Good morning. Andrew here. The conversation we began yesterday about Anthropic’s potential computing constraints is continuing, as OpenAI unveils a new model that supporters say doesn’t suffer from a lack of resources. We look at what that might mean for businesses, customers and national security. Also, Apple blew away earnings expectations. But as it faces questions about its A.I. strategy, the company announced a new plan that could give it more cash to invest in the technology. More below. (Was this newsletter forwarded to you? Sign up here.)
The compute questionEarlier this week, many in the technology industry were asking whether OpenAI was in danger of falling behind in the artificial intelligence race. But the release of the company’s new model may change its fortunes — and is contributing to a wider debate about the importance of computing resources. Meet GPT-5.5-Cyber. The mouthful of a name is OpenAI’s newest cybersecurity model, which Sam Altman, the company’s C.E.O., said yesterday would be released to “critical cyber defenders” in a few days. Altman said OpenAI would work “with the entire ecosystem and the government” to figure out how to quickly roll out wider access for the model in the name of safeguarding global computing systems. “We want to rapidly help secure companies/infrastructure,” he added. It was an implicit rebuke to his rival Anthropic, even as Altman is doing something similar. Anthropic has restricted access to its Claude Mythos Preview model to several dozen companies and a growing number of U.S. federal agencies, arguing that Mythos’s ability to ferret out cybersecurity threats was a potential weapon that could be easily abused. David Sacks cited GPT-5.5-Cyber in a different critique of Anthropic. Sacks, the venture capitalist who served as A.I. czar in the current Trump administration, wrote on social media that GPT-5.5-Cyber “appears not to be token constrained,” unlike Mythos. (Worth noting: Sacks has a long-running beef with Anthropic.) Lack of computing power is becoming a big topic of discussion. The Wall Street Journal reported this week that the Trump administration was worried that Anthropic might not have enough resources to adequately power Mythos for both corporate and government users. Some users have already accused Anthropic of quietly downgrading the performance of its Claude chatbot to conserve compute resources. (An Anthropic executive said the company publicly disclosed the change.) OpenAI, which has been scrutinized for its sky-high investment plans for data centers, has said that running out of compute would be disastrous. Anthropic has criticized OpenAI’s approach, but is also planning major spending on A.I. infrastructure. Is this enough to change the outlook for OpenAI? It has reportedly missed internal business targets as many corporate customers have switched to Anthropic, citing better performance from its tools. But if the new OpenAI model has caught up in performance, and users now worry about computing resource constraints, Altman’s company may have gotten a new chance to retake the lead in the A.I. race.
Berkshire Hathaway prepares for its first annual meeting of the Greg Abel era. Abel, who succeeded Warren Buffett as C.E.O. this year, will preside over an event that has long been called “Woodstock for capitalists,” though attendance looks like it will be down slightly this year. Investors will want to know how Abel plans to lift Berkshire’s moribund stock price and how he might spend some of the company’s $373 billion in cash. The U.S. economy grows, despite a surge in energy prices. G.D.P. grew 2 percent in the first three months of the year, suggesting that private investment — especially on artificial intelligence infrastructure — consumer spending and government expenditures remained in solid shape before the war with Iran and in its first weeks. But economists think persistently higher fuel prices, with the average national gas price nearing $4.40 a gallon, could curb consumer spending. Oil giants’ profits climb in the early days of the war with Iran. Exxon Mobil and Chevron both beat analyst expectations for their first-quarter earnings, benefiting from higher prices for oil and natural gas in the early days of the fighting. (They both took hits from an accounting quirk related to hedging contracts.) Analysts expect the companies to continue to benefit from elevated energy prices, despite production outages related to the war. Senators bar themselves from trading on prediction markets. The Senate voted unanimously to immediately forbid lawmakers and their staff from betting on the platforms, amid growing concern that insiders have used information about classified government operations to profit from wagers on military actions in Venezuela and the Middle East. Congress has also long faced criticism over lawmaker investments in stocks.
