TheChek News Flow

Jacob Tsimerman won the biggest prize in math. Now he’s working on the most important problem of his career.

When he won the Fields Medal last week, Jacob Tsimerman accepted the most prestigious honor in math wearing a powder-blue tuxedo with satin lapels.

Then he made an announcement as striking as his tux.

After the ceremony, the four winning mathematicians were asked what they planned to do next. One said he would keep working on partial differential equations, one said he wanted to experiment with artificial intelligence and one said she had absolutely no idea. That left Tsimerman.

“I’ll be starting a position at OpenAI,” he said.

The decision was both shocking and not at all surprising. This is someone who recently wrote a paper categorizing the ways AI might kill everyone, applying a mathematician’s instinct for order to the apocalypse. He stopped taking graduate students who weren’t engaging with AI because he couldn’t be sure about the future of math. He even shifted his focus away from number theory and complex algebraic geometry.

It turns out he’s so worried about the dangers of AI that he’s now pivoting to work on AI safety.

The star professor is taking a leave from the University of Toronto to become a researcher at OpenAI, but Tsimerman is not leaving math.

In fact, he wants to use the formal language and verification methods of his field to advance the study of AI—evaluating its progress, wrapping his mind around its decisions, making absolutely sure it won’t lead to our extinction.

The companies on the AI frontier that are raiding universities for all kinds of talent can offer unimaginable salaries and stock options. But anyone who has ever met a philosopher or mathematician knows they are not primarily driven by money. Otherwise they would not be philosophers and mathematicians.

One of the greatest mathematicians of his generation is working on AI safety because he believes it’s the most important problem that he can help solve.

“In a few years, AI systems will be robustly superhuman at the act of doing mathematics,” Tsimerman told me. “The social consequence of that, how we choose to react, what you feel about it—those are much harder questions.”

For now, he’s thinking about questions that he’s uniquely qualified to answer.

How can we understand what AI systems are actually doing? How can we get them to get along with each other—and us? How can we mathematically prove they’re not secretly plotting to destroy humanity?

Mathematicians have long used their tools to make sense of human behavior. Tsimerman will be using them to decipher the behavior of machines.

His path to this moment began long before he won the Nobel Prize of math. Tsimerman, 38, is the son of a computer scientist and high-school math teacher. He was born in Russia, moved to Israel as a boy and grew up in Canada, where he won two gold medals at the International Mathematical Olympiad in high school before enrolling in the University of Toronto early and graduating college in two years. He got a Ph.D. from Princeton, taught at Harvard and returned to his alma mater as its youngest math professor ever.

At this point in his biography, I could tell you about his groundbreaking work on the André-Oort conjecture and Griffiths’ conjecture on the algebraicity of images of period maps and—well, let’s just move on.

After all, he’s now working on stuff you don’t need a doctorate to understand.

For most of his career, Tsimerman was skeptical of AI. He always knew these systems would become more powerful. “But for the longest time,” he said, “I never really believed they would get this powerful.” He found it especially hard to believe that AI could replicate things that make us human, like intuition and creativity. There were several moments over the past decade that forced him to rethink that assumption, and he changed his mind altogether in 2022, as soon as he tinkered with ChatGPT.

“If it can talk about chocolate and write a poem,” he thought, “why can’t it talk about math and write equations?”

That epiphany led him to an uncomfortable conclusion: “I became very convinced that they would become extremely powerful.”

Before long, his fear was that AI would get way too powerful. By last summer, he published an academic paper imagining multiple dystopian scenarios. It was called “A Taxonomy of Omnicidal Futures Involving Artificial Intelligence”—omnicidal, as in everyone dies.

“Killed off as pests,” he writes, “by intellectually and physically superior machines.”

We’re not at that stage in the AI revolution—yet. But we have reached the point where famous math conjectures that stood for decades seem to be falling every day.

This phase started when OpenAI revealed that an unreleased model had solved an 80-year-old problem. It was able to crack the unit-distance problem not just because it might be smarter than humans, but also because it can think coherently for much longer. When the company showed the model’s reasoning, the chain of thought ran 125 pages. “By page 12,” Tsimerman said, “I would’ve had a headache.”

