Good morning. Andrew here. For those who think that the Trump administration’s designation of Anthropic as a security risk was another effort to coerce private companies to bend to its will, a federal judge agrees with you. In temporarily blocking the government’s Anthropic move, Judge Rita Lin in San Francisco wrote: “Nothing in the governing statute supports the Orwellian notion that an American company may be branded a potential adversary and saboteur of the U.S. for expressing disagreement with the government.” Corporate leaders may want to take note. The judicial system — all the way up to the Supreme Court — has frequently pushed back against the administration’s most aggressive tactics against businesses. Lin’s ruling might also persuade the Pentagon and Anthropic to return to the negotiating table. More below. (Was this newsletter forwarded to you? Sign up here.)
Taking stockWall Street is still struggling to gauge the fallout of the Middle East war nearly one month into the conflict. Fresh off one of the stock market’s worst trading days of the year, oil prices are rising again today, with analysts at Macquarie warning that Brent crude, the global benchmark, could reach $200 a barrel if the war lasts into June. Such an increase would wallop the global economy and corporate profits. The latest:
Higher fuel prices and borrowing costs are clouding the economic outlook. Mortgage rates in the U.S. have steadily risen since the war began. The Organization for Economic Cooperation and Development raised its inflation forecast yesterday and now expects it to average 4.2 percent this year. That’s well above the Fed’s projection, but the Fed governors Michael Barr and Lisa Cook warned yesterday that they were closely watching the war’s inflationary impact. Traders seem to be getting the message. The futures market this morning foresees the Fed’s next move as an interest-rate increase, in December. It’s the latest sign that Trump’s long-stated desire for lower borrowing costs is at odds with market expectations. Keep watching the Strait of Hormuz. Tehran has signaled it will grant safe passage through the strait to countries it deems friendly. That said, traffic through the vital shipping route, through which roughly 20 percent of the world’s exports of oil and natural gas pass, is at a virtual standstill. This is creating a huge supply disruption that could also slam food production. Through the first four weeks of the war, spikes in oil prices have been somewhat subdued as countries have been able to tap strategic reserves or supplies already at sea, Paola Rodriguez-Masiu, an oil analyst at the research and advisory firm Rystad Energy, wrote to investors yesterday. “That phase is now ending,” Rodriguez-Masiu added. With “inventories already drawing down, the system has shifted from buffered to fragile.”
The Senate passes a Homeland Security funding bill. The legislation was approved by voice vote early this morning, after Republicans acceded to Democrats’ demands not to fund ICE and Border Patrol amid uproar over the Trump administration’s immigration crackdown. If passed by the House and signed into law by President Trump, the bill could end the longest partial shutdown on record, which has led to huge security lines at airports as unpaid T.S.A. agents called in sick or quit. Trump’s signature will appear on the U.S. dollar. The Treasury Department said that this unprecedented move, which will have the president’s John Hancock appear along with Treasury Secretary Scott Bessent’s, was being made in honor of the country’s 250th anniversary. It’s the latest decision by Trump to emblazon his brand on national institutions. The makers of Absolut Vodka and Jack Daniel’s are in talks to combine. Pernod Ricard and Brown-Forman confirmed they were discussing a “merger of equals” that would create a drink giant with a combined value around $30 billion. The discussions come as the drink industry faces a broad downturn in alcohol consumption, driven in part by GLP-1 weight-loss drugs and the consumption of cannabis products. Anthropic takes Round 1The battle between the Pentagon and Anthropic over the government’s designation of the artificial intelligence start-up as a security risk is far from over. But Anthropic just scored a big early win in court, with a federal judge temporarily sparing the company from that debilitating label. The bigger question remains how long the fight will drag on, as Anthropic races to go public before year-end and tries to stay ahead of rivals like OpenAI and Google. Anthropic can keep its federal contracts for now, Judge Rita Lin of the Federal District Court for the Northern District of California ruled yesterday. In her order, the judge excoriated the government’s effort to classify the company as a supply chain risk after a breakdown in negotiations over limits on the use of Anthropic’s Claude software in classified military applications. Under the law, the supply-chain-risk designation would prohibit military contractors from using Anthropic software in their Pentagon-related work. The Trump administration went further than that, however, ordering all federal agencies to stop using Anthropic. From the ruling: The Department of War’s records show that it designated Anthropic as a supply chain risk because of its “hostile manner through the press.” Punishing Anthropic for bringing public scrutiny to the government’s contracting position is classic illegal First Amendment retaliation. The Pentagon offered “no legitimate basis” to assume that Anthropic might seek to sabotage the government, the judge added. The case is a reminder of the challenges that Anthropic faces. The company has become the darling of the A.I. world in recent months, racking up new customers eager to deploy its Claude coding tools. Anthropic was a primary driver of the so-called SaaSpocalypse that decimated many software, private equity and private credit stocks. But if the administration prevails in court, Anthropic could lose enterprise customers who must comply with the government directive to strip out its software. So far, it’s unclear how many companies have done that, other than military contractors removing Claude from their Pentagon-facing work. It’s just another complication for Anthropic, which is racing to stage an I.P.O. — reports suggest that could come as soon as October — and keep innovating in the face of competition, particularly from OpenAI and Google. (Anthropic told Fortune that it had started testing its newest model with some customers.)
