Earlier this week, Chinese artificial intelligence startup DeepSeek sparked a sell-off of global tech stocks that lost the AI-driven stock rally $1 trillion in value.

Read more...

Meta (META) CEO Mark Zuckerberg told employees during a company-wide Q&A session on Thursday that “its going to be a crazy year.”

Read more...

Only rarely does a single company’s new product provoke a major market sell-off. But that’s exactly what happened on Monday, when a large language model from a Chinese company named DeepSeek drove the entire Nasdaq index of tech companies down more than 3 percent and shaved more than 17 percent off the market capitalization of the chipmaker Nvidia—which, until that moment, had been the most valuable company in the world.

The panicked selling of Nvidia had a surface logic. The company provides almost all of the computer chips (called GPUs) that companies such as Alphabet, OpenAI, Microsoft, and Meta rely on to train their LLMs. (The Atlantic entered into a corporate partnership with OpenAI in 2024.) Consequently, it has been the biggest beneficiary of the huge boom in corporate spending on AI that we’ve seen over the past few years. (Nvidia’s annual revenue has quadrupled since 2022.) Although DeepSeek also used Nvidia chips to train its model, the company said that they were an older type of GPU—U.S. export controls imposed by the Biden administration have prevented Chinese companies from buying cutting-edge chips. DeepSeek’s disclosure raises the possibility that future progress in training LLMs could be made with fewer, simpler chips, and at a lower cost than previously anticipated. That would obviously put a big dent in Nvidia’s profits. So investors dumped its stock.

[Read: The DeepSeek wake-up call]

If investors are very concerned about how DeepSeek might hurt chipmakers, they seem surprisingly unconcerned about how it might affect big AI software companies. Meta’s stock price, for instance, actually rose on Monday, and although the stocks of Alphabet and Microsoft did take a hit, they bounced back over the next couple of days. Some of that is because the underlying business of these companies, independent of AI, remains enormously profitable. But it also suggests that investors aren’t paying enough attention to the way DeepSeek’s success could disrupt the AI market, and in doing so threaten the future profits of the tech companies that are currently spending many billions of dollars every year on their LLMs.

Tech investors have historically profited by spotting the new new thing. But at the moment, they seem implicitly to assume that all of the fundamental change in the LLM business has already happened and that its future will look much like its present, with the companies that currently dominate the space—many of which are not simply competitors but also financial partners—continuing to do so indefinitely. What happened over the past week is a reminder that these assumptions may not be so solid.

The large language model that caused such a stir on Monday, DeepSeek-R1, is clearly comparable with LLMs such as ChatGPT o1-mini and Claude 3.5. Measured by industry benchmarks that rate subject knowledge, reasoning, and accuracy, the DeepSeek model seems to deliver similar performance while costing much less to develop—though just how much less remains a matter of debate. Beyond dispute is that it’s cheaper to use: Consumers can get access to DeepSeek’s core functions for free, and third-party developers are being charged a fraction of the cost of a product such as ChatGPT. DeepSeek also uses open-source technology, meaning that, in theory, you could download the program and run your own AI on your desktop if you had a powerful-enough computer. The fact that the LLM offers reasonable performance—results that, even a year ago, would have seemed startlingly good—at a significantly lower cost means that it has to be taken seriously as a competitor.

From one angle, in fact, DeepSeek looks like what the business-school professor Clayton Christensen, in his book The Innovator’s Dilemma, dubbed a “disruptive technology”: a product that’s less powerful than the products at the top of the market but also much cheaper, and that has the possibility of improving in quality over time to the point where it offers a superior combination of price and performance for most customers. In this regard, the rapid uptake of DeepSeek by users around the world has been striking. The LLM still has miles to go in market share to catch ChatGPT, which has more than 300 million weekly users, but since its release on January 20, its mobile-app version has been downloaded more than 3 million times from Google Play and Apple, making it the most popular app on both stores. That suggests that the cost of switching from one AI tool to another is very low, and that the moats big AI companies are building around their business may be much shallower than they’d hoped.

[Read: China’s DeepSeek surprise]

The underlying wager that these companies have made is that the big money they’re investing will result in radically better performance, which in turn will enable them to charge hefty sums to businesses and, to a lesser extent, consumers. (OpenAI, for instance, is reportedly targeting $100 billion in revenue by 2029.) And these companies remain committed to that bet. This week, the CEOs of both Microsoft and Meta said that enormous spending is essential to staying competitive in the market. Dario Amodei, a co-founder and the CEO of Anthropic (in which both Amazon and Google have invested heavily), wrote in a blog post that companies are going to continue to “spend more and more on training powerful AI models, even as … the cost of training a given level of model intelligence declines rapidly,” because “the economic value of training more and more intelligent models is so great.” In the long run, such investment may well result in the kind of performance improvement that a company like DeepSeek (which can’t even get access to the most powerful GPUs)—or the many other low-cost LLM developers that are sure to try to emulate it—cannot keep up with.