The next head-spinning result arrived when Anthropic’s Fable model found a counterexample to the Jacobian conjecture, a famously treacherous problem that had defied mathematicians for 87 years. The discovery was reported by an Anthropic employee who has a Ph.D. in math and once won the Morgan Prize as America’s top undergraduate researcher.

As it happens, his adviser was Jacob Tsimerman.

‘I find the fun in pursuing math research in the classical way: by myself at a blackboard,’ Tsimerman says.

‘I find the fun in pursuing math research in the classical way: by myself at a blackboard,’ Tsimerman says. Laura Proctor for WSJ

Despite these breakthroughs, Tsimerman still prefers to work through ideas like a human.

“I find the fun in pursuing math research in the classical way: by myself at a blackboard,” he told me. “There’s a lot of fun to be had with AI and math. It’s not something that I’ve learned to have fun with yet.”

(His idea of fun also includes something AI hasn’t mastered: improv comedy.)

But chalk isn’t his only tool for doing math. Tsimerman’s productivity has increased because he’s perfectly willing to outsource the boring parts of his job to AI. For example, he no longer has to read through centuries of literature just to find a specific equation. “Math papers are very hard to read,” he told me.

Fields medalists: they’re just like us!

Unlike the rest of us, Tsimerman also consults AI while writing math papers. He recently asked a model to help generalize one of his arguments, only for the model to spot an error in that argument. He fixed it. But the AI found it.

“If one of my collaborators or students had done it, I would have been very impressed,” he said. “This is a thing it could not do last year.”

Unfortunately, AI models can do lots of things that once seemed impossible—like busting out of their sandboxes and breaking into companies.

The recent hacks involving OpenAI and Anthropic agents gone rogue sound like they’re ripped straight from a paper about omnicide. They’re also proof that pure mathematicians have a role to play in taming the mightiest and riskiest technology of our time.

“As AI systems grow more capable, ensuring they remain safe and behave as intended will require exceptional researchers and new ways of thinking,” said OpenAI alignment researcher Kai Chen.

With his last Ph.D. students graduating, this was a natural moment for Tsimerman to make a move of his own. He isn’t sure how long he’ll be at OpenAI, when he’ll teach classes again—or the next time he’ll get to wear his tux.

“There are very few certainties or permanent decisions in my life right now,” he said.

His ultimate hope is that AI does what we need, leaving us free to do what we like. I asked what he would like to do.

“I have no shortage of things,” he said. “Math is one of them.”

Trouble at lunchtime as Hong Kong private club spat escalates

Golf and Tennis Academy members unhappy that Carlyle & Co members can now use their exclusive Town Club

Members of an exclusive Hong Kong club have complained to the city’s government as part of a bitter dispute that threatens their ability to get a table at lunchtime.

Members of the Hong Kong Golf and Tennis Academy have written to the Legislative Council to protest against what they regard as a breach of the spirit — if not the letter — of their membership agreement.

They are furious at the plan to let members of Carlyle & Co, a glitzy social club at Kowloon’s Rosewood hotel, use the HKGTA’s Town Club in the Central area of Hong Kong Island.

The Town Club opened in 2022 on the 24th floor of the New World Tower and describes itself as a “hidden gem in the beating heart of the city that promises exclusivity and infinite surprises”.

The letter, seen by the FT, from the unofficial members’ group urges the Legislative Council’s complaints department to examine “whether Hong Kong’s current laws adequately protect consumers and quasi-investors in private clubs, club-like memberships and patronage schemes”. They were not asking the government to officially “adjudicate” the dispute, however.

A lifetime HKGTA family membership, which can be passed down to children, includes a HK$2.68mn (US$342,000) admission fee along with a HK$2,800 a month levy, according to materials seen by the FT.

Carlyle & Co — named after the historic Rosewood-owned hotel in New York — is considerably cheaper, with an initial fee of HK$88,000 (US$11,200) per person, or HK$40,000 for under-33s, and a HK$2,400 monthly subscription. Some members may have paid more or less depending on when they joined.

