DealBook: Nvidia isn’t saving the day
Also, Kalshi gets tough on insider trading.
DealBook
February 26, 2026

Good morning. Andrew here. Nvidia’s windfall from the A.I. boom keeps growing — but is it enough to keep lifting the sector? We go behind the numbers.

And here’s a new wrinkle for prediction markets: Kalshi just brought its first enforcement actions over insider trading, which is most likely just the beginning. But that raises all sorts of questions about what constitutes inside trading on these platforms and whether anyone has the equivalent of a fiduciary duty. More below. (Was this newsletter forwarded to you? Sign up here.)

Jensen Huang, the C.E.O. of Nvidia, is seen wearing dark clothing as he speaks on stage with outstretched hands.
Jensen Huang’s Nvidia just blew away Wall Street estimates. But to investors’ chagrin, that performance has failed to move the market. Mikayla Whitmore for The New York Times

Tough crowd

Nvidia, the North Star of the artificial intelligence boom, delivered another knockout quarter. Yet shares in the chipmaker are up only modestly in premarket trading today, and the tech-heavy Nasdaq Composite index looks set to open lower.

What gives? In previous quarters, a big beat by Nvidia lifted the entire A.I. ecosystem. But investors appear to be acknowledging that the industry’s central player alone can no longer allay growing concerns about the sector.

Nvidia did its best to revive markets’ animal spirits. The company said that its most recent quarterly profit nearly doubled year-on-year, to $43 billion; revenue climbed 73 percent, to $68.1 billion. It has benefited from huge demand by tech giants, including Alphabet, Meta, Microsoft and OpenAI, for the chipmaker’s top-end A.I. processors.

Perhaps more important, Nvidia sees a bright future ahead, predicting $78 billion in revenue for its current quarter, beating analyst expectations. “A.I. is only going to get better from here,” Jensen Huang, the company’s C.E.O., told analysts yesterday.

Wall Street wants more reassurance about the A.I. boom. In recent months, investors have fretted about the sheer cost of the A.I. buildout. (The top hyperscalers have said they plan to spend about $650 billion this year alone on infrastructure like data centers.)

Several other factors are dampening investor fervor:

  • Supply constraints for chipmakers and other tech companies, particularly when it comes to components like memory chips. (Techies are calling it “RAMageddon.”)
  • Other costs for building out data centers, like power, may go up too, partly because of pressure from political leaders like President Trump.
  • Steeper competition for Nvidia, including from AMD — which just signed a huge chip supply deal with Meta — Google and others.
  • The growing worry among investors about the mind-boggling amount of debt tech companies are raising for their data centers. Private credit lenders, which are helping fund big projects for the likes of Meta, are already getting battered by A.I. fears around the software sector. (More on that below.)
  • The future of A.I. chip demand in China, once one of Nvidia’s biggest markets. Washington has allowed the company to sell only small numbers of its H200 chips, a lower-powered model earmarked specifically for the Chinese market, but it’s unclear how many sales Beijing will allow.

Some investors are betting against the A.I. boom, including by shorting shares of Oracle, a major partner to OpenAI, or the debt of hyperscalers. (Wagering against tech giants’ bonds is considered safer than trying to short their stocks, which could spike suddenly from a meme-stock frenzy, strategists told The Wall Street Journal.)

To be clear, Nvidia’s latest results are still impressive: The chipmaker’s gross margins topped 75 percent last quarter.

But it’s also worth noting that its free cash flow exceeds those of all of the tech titans save Apple. That’s great for Nvidia — but perhaps worrisome for its legions of customers that are hoarding debt to build out the A.I. ecosystem.

HERE’S WHAT’S HAPPENING

The Pentagon reportedly puts more pressure on Anthropic. The Defense Department has asked two major military contractors, Boeing and Lockheed Martin, about their ties to the artificial intelligence start-up Anthropic, according to Axios. Defense Secretary Pete Hegseth has threatened to declare it a supply-chain risk — which would require other contractors to cut ties to it — if the company doesn’t relax restrictions on the military’s use of its Claude software.