The test for Apple’s next C.E.O.John Ternus made his first public comments as Apple’s incoming C.E.O. yesterday, during the company’s quarterly earnings call, telling Wall Street analysts that he planned to maintain the financial discipline that defined the tenure of his predecessor, Tim Cook. Ternus expressed enthusiasm about Apple’s prospects, as the company announced some of its best-ever financial results. But analysts raised questions about the company’s readiness for the artificial intelligence era. The highlights:
“This moment for the transition is the right one,” Cook told analysts during the call. Ternus said it was “the most exciting time in my 25-year career at Apple to be building products and services.” But analysts wanted to know about Apple’s A.I. plans, a big concern for the company given its stumbles with upgrading its Siri digital assistant and its struggles to develop in-house models. “The key is less about the March results and June guide, and more about Ternus’ ability to pull AI rabbits out of his hat,” Gene Munster, the managing partner at the investment firm Deepwater Asset Management, wrote on social media. Apple signaled it’s making a big move on that front: It will retain more cash instead of returning it to shareholders via stock buybacks and dividend payouts. The company has long been a cash-generation machine, holding $62 billion on its balance sheet even with its voluminous stock buybacks. That financial firepower could help pay for further investments in A.I. “We believe A.I. is a really important investment area for Apple,” Kevan Parekh, Apple’s C.F.O., told analysts. “We’re going to be doing that incrementally, on top of what we normally invest in our product road map.” The Epstein wedge in the Rothschild empireThe Jeffrey Epstein files continue to jolt the business world with their revelations of the convicted sex offender’s connections with global business leaders. Among these are Ariane de Rothschild, the billionaire C.E.O. of the bank Edmond de Rothschild. A police raid in March of a Paris townhouse belonging to the bank, over a former employee who had been an Epstein associate, put the focus on Ariane’s friendship with the disgraced financier. Edmond de Rothschild paid Epstein a $25 million consulting fee, and the C.E.O. had visited with her daughters on his island. (After Epstein’s arrest in 2019, the bank said Ariane had never met him, and she has denied knowledge of Epstein’s crimes.) The mess has also reignited tensions between Edmond de Rothschild and another part of the Rothschild empire, the investment bank Rothschild & Co., The Wall Street Journal reports: Now, the Epstein headlines were a concern for clients. Not to mention, it was personally unpleasant for family members to read Ariane’s unvarnished thoughts. In a 2017 email to Epstein that had just been made public, she called the cousins at Rothschild & Co. “a dead breed.” Alexandre de Rothschild, one of those cousins and the young CEO of Rothschild & Co., had a simple message for his teams, according to people familiar with the matter. Relationship managers should tell concerned patrons: Not us. The other ones.
Talking A.I. with the C.E.O. of DoximityEvery week, we’re asking a leader how he or she uses artificial intelligence. This week, Jeff Tangney, who leads Doximity, a maker of digital tools for physicians, told Sarah Kessler that he uses A.I. to digest customer feedback. The interview has been condensed and edited. How do you personally use A.I.? My favorite thing to do with A.I. around the house — I have three kids — is to use Google’s Nano Banana to do superhero images of them crushing their math exam or doing well on their track meet. What about at work? A.I. is very good at taking lots of text comments and sorting them into buckets and reporting that, say, 23 percent of the feedback you’re getting is about this. It’s great for market research and summarizing that research. I think there’s a future where instead of doing focus groups, we’re going to have a much more flexible model of folks giving video feedback to A.I., which distills it. What have you told your employees about how you want them to use A.I.? For our software engineers, we first did these “lunch and learns” where people would share what was working. But last quarter, we shifted to expecting that a quarter of all code committed should be done using A.I. It really makes sure everyone is pushing through the first few weeks to actually get past the productivity plateau. We hope you’ve enjoyed this newsletter, which is made possible through subscriber support. Subscribe to The New York Times.
Deals
Politics, policy and regulation
Best of the rest
Thanks for reading! We’ll see you tomorrow. We’d like your feedback. Please email thoughts and suggestions to dealbook@nytimes.com.
|