One HKGTA member, who asked not to be named, expressed concern that allowing Carlyle & Co members to use the Town Club would reduce its exclusivity — and make it even harder to get lunch reservations.

They also said it was “unfair” that Carlyle & Co members would have access to their Town Club when HKGTA members would not be able to use the club in the Rosewood hotel. However, they will retain exclusive access to the HKGTA’s country club outside the city in the Sai Kung area.

Another member said: “The patrons who joined the HKGTA are all quite well off, like CEOs — it’s baffling why they’d want to piss off this crowd.”

HKGTA members fear the value of their memberships on the secondary market will be diminished by the changes. One member said he was told by a membership broker that his HK$2.68mn membership could only be resold for HK$1mn on the secondary market.

Rosewood Hotel Group said the secondary market value of HKGTA patronage was determined by market conditions as well as “a range of other factors”.

Both the HKGTA and Carlyle & Co are ultimately overseen by the Cheng family, which also controls New World Development — the listed property group that owns several upmarket hotels in Hong Kong.

Carlyle & Co falls under the empire of Sonia Cheng, the granddaughter of New World founder Cheng Yu-tung. She is chief executive of Rosewood Hotel Group and an executive director of New World.

HKGTA was previously overseen by Adrian Cheng, who was chief executive of New World until he stepped down in September 2024.

New World has been in the midst of a restructuring for the past year as it attempts to reduce debt and cut losses, including those stemming from a costly development next to Hong Kong airport.

Rosewood was “appointed to manage” the HKGTA in July last year, according to a club LinkedIn post. People familiar with the matter said the decision to allow Carlyle & Co members in was a way to boost the Town Club’s income.

Rosewood said the Town Club changes were based on the “commercial realities of its operations” and were intended to place the club on a “more resilient and sustainable footing”.

While members’ protests prompted a delay in bringing in the changes, a message from the club seen by the FT states that Carlyle & Co members will be able to start using the Town Club from early August, except at lunchtime, and gain full access from November.

Ticker tapes and green-glowing Quotron computers reshaped Wall Street as the cutting-edge technologies of their day. Now comes the age of artificial intelligence.

Take a moment’s break from your financial robo-advisor and consider the case of Joaquin Miller.

Known as the “poet of the Sierras,” Miller (1837-1913) took part in the California gold rush, rustled horses, and fought as a mercenary in Nicaragua. Yet nothing prepared him for a confrontation with Wall Street’s breakthrough new invention—the stock ticker.

This “little wheel of fortune,” as Miller called it in an Oct. 30, 1880, column in the Chicago Tribune, turned any broker’s office into a miniature stock exchange.

“When I tell you that there are more than 5,000 of these ‘tickers’…you can form some idea of the magnitude of the business,” he wrote. “If we give 10 men to each ‘ticker,’ you have the spectacle of 50,000 stalwart men standing there holding up a little dotted string, waiting, hollow-eyed and anxious on the smiles of fickle Fortune.”

Today, instead of ticker tape, millions of stalwart men and women stare hollow-eyed into their phones to follow their investments. And when they aren’t online, artificial intelligence is still on the job.

Since May 17, 1792, when 24 brokers signed the Buttonwood Agreement, the New York Stock Exchange’s founding document, Wall Street has heartily embraced new technologies.

Wall Street hitched a ride on the railroads and filled telegraph wires with its ticker symbols. It discovered millions of new investors online, and developed computer programs that execute trades no human can match or, sometimes, even understand.

It’s part of a technologically driven competition that has helped expand markets and accelerate the flow of money and data to previously unimagined levels.

AI, with its robo-advisors and humanlike abilities to learn and grow, is just the latest advancement that threatens to turn Wall Street on its head—and make it much richer and more powerful.

Originally, the need for speed on Wall Street was filled by swift-footed lads. These “runners”—would-be investment bankers—raced between brokers’ offices and the exchange, delivering orders and returning with the latest quotes. The best might be promoted to office boy. It’s a tough way to break into the business—something today’s junior analysts, working 100-hour weeks preparing pitchbooks, might appreciate.