Google jolts the chip marketShares in some memory-chip makers, including Samsung and SK Hynix, are down again this morning, after the sector tanked yesterday on news of a technological breakthrough from Google. It is the latest sign of Google’s growing influence in the semiconductor industry, Niko Gallogly reports. Shares in Google’s parent, Alphabet, were also down yesterday, by 3 percent. But that dip was probably in part because of the verdict against the company on Wednesday in a landmark case about social media addiction. Memory-chip companies had been on a roll, driven by robust demand from A.I. companies looking to train and run large language models. Google’s latest software innovation, known as the TurboQuant algorithm, promises to run models using one-sixth of the memory — prompting fears among investors about demand for such chips. Some analysts played down the potential impact. It makes “little difference to demand given the extreme supply constraints,” wrote an Ortus Advisors analyst, Andrew Jackson, according to Bloomberg. Google has been gaining ground in chips. The company’s tensor processing units, or TPUs, specifically target A.I. workflows. The tech giant’s progress is beginning to threaten Nvidia, whose chips have long dominated the chip market. Google’s TPUs are more energy-efficient, though less broadly applicable than Nvidia’s graphic processing units, or GPUs. TPUs are no longer a “nascent” alternative to Nvidia, Abhijoy Mitra, a founder of the venture capital firm CIV, which invests in A.I. companies, told DealBook. One of the most important proof points for Google’s success in the chip market is Anthropic, which has used TPUs to build out a highly compute-efficient business. Anthropic has bought $21 billion of TPUs. Nvidia has since responded by unveiling new chips that work more like Google’s.
Talking A.I. with the C.E.O. of AntithesisEvery week, we’re asking a leader how he or she uses artificial intelligence. This week, Will Wilson, who leads Antithesis, an autonomous-software testing company, told Sarah Kessler that he was teaching his kids to code the old-fashioned way. The interview has been condensed and edited. How do you personally use A.I.? My oldest daughter really, really likes programming, and she’s good at it. And so she and I have been writing a small computer game together. Vibecoding makes it easier to do the grungiest parts of coding and lets you focus more on the fun parts. That’s injected new life into this shared project. You’ve been a software engineer at Google and Apple. You obviously know how to code. But you’re choosing to vibecode? None of the A.I. companies want to say this, but it’s totally true: Vibecoding is most useful for people who are already really good at coding. You know when the A.I. is doing a good job, and you know when it’s doing a bad job. So given that, are you teaching your kids to code the old-fashioned way, or are you letting them vibecode? Maybe in six months the models will get better and none of this will matter. But how do you become a good programmer? By writing code the old-fashioned way, millions and millions of times, and bashing your head against all the problems. Late at night, if I’m adding some cool new feature to the game that I’m going to surprise my kids with, I’ll vibecode. But I don’t let them see that. What have you told your employees about how you want them to use A.I.? I’ve basically said to all of our engineers, Look, I don’t care if you generate the code by typing yourself or by asking Claude to type for you. It’s still your code. And so you need to understand every single detail of how it works before you ask another person to spend their time thinking about it or looking at it. This generalizes to a “no work slop” policy. You can’t just generate a zillion-paragraph document out of ChatGPT and email it to all your co-workers. You’re making an imposition on their time. 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
|
