on AI-related hardware and products.

Demand for the chips has given Nvidia the cash for the investor payday. The company reported its adjusted gross margins nearly doubled to 71.2% in its second quarter, when most semiconductor companies have gross margins between 50% and 60%.

Kinngai Chan, an analyst at Summit Insights Group said Nvidia’s inventory of $4.3-billion is “light”.

“We think [Nvidia] will continue to beat the $16-billion guide for the October quarter as demand continues to outstrip supply,” Chan said, referring to the company’s third-quarter revenue outlook.

To be sure, some analysts don’t see unlimited demand. Dylan Patel of SemiAnalysis said many tech companies are spending heavily on Nvidia graphics processing units (GPUs) this year before determining how they will actually make money off products developed with those chips.

“They must over-invest in GPUs or risk missing the boat. At some point, the true use cases will shake out, and many of these players will stop investing, though others will likely continue accelerating investment,” Patel said.

Huang declined to comment on whether the AI boom will last past next year. He said the biggest risk Nvidia faces is securing supplies.

Read: The trillion-dollar man: who is Nvidia’s Jensen Huang?

The company said the biggest sales driver this quarter was its HGX system, which is an entire computer built around Nvidia chips. That system is much more complex than just the chip itself, and any missing piece can delay shipments.

“We’re getting great cooperation from our supply chain. And it’s a complicated supply chain,” Huang said. “People think it’s a GPU chip. But it’s a very complicated GPU system. It’s 70 pounds. It’s 35 000 components. It’s $200 000.” — Stephen Nellis and Max Cherney, with Chavi Mehta, (c) 2023 Reuters