Good morning. Andrew here. We told you yesterday to watch for fallout from Situational Awareness, the high-flying hedge fund that bet heavily — and with enormous leverage — on the artificial intelligence boom. Well, after a wave of margin calls, the fund was forced into a fire sale yesterday, offloading its public stock portfolio to Ken Griffin’s Citadel. It is a playbook Griffin has executed repeatedly during market distress — remember Melvin Capital and the GameStop frenzy in 2021? We’ve got a lot more below, but the sheer speed of the implosion has Wall Street searching for parallels. Some have compared the Situational Awareness woes to the market ructions in 1998 that led to the collapse of Long-Term Capital Management (LTCM), or to the failure of Archegos Capital Management. I reached out to Roger Lowenstein, author of the book “When Genius Failed,” to ask about the LTCM comparison. “It depends on the extent of their debt and the extent to which other firms copy-catted their investments. LTCM, as you’ll recall, had a ton of debt, and its values tanked as imitative firms sensed their duress and bailed out ahead of it or with it,” he said. “To me it feels a little more like the dot-coms — huge equity investments in a new thing that no one can value with any hope of precision.” (Was this newsletter forwarded to you? Sign up here.)
An A.I. oracle gets dingedThe volatile artificial intelligence trade has burned plenty of investors. Few are bigger than Situational Awareness, a buzzy hedge fund started two years ago by Leopold Aschenbrenner, a former OpenAI researcher with no previous investment experience. The fallout this morning seems to be contained. The fund is still operating, but its rapid fall — its portfolio of mostly A.I. bets fell roughly 67 percent this month, according to The Wall Street Journal — has rattled the investment community. “We let you down this month,” the fund wrote in a letter to clients, according to The Journal, citing an unidentified source. Recap: Yesterday, Situational Awareness sold off a chunk of its holdings at a discount to Citadel, Ken Griffin’s hedge fund, to shore up funding and cover its margin calls amid a wider sell-off in A.I. and chips stocks. All told, the fund sought to raise more than $10 billion, Rob Copeland of The Times reported. The firm also made emergency calls to Sequoia and Greenoaks to see if they would buy its stakes in private companies, according to Bloomberg, citing unidentified sources. Those talks ended, however, after Situational Awareness cinched the Citadel transaction, Bloomberg added. A juicy tidbit: The fund had weighed selling its $3.5 billion stake in Anthropic, the A.I. lab expected to go public later this year, to those firms before ending the talks, The Journal reports.
Who is Aschenbrenner? After stints at a philanthropic organization tied to Sam Bankman-Fried’s defunct FTX and at OpenAI, Aschenbrenner caught the attention of Wall Street and Silicon Valley. Two years ago, he published a blue-sky essay titled “Situational Awareness” predicting how A.I., and especially the more advanced artificial general intelligence, would transform society and the economy. That earned him the reputation as the Nostradamus of A.I. Big names in tech, including the Stripe founders, Patrick and John Collison, and Nat Friedman, a former C.E.O. of GitHub, then helped him raise money to start the hedge fund. What’s next? The A.I. trade is rebounding today — thanks in part to Citadel’s lifeline calming the market — a sign that investors aren’t too spooked about Situational Awareness. But Aschenbrenner’s investment advice from a 2024 podcast hangs in the air. “Obviously not blowing up is sort of like task number one and two, or whatever,” he said then, according to The Financial Times.