The Trump administration plans to raise tariffs again. Jamieson Greer, the U.S. trade representative, told Bloomberg that the U.S. would increase newly imposed global tariffs to 15 percent “where appropriate.” He added that the administration would seek “continuity” in existing trade deals with other countries, an effort to dispel uncertainty that arose after the Supreme Court struck down many of President Trump’s levies.

Here’s the latest Epstein files fallout. More prominent men will resign their posts after revelations about their ties to Jeffrey Epstein: Larry Summers, the former Treasury secretary, will leave Harvard at the end of the academic year; Borge Brende will step down as C.E.O. of the World Economic Forum; and Bob Kerrey, the former Nebraska senator, quit the board of a clean-energy start-up. The Times also reports that the Justice Department’s latest Epstein release didn’t include details related to a woman who made an accusation against Trump.

Jimmy Donaldson, the YouTuber known as MrBeast, in a dark suit, gray sweater and sneakers, sitting in a light chair in front of a DealBook Summit backdrop.
A video editor for the YouTuber known as MrBeast, pictured, was accused by Kalshi of insider trading on its platform. Karsten Moran for The New York Times

What’s behind the MrBeast insider trading case

Kalshi just announced its first major enforcement actions for insider trading. It punished a former candidate for governor and a video editor for the YouTuber known as MrBeast.

The fast-growing prediction market is seeking to show how seriously it takes accusations of unfair betting on its platform at a time when critics say Kalshi and its main rival, Polymarket, are awash with insiders taking advantage of those less in the know, Michael de la Merced writes.

Who was punished:

  • Kyle Langford, a former long-shot Republican candidate for governor of California, who Kalshi said traded about $200 on his own candidacy before posting about it on social media. That violated a Kalshi rule that prevents people with power to influence an event from trading on it.
  • Artem Kaptur, a video editor for MrBeast, who traded on material nonpublic information in contracts related to MrBeast content. That is, he appears to have bet on contracts about what might happen in one of the creator’s highly popular videos.

The punishments: Kalshi suspended Langford from trading on its platform for five years and fined him about $2,246. It suspended Kaptur for two years and fined him roughly $20,397.

Kalshi said they would be donated to a nonprofit group that helps educate consumers about derivatives market trading. Kalshi can drag those who won’t pay into arbitration.

The cases appear meant to send a zero-tolerance message, as skeptics of prediction markets argue that they’re rife with insider trading. This year alone, an anonymous bettor on Polymarket appeared to have profited from a contract on the capture of Nicolás Maduro of Venezuela — raising fears that a U.S. government official was involved. Another individual correctly predicted who would and wouldn’t perform during Bad Bunny’s Super Bowl halftime show performance.

Kalshi disclosed yesterday that it had opened about 200 investigations, with more than a dozen becoming active cases.

A big question: What will government authorities do now? Kalshi said it had referred the Langford and Kaptur cases to the Commodity Futures Trading Commission, which oversees prediction markets. That said, the C.F.T.C.’s standard for insider trading in commodities markets is looser than the S.E.C.’s better-known one for stocks.

Some lawmakers have also introduced legislation to bar some insiders from betting on prediction markets at all. Kalshi has endorsed one proposal.

Marc Benioff with blue World Economic Forum signage behind him.
Marc Benioff, the C.E.O. of Salesforce, has acquired 11 start-ups over the past 12 months to beef up his company’s agentic artificial intelligence. Yves Herman/Reuters

The “SaaSpocalypse” trade returns

Software stocks are under pressure in premarket trading after Salesforce spooked investors with an underwhelming sales forecast.

Salesforce and other software makers have been walloped by the artificial intelligence scare trade — “SaaSpocalypse” to many on Wall Street, referring to the software-as-a-service category — that has gripped global markets. But companies are looking to team up with or buy rivals to adapt to the changing marketplace, Niko Gallogly reports.

The latest: Anthropic, the A.I. start-up that’s seen as a prime disrupter, gave a presentation on Tuesday that highlighted new integrations of Claude Code, its autonomous coding tool, with several software heavyweights, including Intuit, LegalZoom and Slack, which is owned by Salesforce.