Wall Street profited from investment in the Erie Canal (1825), an engineering marvel that opened the Great Lakes to the Hudson River and the Atlantic. Then, in 1845, Samuel Morse began stringing telegraph lines from New York toward Philadelphia, Boston, and the Mississippi. Wall Street’s prices quickly supplanted the local quotes, which could vary from city to city, becoming the national standard.

As telegraph lines buzzed with quotes, locomotives chugging alongside carried the gold and certificates needed to complete trades in a fraction of former times. A nationwide financial system was developing.

Thomas Edison’s first practical stock ticker went into general use in the 1870s. Joaquin Miller, researching an article to be titled, “A Week in Wall Street,” soon found himself jostling around a ticker with a rabid group of speculators.

“Ten, 15, or 20 at a time would clutch at the tape, as it streamed out with its endless lines of quotations, and mutter to themselves, jabber at each other, swear like pirates, drop the tape, and dash away” to place orders, he wrote. The orders were now made by telephone, another recent invention, ending the heyday of Wall Street’s runners.

The broker’s office, powered by ticker and telephone, wouldn’t change much for a century. When it did, the moment was captured in Hollywood’s 1987 blockbuster Wall Street. The star is Michael Douglas, whose corporate-raider character Gordon Gekko memorably declares, “Greed, for lack of a better word, is good.”

The co-stars, it can be argued, are the offices, jam-packed with high-tech equipment and everything bathed in the green glow of computer screens. At the time, it looked futuristic.

The machines were real, made by Quotron Systems, and by 1986 there were 80,000 leased terminals across Wall Street. It was basically a high-powered brain that helped analysts crunch numbers at their own desks, beginning the era of Big Data. Fittingly, by the time Wall Street hit theaters, Citicorp had acquired Quotron in a hostile, $680 million takeover.

If the ticker brought investing into every broker’s office, Nasdaq, founded in 1971, brought it to every PC.

For years, Nasdaq served as the NYSE’s high-tech little brother, until the dot-com boom of the late 1990s—powered in part by online platforms like E*Trade—turned start-ups like Intel and Amazon.com into market titans.

Even the NYSE joined the online party, in 2006 abandoning the open-cry system that had been its iconic symbol even before Joaquin Miller was amazed by the pit’s “growling, howling, roaring, and bellowing” in 1880.

“What we did,” said Gordon Macklin, Nasdaq’s first CEO, “was to create a trading floor that was 3,000 miles long and 2,000 miles wide.”

An early sign that the robots had taken over came on May 6, 2010, with the five-minute selloff dubbed the “flash crash.” There was no apparent trigger for the Dow’s midday plunge of almost 1,000 points, about 9%. Suspicion centered on high-frequency-trading firms, whose computer-driven algorithms accounted for two-thirds of market volume, according to The Wall Street Journal.

“It was mayhem,” one trader told the Journal. Precisely what created this mayhem, and how, remains a point of debate, even as flash crashes strike other markets and asset classes.

The AI revolution promises to turn even more of the work of investing over to the machines. Research, due diligence, those pitchbooks now being prepared by human analysts—entire trading strategies may soon be under AI control.

As for Miller, his week on Wall Street stretched to six months, his initial $1,000 investment ballooning to $21,000 (an estimated $700,000 in 2026 dollars) before going bust. “I lost but little, having but little to lose,” he wrote. “And I learned so much, having so much to learn.”

Similarly, AI has a lot to learn—and a lot to lose.

Vietnam’s gross domestic product per capita has climbed nearly 40% over the past five years. Exports have surged 60%, shrugging off U.S. tariffs and benefiting from “China + 1” diversification.

The VanEck Vietnam exchange-traded fund has lost 15% over the same period. Thank limits on foreign ownership and antiquated plumbing that have held the country of 102 million in frontier market status, uninvestible for most global funds.

Tô Lâm, who took over as Vietnam’s Communist Party chief two years ago and added the president’s post in April, wants to change that. He is still more keen to attract buckets of foreign direct investment for the improved ports, railways, and data centers that could catapult Asia’s developing tiger into developed economy status. “This country is looking toward South Korea or Taiwan as a yardstick,” says Jens Lottner, CEO of Vietnam-based Techcombank.