When you look at ordinary users’ embrace of DeepSeek, though, you can also see an alternative future. In this one, AI performance improves so much that most customers are happy with cheap, good-enough LLMs, and AI models end up as essentially interchangeable, commoditized products, with the small profits that always follow that type of commercial diffusion. We’re going to find out whether the great authors of the disruptive technology that’s transforming the business world might themselves get disrupted.

This is an edition of The Atlantic Daily, a newsletter that guides you through the biggest stories of the day, helps you discover new ideas, and recommends the best in culture. Sign up for it here.

Americans keeping close track of political news may have been toggling their screens today between Senate confirmation hearings: the second day of Robert F. Kennedy Jr.’s for secretary of Health and Human Services, and the first for Tulsi Gabbard’s for director of national intelligence and Kash Patel’s for FBI director. But each of those three hearings deserves the public’s full attention: Donald Trump’s nominees offered new glimpses into their approaches to policy, truth, and loyalty to the president.

Robert F. Kennedy Jr., Day Two

Ahead of Kennedy’s first day of hearings, our colleague Nicholas Florko noted that the HHS nominee is no stranger to conspiracist statements: “RFK Jr. has insinuated that an attempt to assassinate members of Congress via anthrax-laced mail in 2001 may have been a ‘false flag’ attack orchestrated by ‘someone in our government’ to gin up interest in the government preparing for potential biological weapon threats. He has claimed that COVID was ‘targeted to attack Caucasians and Black people,’ and that 5G is being used to ‘harvest our data and control our behavior.’ He has suggested that the use of antidepressants might be linked to mass shootings.”

“If Republican senators skirt around [Kennedy’s] falsehoods during today’s confirmation hearings,” Nicholas wrote, “it will be evidence of their prevailing capitulation to Trump. And it also may be a function of Kennedy’s rhetorical sleights … He’s capable of rattling off vaccine studies with the fluency of a virologist, which boosts his credibility, even though he’s freely misrepresenting reality.” But Kennedy’s sleights didn’t serve him quite as well today as he might have hoped.

At several points, senators encouraged Kennedy to acknowledge that vaccines are not the cause of autism, but instead of confirming what numerous studies have shown to be true, Kennedy insisted that he would need to “look at all the data” before coming to any conclusions. “The room went silent today during Senator [Bill] Cassidy’s closing questions,” Nicholas noted when we spoke this afternoon. “Cassidy was practically begging Kennedy to recant his previous statements on vaccines. Kennedy, like everyone else in the room, had to know this was a make-or-break moment for his confirmation. But despite the potential fallout, Kennedy refused, promising only that he would look at any studies presented to him disproving a link between vaccines and autism.”

The nominee for HHS secretary also showed, for the second day in a row, his lack of understanding about basics of the Medicare system, fumbling his answers to a series of rapid-fire questions from Senator Maggie Hassan, a Democrat from New Hampshire. Hassan also shared that she is the mother of a 36-year-old with cerebral palsy, and accused Kennedy of relitigating settled science on the fact that vaccines do not cause autism. “It’s the relitigating and rehashing and continuing to sow doubt so we can’t move forward, and it freezes us in place,” she argued.

Cassidy, whose vote could prove key to whether RFK Jr. is confirmed, said after today’s hearing that he is “struggling” over whether to confirm Kennedy.

Tulsi Gabbard

Gabbard came into her confirmation process with a history that raises questions about her commitment to national security (she has, among other things, met with former Syrian President Bashar al-Assad and expressed sympathetic views toward Russian President Vladimir Putin). As our colleague Tom Nichols wrote in November, “Gabbard has every right to her personal views, however inscrutable they may be. As a private citizen, she can apologize for Assad and Putin to her heart’s content. But as a security risk, Gabbard is a walking Christmas tree of warning lights. If she is nominated to be America’s top intelligence officer, that’s everyone’s business.”

The topic that ultimately received much attention in her confirmation hearing today was her refusal to say whether Edward Snowden is a traitor. Despite pressure from Democratic and Republican senators, Gabbard refused to answer the question, repeating that Snowden had broken the law and that she would take steps to make sure whistleblowers know how to properly make a complaint. Gabbard also revealed that she was unable to extract any concessions in her 2017 meeting with Assad. “I didn’t expect to,” she said.

Gabbard’s potential confirmation will depend on how her somewhat incoherent set of policy views sits with Republican senators. Last week, our colleague Elaine Godfrey explored the one through line—besides ambition—that has guided Gabbard’s otherwise inconsistent political career.