Don’t call it a comeback — yet. Salesforce’s stock fell 3.6 percent in premarket trading despite reporting solid sales growth for its most recent quarter.

In the face of A.I. uncertainty, software companies are turning to M.&A. Last year, the software sector spent $78 billion purchasing A.I. start-ups — more than five times as much as the $15 billion it spent the year before, according to PitchBook. In addition to bolstering their technology, the acquisitions provide software executives with a product to show to Wall Street investors eager to see A.I. innovation.

A look at the major software companies reporting earnings this week reveals the trend:

  • The workplace software company Workday spent more than $2.1 billion in acquisitions last year, including on the A.I. platforms Paradox and Sana. Those businesses are “accelerating agentic A.I. adoption,” Aneel Bhusri, a Workday founder who returned as C.E.O. this month, said on Tuesday’s earnings call.
  • Salesforce has announced 11 acquisitions over the past 12 months, according to PitchBook. The deals were aimed at beefing up the company’s agentic A.I. product, which accounted for $800 million in revenue last quarter. “SaaS just got a lot better with agents,” Marc Benioff, Salesforce’s C.E.O., told investors.
  • This month, the data cloud platform Snowflake acquired TensorStax, an A.I. data engineering platform. It was the company’s fifth acquisition in the past 12 months, according to PitchBook.

Investors aren’t fully convinced of the sector’s acquisition strategy, Jordan Klein, an analyst at Mizuho, told DealBook. Salesforce, Workday and Snowflake have all seen their stocks fall more than 20 percent since the start of the year.

“These acquisitions are defensive, they’re not offensive,” Klein said.

A picture of Revel's software interface, showing off mechanical hardware.
Revel pitches its software as a next-generation way for companies to test and control hardware like rocket engines and nuclear reactors. Revel Software

A big bet on software for hardware

Much of the excitement, and fear, in the software sector has to do with artificial intelligence models transforming enterprise products for sales, collaboration and more.

But Revel, a 16-month-old start-up founded by a veteran of SpaceX, is focused more on software to run hardware like rocket engines and nuclear reactors. That has drawn prominent investors, Michael de la Merced is first to report.

Revel has collected $150 million in its Series B round, the company plans to announce today. (The round values the start-up at just over $1 billion, according to two people with knowledge of the matter, who spoke on the condition of anonymity because they weren’t authorized to publicly discuss it.)

Index Ventures led the deal; other investors include Redpoint Ventures, Thrive Capital and Felicis.

Revel’s software controls complex industrial hardware. Its development was driven by the experiences that Scott Morton, the start-up’s founder, gained as the top software engineer for SpaceX’s enormous Starship spacecraft.

Morton and his team led efforts to develop custom software for SpaceX after being frustrated by existing options, such as the 40-year-old LabVIEW program.

The idea: Help companies develop their products faster. Dane Sarcone, the principal test engineer at Impulse Space, a spacecraft start-up, said that Revel’s software helped his company deploy an engine igniter on a test stand in about four days, down from two to three weeks.

Investors say it’s a field that A.I. probably won’t disrupt. Even as public-market investors fret over whether A.I. models from Anthropic and OpenAI can decimate whole software sectors, Revel and its backers say that’s less likely here.

“This is not something you can vibecode,” Alex Bard, a partner at Redpoint, told DealBook.

Revel is betting that it can accelerate manufacturing across the spectrum. Nina Achadjian, a partner at Index, said that speeding up customers’ testing processes means that they can get their products to market faster, and bolster productivity. Morton added that he planned for Revel to move beyond testing, too.

We hope you’ve enjoyed this newsletter, which is made possible through subscriber support. Subscribe to The New York Times.

THE SPEED READ

Deals

  • The investment giant General Atlantic is said to be planning the sale of some of its holdings in ByteDance, valuing the Chinese internet giant, which once owned the U.S. operations of TikTok, at $550 billion. (Reuters)
  • The London Stock Exchange Group will buy back up to £3 billion (about $4 billion) worth of stock, amid pressure from the activist investor Elliott. (FT)

Politics, policy and regulation

Best of the rest