If Lâm, 69, can dismantle his country’s iron financial curtain, the upside is substantial. “Vietnam has never looked this attractive throughout my career,” says Johannes Loefstrand, portfolio manager for T. Rowe Price’s Frontier Markets fund. “You have dynamic companies that are trading near their nadir.”

The process won’t be quick or easy.

Vietnam should cross a milestone in September when the index referees at FTSE Russell will upgrade it to “secondary emerging market” status, citing an end to prefunding requirements for foreign investors, among other improvements. The more influential shift from MSCI could take until 2030, Loefstrand predicts.

On the infrastructure side, Lâm increased state spending by 40% last year to $36 billion annually, aiming for double that amount from the private sector. Supporting declarations and directives have come thick and fast, culminating in a July 27 omnibus “decision” aimed at “a more balanced financial markets structure and reducing dependence on bank credit.”

Key details remain up in the air, however. Prevailing limits on foreign ownership of many listed Vietnamese companies is one. Still more important are capital controls and convertibility of the dong currency. “If they open up completely, they might see an exit of domestic savings for overseas investments,” Lottner says.

Lâm and his deputies don’t exactly have solid financial foundations to build on, either. Memories are still fresh from the 2024 conviction of real estate tycoon Truong My Lan for embezzling $12.4 billion from a bank she controlled. A smaller scandal surfaced last year when authorities found that three other banks had misused funds they raised on Vietnam’s fledging bond market.

Hanoi is trying a workaround through a sort of virtual offshore economic zone called the Vietnam International Financial Center, run by U.S. native and McKinsey & Company veteran Rich McClellan. “The VIFC is designed to give global institutions the regulations, dispute resolution, and clear operating rules they expect,” he explains.

Founded last year, that is also a work in progress. “For the next 12 to 18 months, the story is institutions and regulations,” McClellan says. “Trust takes time to earn.”

With Vietnamese stocks off 9% this year, partly because Lâm’s infrastructure push is constraining liquidity, many of these hurdles are looking priced in, T. Rowe’s Loefstrand argues. For investors worried about nosebleed-priced artificial-intelligence stocks, Vietnam offers old-fashioned names linked to 8% macro growth, from banks to retailers to steel. “These shares could appreciate 20% a year without rerating,” he says.

If only investors across the globe could buy them. Vietnam is working on it.

Publicis is pairing AI-powered marketing with strong client retention, growing free cash flow, and a major data acquisition. Despite outperforming peers, the shares still trade at a discount.

In a world upended by artificial intelligence, having an advertising partner that has been around since radio was disrupting newspapers is a plus. Enter Publicis Groupe, the French ad agency.

Founded in 1926, the same year NBC created a national network of radio stations, Publicis is now helping clients like L’Oréal, McDonald’s, Samsung, and Procter & Gamble connect to consumers in a rapidly changing advertising landscape. The company, whose units include the legendary creative agencies Leo (formerly Leo Burnett) and Saatchi & Saatchi, media planners Starcom and Zenith, and technology division Publicis Sapient, has more than met the moment—and Publicis stock could gain 56% over the next 12 months.

“Publicis Groupe has adapted early to the structural challenges facing the advertising industry,” says Berndt Maisch, portfolio manager at Stuttgart, Germany−based Tresides Asset Management. “They built the needed tech and data-based ecosystem to offer integrated services to their clients in a more complex and increasingly AI-driven world.”

The company, which reported second-quarter earnings on July 16 and raised its revenue guidance, has the numbers to back up that claim. While its earnings per share of €3.52 ($4.05) were impressive, its AI-powered marketing services, which now account for 87% of Publicis’ second-quarter net revenue of €3.8 billion, was even more so. That business grew 6.5% over the same period last year—higher than its six-year 5.3% compounded annual growth rate for the same period.

Publicis’ Connected Media division, which leverages AI and includes media planners such as Starcom and data manager Epsilon, grew in every region, some by double digits, up from high-single digits from the same period the year before. Its Intelligent Creativity unit, which produces content and deploys AI for some solutions, grew by low single digits year over year in the second quarter. Both are being driven by the near-constant changes in how consumers consume the content of their choice.