Kash Patel

Donald Trump is not always clear about what he means when he refers to “DEI,” but presumably it involves how someone’s identity is taken into consideration during the hiring process. In this morning’s press conference addressing the tragic plane crash last night, Trump asserted, without evidence but crediting his “common sense,” that DEI hiring at the Federal Aviation Administration was at fault.

It was odd, then, that a few hours later, Republican senators used Patel’s confirmation hearing to highlight his identity: Senator Lindsey Graham of South Carolina asked about examples of racism Patel has experienced, and Senator Mike Lee of Utah acknowledged the struggles Patel and his father must have faced as racial minorities in the United States and Uganda, respectively. Senator Thom Tillis of North Carolina, as if he were reading from a book report about the Gujarati people, lauded the religious diversity in Gujarat, India, where Patel’s family is originally from, omitting the state’s extreme tensions and violent history. Patel opened his own remarks by acknowledging his family’s journey from abroad. He invoked the phrase Jai Shri Krishna, a standard greeting for a sect of Hindus seeking blessings.

Patel was calm and still—he became riled up only when questioned by Senator Amy Klobuchar about his past suggestion that he would “shut down the FBI Hoover Building on day one and reopen it the next day as a museum of the ‘deep state.’” But he was walking a tightrope. Today’s hearing may be the rare instance when Patel has publicly broken with Trump, to whom he has otherwise been unequivocally loyal. He refused to explicitly state that Trump lost the 2020 election, but he also said, “I do not agree with the commutation of any sentence of any individual who committed violence against law enforcement.”

Overall, Patel seemed to be trying to carefully toe a line, answering questions about the culture-war issues that Trump and congressional Republicans care about—Senator Marsha Blackburn, for example, asserted during the hearing that the FBI prioritizes DEI and “counting Swiftie bracelets” over conducting investigations—while attempting not to alienate the employees he hopes to lead. Pressed by Blackburn, Patel made a vague statement about the “high standards” FBI employees must meet.

Related:

RFK Jr. has a lot to learn about Medicaid. What everyone gets wrong about Tulsi Gabbard

Here are four new stories from The Atlantic:

The memo that shocked the White House The near misses at airports have been telling us something. Donald Trump is just watching this crisis unfold. Jonathan Lemire: “What I saw at Trump’s first press conference”

Today’s News

Officials announced that there are no survivors in the crash last night between a U.S.-military Black Hawk helicopter and a regional American Airlines passenger jet landing at an airport near Washington, D.C. Three soldiers were aboard the helicopter, and 64 people were on the flight from Wichita, Kansas. Donald Trump appointed Christopher Rocheleau as the acting administrator of the Federal Aviation Administration. The agency had not had an administrator since the start of Trump’s new term. Eight hostages were released from Gaza by Hamas, and Israel released 110 Palestinian prisoners.

Dispatches

Time-Travel Thursdays: Don Peck interviews Nicholas Carr about h​​ow online life has rewired our brains.

Explore all of our newsletters here.

More From The Atlantic

“The ‘exciting business opportunity’ that ruined our lives” Trump’s war on meritocracy If Iranian assassins kill them, it will be Trump’s fault, Tom Nichols writes. Don’t politicize aviation safety. The return of snake oil Why Meta is paying $25 million to settle a Trump lawsuit

Evening Read

Illustration by Ben Kothe / The Atlantic. Source: Getty.

Your FOMO Is Trying to Tell You Something

By Faith Hill

I feel deeply haunted by the thought that if I don’t go to the party or the dinner or the coffee stroll, my one wild and precious life will be void of a joyful, transformative event—one I’d surely still be thinking about on my deathbed, a friend at my side tenderly holding my hand and whispering, Remember? That time we went bowling and the guy in the next lane over said that funny thing? Every year, my New Year’s resolution is to keep one night of the week free from social plans. Almost every week, I fail.

Read the full article.

Culture Break

Illustration by The Atlantic. Source: Jamie McCarthy / Getty.

Listen. In the latest episode of Radio Atlantic, the MSNBC host Chris Hayes speaks with Hanna Rosin about how bad the war for your attention has really gotten.

Read. The Stranger Things effect is coming for the novel, Mark Athitakis writes.

Play our daily crossword.

Stephanie Bai contributed to this newsletter.

When you buy a book using a link in this newsletter, we receive a commission. Thank you for supporting The Atlantic.

This story incorporates reporting from Yahoo, Barron’s on MSN.com and The Associated Press on MSN.com.

On January 30, several major stocks, including those of Tesla, Microsoft, Meta, and UPS, saw notable movements in the U.S. stock market.

Read more...

Read more...

The Dow and other major indices edged higher following earnings reports from tech giants Tesla (TSLA), Meta (META), and Microsoft (MSFT). However, AI stock Nvidia (NVDA) dropped another 2%, continuing its decline after it was hit hard by the launch of DeepSeek’s efficient-yet-thrifty AI model.