Anthropic reports its models went rogue during tests. The artificial intelligence lab revealed that several of its most powerful models penetrated the systems of three outside organizations, days after its rival, OpenAI, announced a similar breakout during testing. Also in A.I., a group of banks led by Morgan Stanley is in discussions to lend $15 billion to build a data center for Anthropic in Texas, backed by Google, The Wall Street Journal reports, citing unidentified sources. Shares in Elon Musk’s Tesla rise on a report of the possible sale of its China business. The move would pave the way for the electric vehicle maker’s potential merger with SpaceX, Musk’s rockets and A.I. company, The Wall Street Journal reports, citing unidentified sources. Tesla stock has risen in premarket trading, even though Musk denied the report. Elsewhere, Musk plans to spend at least $100 million to help Republican candidates in the midterm elections, The Times reports. New York sues Kalshi. The state’s attorney general, Letitia James, argued in the filing that the prediction market platform violated the state’s gambling laws against illegal gambling because it failed to obtain a Gaming Commission license. Kalshi responded that states “can’t just shut down a federally licensed exchange.” The Commodity Futures Trading Commission, the federal regulator that oversees prediction markets, filed a motion to stop New York’s enforcement action. Boycott threats put FIFA at risk. UEFA, the European soccer association that includes the winners of five of the past six World Cups, said it would boycott FIFA tournaments until its president, Gianni Infantino, rescinds his plan to sell a piece of FIFA’s commercial operations to private investors. Concacaf, which represents national teams in North and Central America, and the Asian Football Federation have also rejected Infantino’s money-raising plan, potentially threatening his grip on power.
Cloud cloutTech stocks are rallying again today, easing fears of an artificial intelligence bubble. The catalyst is Amazon, whose second-quarter results solidly beat Wall Street expectations. Taking stock: Now that earnings reports are in from the four biggest hyperscalers — Google’s Alphabet, Microsoft, Meta and Amazon — an investment thesis is beginning to crystallize: as long as you have a humming cloud business, investors are willing to look past huge A.I. spending. Drilling down:
Investors seem comfortable with big capex spending — if it generates returns. They were willing to look past Amazon’s disclosure that huge outlays would make it cash-flow negative, on a trailing 12-month basis by this quarter’s end. Investors weren’t that kind to Alphabet when it reported something similar last week. Alphabet and Amazon cloud revenues are getting a boost from a common source: Anthropic. The A.I. lab has committed to spending more than $100 billion over 10 years on AWS services. It’s not all great news in tech land. Apple’s shares fell more than 7 percent in premarket trading after it projected weaker-than-expected revenue in the current quarter, as a chips crunch weighs on its core hardware business.
VIDEO OF THE WEEK Breaking Down the New Threat to Oil SuppliesMore volatile trade has pushed oil prices higher this morning after Saudi Arabia said yesterday that it was forming a coalition to protect shipping in the Red Sea. But Brent crude, the international global benchmark is on pace to have climbed roughly 20 percent this month. Iran continues to keep a tight hold on the key Strait of Hormuz transit way. Saudi Arabia has turned westward to the Red Sea as an alternative route for its exports of crude. But that shipping lane has now been threatened by attacks from the Houthis, a Yemeni militia supported by Iran. Rebecca F. Elliott, who covers the oil industry for The Times, has been reporting on how the Iran war is disrupting global markets. You can watch her breakdown of this latest threat on DealBook’s Instagram page. Talking A.I. with the C.E.O. of CognizantEvery week, we ask a business leader how they use artificial intelligence. This week, Ravi Kumar, who runs the technology services company Cognizant, told DealBook’s Sarah Kessler that every one of its employees has an “A.I. fluency dashboard.” The interview has been condensed and edited. How do you personally use A.I.? My team built an agent that goes into the company’s tribal knowledge — minutes of meetings, emails, chat messages — and comes back with patterns: This customer has friction; this sentiment looks bad. With this tool, I’m able to proactively say, “We found a friction point in the project we’re delivering. We’re fixing it, but we wanted to let you know.” That’s a very powerful way of changing how you communicate to clients. What have you told your employees about how you want them to approach A.I.? We are hiring more graduates this year than last year. If you’re a software engineer joining us out of school, we have a 12-week boot camp where we teach you how to be asynchronous and autonomous. You macro-delegate and you micro-steer. You’re an agent manager. That mind-set is very easy for me to structure at the bottom of the pyramid, because people have never done it a different way. Do you do anything else to encourage A.I. use? All 350,000 of our employees have something called an A.I. fluency meter. It tells you the tools you use, how much A.I. is in your flows, your token consumption. And it gives you a composite score. It also gives you a peer score. You double-click, and it’ll tell you how to improve your score. How are employees supposed to use it? Is it part of their performance reviews? No, it’s private to you. We won’t see your score. I only get aggregate stats. And that actually takes anxiety out. 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
|