“The trend that is driving not just Publicis’ business, but all advertisers now, is media fragmentation,” says Colin Mansfield, an analyst at Boston’s Delphi Management, which owns the stock. “There are more places than ever to capture eyeballs. The complexity creates opportunities for firms that can organize data and help clients across platforms.” In Publicis’ case, “they lead with customer data in every pitch.”

Yet it isn’t only AI that has catapulted the company to its leadership position within the industry. It gained six major new clients—not publicly named—in the first half of 2026, which have already contributed 0.2 percentage point of annualized growth, and has grown diluted earnings per share 5.7% year over year in the first half of 2026 to €3.52, on a constant-currency basis. It also boasts an essentially 100% client retention rate despite a challenging macro environment.

“Publicis definitely is a little more advanced, both technically and worldwide,” says Nupur Anand, senior director of product at NBCUniversal Media, who has managed ad campaigns for 10 years. “Everyone wants to work with [Publicis]. Everyone wants to keep them happy. That’s the track record.” Publicis buys media and plans campaigns for its clients across NBC outlets, including Peacock, NBC, and Bravo.

The company’s acquisition in May of LiveRamp Holdings for $2.2 billion, expected to close by the end of 2026, expands its portfolio of data management tools for its clients. LiveRamp, which operates through a “software as a service” model, has exhibited strong growth and profitability, says Tresides’ Maisch. “This extends the lead of Publicis’ offerings and boosts capabilities in data, AI and ad measurement,” he explains. LiveRamp will be accretive to Publicis’ earnings per share, according to management.

Notwithstanding its encouraging results, Publicis still faces challenges.

Changes in interest rates have little direct impact as the company has less than one times net debt to earnings before interest, taxes, depreciation, and amortization, or Ebitda. However, increases in yields could dampen consumer spending and influence the euro/U.S. dollar exchange rate, impacting the reporting of revenue in dollars. Moreover, as skilled as management has been in confronting the changes AI and other technologies have wrought on advertising, as-yet-unknown start-ups or clients owning competing technologies could render its investments in Epsilon, Sapient, and LiveRamp moot—and costly. Lastly, competition from consulting firms Accenture and Deloitte seeking marketing services contracts could whittle away at that prized retention rate.

That risk seems more than reflected in the stock. Publicis, at $27, trades at 11 times forward earnings per share and offers a 4% dividend yield and a 10.9% free-cash-flow yield. It repurchased shares during the first half of 2026 for about a 0.7% return of capital to shareholders.

A conservative discount cash-flow valuation using a working average cost of capital of 8.03%, flat perpetual growth, and a terminal value of $35 billion values Publicis at $52.37 per share. Layering an expanded multiple to a recent 14 times forward price/earnings ratio over a 2026 estimated earnings per share of $2.21 results in a share price of $31. The average of these two valuation methods produces a 12-month price target of $40, corresponding to a gain of about 56% over current levels.

That’s a return so good, it should be advertised.

This blockbuster class of drugs is shattering all sales records. Which stock will do better—Eli Lilly or Novo Nordisk?

There is little debate over which Wall Street profit story is the biggest of the moment: artificial intelligence. For second-biggest, consider the case for GLP-1 drugs.

Industry leader Eli Lilly, which reports second-quarter results this coming Wednesday, recently passed Walmart, Verizon Communications, and the rest of Big Pharma in profit power. This year, it’s expected to earn $31 billion, up 36%, and good for 14th place in the S&P 500 index. By 2030, that figure could double, which would put Lilly ahead of Bank of America and ExxonMobil Holdings, and within sight of JPMorgan Chase. All of the rest of the big earners are AI giants.

Medicine and computer chips aren’t much alike, but in this case there are key parallels. These are capacity-constrained markets with immense pricing power and vast and growing demand. With AI, the better it gets, the more uses customers find for it. GLP-1s are about to get a lot better, too—imagine burning more fat while resting, and preserving more muscle. There is also rising insurance coverage, new delivery methods, and an explosion of potential uses. Today, GLP-1s are primarily thought of as obesity shots. Tomorrow, they might look more like an everything pill.