Read more...

This is Atlantic Intelligence, a newsletter in which our writers help you wrap your mind around artificial intelligence and a new machine age. Sign up here.

Earlier this week, almost overnight, the American tech industry entered a full-on panic. The latest version of DeepSeek, an AI model from a Chinese start-up of the same name, appeared to equal OpenAI’s most advanced program, o1. On Monday, DeepSeek overtook ChatGPT as the No. 1 free app on Apple’s mobile-app store in the United States.

So far, China has lagged the U.S. in the AI race. DeepSeek suggests that the country has gained significant ground: The chatbot was built more quickly and with less money than analogous models in the U.S., and also appears to use less computing power. Software developers using DeepSeek pay roughly 95 percent less per word than they do with OpenAI’s top model. One prominent AI executive wrote that DeepSeek was a “wake up call for America.” Because DeepSeek appears to be cheaper and more efficient than similarly capable American AI models, the tech industry’s enormous investments in computer chips and data centers have been thrown into doubt—so much that the top AI chipmaker, Nvidia, lost $600 billion in market value on Monday, the largest single-day drop in U.S. history. Sam Altman, the CEO of OpenAI, said that it was “invigorating to have a new competitor” and that, in response, the company would move up some new software announcements. (Yesterday morning, OpenAI said that it is investigating whether DeepSeek used ChatGPT outputs to train its own model.)

But many prominent American researchers and tech executives celebrated DeepSeek, as well. That’s because “the most notable feature of DeepSeek may be not that it is Chinese, but that it is relatively open,” I wrote on Monday. Whereas the top American AI labs at OpenAI, Google, and Anthropic have kept their technology top-secret, DeepSeek published an in-depth technical report and is allowing anybody to download and modify the program’s code. “Being democratic—in the sense of vesting power in software developers and users—is precisely what has made DeepSeek a success,” I wrote. Start-ups and researchers love this relative transparency. In theory, competitors can use DeepSeek’s code and research to rapidly catch up to OpenAI with far fewer resources—you might not need colossal data centers to get to the front of the AI race. (The Atlantic recently entered into a corporate partnership with OpenAI.) However, there’s substantial uncertainty about just how much cheaper DeepSeek was to build, based on reports about the start-up’s hardware acquisitions and uncertainty about how the model was trained.

Meanwhile, for national-security hawks, the fear is that an open-source program that won’t answer questions about the Tiananmen Square protests could become a global technological touchpoint. DeepSeek could face similar privacy concerns as TikTok: Already, the U.S. Navy has banned its use, citing security concerns.

Any predictions, for now, are highly speculative. The global AI race is far from over, and forthcoming products from Silicon Valley could leap ahead once again. At the very least, U.S. tech companies may have to reconsider whether the best way to build AI is by keeping their models a secret.

Illustration by The Atlantic. Source: Getty.

China’s DeepSeek Surprise

By Matteo Wong

One week ago, a new and formidable challenger for OpenAI’s throne emerged. A Chinese AI start-up, DeepSeek, launched a model that appeared to match the most powerful version of ChatGPT but, at least according to its creator, was a fraction of the cost to build. The program, called DeepSeek-R1, has incited plenty of concern: Ultrapowerful Chinese AI models are exactly what many leaders of American AI companies feared when they, and more recently President Donald Trump, have sounded alarms about a technological race between the United States and the People’s Republic of China. This is a “wake up call for America,” Alexandr Wang, the CEO of Scale AI, commented on social media.

But at the same time, many Americans—including much of the tech industry—appear to be lauding this Chinese AI. As of this morning, DeepSeek had overtaken ChatGPT as the top free application on Apple’s mobile-app store in the United States. Researchers, executives, and investors have been heaping on praise. The new DeepSeek model “is one of the most amazing and impressive breakthroughs I’ve ever seen,” the venture capitalist Marc Andreessen, an outspoken supporter of Trump, wrote on X. The program shows “the power of open research,” Yann LeCun, Meta’s chief AI scientist, wrote online.

Read the full article.

What to Read Next

The GPT era is already ending: “The release of o1, in particular, has provided the clearest glimpse yet at what sort of synthetic ‘intelligence’ the start-up and companies following its lead believe they are building,” I wrote in December. The new AI panic: “The obsession with frontier models has now collided with mounting panic about China, fully intertwining ideas for the models’ regulation with national-security concerns,” Karen Hao wrote in 2023.

P.S.

After several major tech executives announced their support for Donald Trump, many liberal internet users are now alleging that they are being censored on certain social-media platforms. “To some, this pattern was as unmistakable as it was malicious,” my colleague Kaitlyn Tiffany writes. “Social media was turning against Democrats.” And they are panicking.

— Matteo