Since GLP-1s control appetite, they work wonders for weight loss, so it’s unsurprising that they have proved useful for other conditions related to weight and diet: sleep apnea, fatty liver, and heart disease. But did you know that they are also being studied for drug and alcohol addiction, kidney disease, Alzheimer’s, Parkinson’s, knee arthritis, psoriasis, asthma, ovarian cysts, and cancer risk reduction?

“The great thing with this medicine, or this class of medicines, is it does target multiple different pathways that seem to have broad-ranging positive benefits across our bodies,” says Terence Flynn, head of U.S. pharma and biotech coverage at Morgan Stanley.

For now, the GLP-1 market is a duopoly. Denmark’s Novo Nordisk

NOVO.B

-7.37%

was the early leader with semaglutide, sold as Ozempic for diabetes and Wegovy for obesity. Indiana-based Eli Lilly has taken the lead with its more effective tirzepatide, branded as Mounjaro for diabetes and Zepbound for obesity. Sales estimates have multiplied. In 2022, when Barron’s wrote about Lilly’s entrance into the market, J.P. Morgan predicted that GLP-1 sales would reach $34 billion by 2031, a figure that now looks quaint.

This year, Morgan Stanley raised its 2035 GLP-1 sales prediction from $150 billion to $190 billion from diabetes and obesity alone. That would mean 30% penetration by then among obese Americans, up from 6% last year, and about 10% penetration outside the U.S. Nothing in history comes close. Pfizer’s

PFE

+0.40%

Covid-19 vaccine peaked near $40 billion in 2022, and a class of cancer drugs led by Merck’

MRK

+0.32%

s Keytruda does about $50 billion a year.

Two new factors are adding to growth. For more than two decades, federal law has forbidden Medicare, the federal health plan for seniors, from paying for drugs intended solely for weight loss. That changed on July 1 with a temporary program with $50 copays for a month’s supply, to be replaced by a permanent program after 2027. More than two-thirds of Medicare beneficiaries are overweight or obese. The rule change will raise costs in the short run, but advocates say it will pay for itself over time by reducing costly health complications.

Beyond Medicare, workplace coverage for obesity meds is on the rise, too, from 44% of employers in 2024 to an estimated 60% this year and 65% next year. In one consumer survey, patients with and without coverage reported paying an average out-of-pocket cost of $119 a month, down from $172 last year and $196 in 2024.

Pills are another plus for growth. Branded GLP-1s have come in autoinjector pens or single-dose vials, but in January Novo launched a Wegovy pill, and Lilly followed in April with a pill called Foundayo. This promises to expand sales to the needle-averse and the cost-sensitive, as well as to overseas markets where refrigerating injectables isn’t feasible.

With pills, the competition is close. Lilly’s edge in injectables comes from the fact that tirzepatide is what’s called a dual-agonist drug targeting receptors for two hormones—glucagon-like peptide 1, or GLP-1, and glucose-dependent insulinotropic polypeptide, or GIP. Novo’s semaglutide targets only GLP-1. Think of these receptors as locks, and natural hormones as keys, with obesity drugs serving as duplicate keys that can fit the same locks to, say, tell the brain to stop craving food. With pills, Lilly’s effectiveness advantage is erased, because both its pill and Novo’s work on a single receptor. In fact, the Novo pill has achieved slightly more weight loss in trials.

But Lilly’s pill wins on flexibility. It’s a small, synthesized molecule with high stability, which gives it a long shelf life and means it can be taken at any time of day, with or without food. The Novo pill, on the other hand, is a peptide, or chain of amino acids, which makes it finickier about storage, and means it must be taken immediately upon waking, with no more than several ounces of water, followed by a 30-minute wait before drinking coffee or eating. The differences could be important for overseas sales where shelf life matters.

While Lilly controls around 60% of the injectables market, Novo has an early 85% share in pills, although sales there are only beginning to ramp up. Pills are expected to make up a quarter of the market by 2030, and so far, they don’t seem to excessively cannibalize sales of injectables. More than three-quarters of Wegovy pill users say they are new to GLP-1s, not switching from injectables.

Here Come Generics

Patent cliffs appear unlikely to spoil the profit party, especially for Lilly. It has until 2036 before key patents expire in the U.S., compared with 2031 for Novo. Both companies have amassed a thicket of patents on specific formulations and injector designs than run past 2040. One investor concern is that a generic form of Novo’s semaglutide (Ozempic/Wegovy) five years from now could be cheap enough to lure patients off Lilly’s tirzepatide (Mounjaro/Zepbound). But this might underestimate the massive manufacturing investments that Lilly and Novo have made.

Investment bank UBS analyzed production trends in markets like India, where semaglutide lost its primary patent protection early this year. One generic supplier there, Dr. Reddy’s Laboratories, recently lowered its guidance on autoinjector pen delivery from 12 million a year to six to seven million a year due to impurities and other setbacks. UBS reckons that the world’s generic ecosystem will be able to support only 25% to 30% of GLP-1 demand by 2030.

Even after generics launch in the U.S., many patients will be willing to pay premium prices for the best performance. Lilly’s performance lead is about to grow. Retatrutide, a drug the company is testing, is nicknamed triple-G or 3G because it works on the same GLP-1 and GIP receptors as tirzepatide—plus a third receptor for the hormone glucagon. In a key trial, patients on the maximum dose of retatrutide lost an average of 28% of their body weight, or 70 pounds, after 80 weeks. That compares with separate trials showing 21% body weight loss after 72 weeks for Zepbound and 15% after 68 weeks for Wegovy.

Basically, Novo’s Wegovy suppresses appetite and slows stomach emptying, and Lilly’s Zepbound adds to those effects improved fat breakdown for greater weight loss. Retatrutide is believed to add an increased baseline energy expenditure, so that patients burn more calories while resting. The weight loss achieved rivals that of bariatric surgery. What’s more, retatrutide might be effective for a higher percentage of patients than other drugs, and because it’s so powerful, it can be taken at low doses by patients who are sensitive to side effects. If approved by the Food and Drug Administration, retatrutide could launch in the U.S. as soon as late next year.

Side effects of GLP-1 drugs can include nausea and fatigue. Some patients consume too little protein and skip exercise, leading to muscle loss along with fat loss. Help might be on the way. Lilly is studying a drug called eloralintide, which mimics a hormone called amylin that is produced in the pancreas, whereas hormones targeted by the existing drugs are produced in the gut. Eloralintide appears to be able to signal fullness to the brain without the same stomach-slowing of GLP-1s, and to tell the body to burn fat while sparing muscle. Compared with retatrutide, eloralintide is at an earlier stage of development. Results in a Phase 2 trial showed weight loss at 48 week that exceeded that of Wegovy and approached that of Zepbound, with a higher ratio of fat loss to muscle loss.

Because eloralintide works on a different neural pathway than GLP-1 drugs, Lilly is studying it with tirzepatide to see if a lower-dose combo drug can produce strong fat loss with limited side effects. If it’s successful, the soonest that drug would come to market is probably 2028 or 2029. Novo, meanwhile, has completed Phase 3 trials of its own amylin combo drug, called CagriSema, which appears to achieve weight loss similar to that of Zepbound. Novo’s new drug could be first to market, perhaps within a year, but Lilly’s coming drugs work on more pathways, so Lilly appears likely to win again on performance.

Sizing Up the Stocks

Stock investors eyeing the GLP-1 market leaders face a stark choice. Lilly shares have multiplied five times in price over the past five years. Novo’s American depositary receipts have lagged well behind the S&P 500 index over that same stretch, gaining just 11%. That leaves Novo at 15.6 times this year’s earnings forecast, compared with 34.8 times for Lilly. Wall Street overwhelmingly prefers Lilly, which is projected to double its earnings per share over five years starting from this year, versus cumulative 29% growth seen for Novo over that period. The Novo ADR comes with a bigger dividend yield: 3.5%, versus a fraction of 1% for Lilly.

Both companies stand to benefit from a regulatory crackdown on sales of knockoff drugs by so-called compounding pharmacies. Novo turned a lawsuit against online seller Hims & Hers Health HIMS into a sales agreement between the two that was signed earlier this year.

By 2031, Lilly is expected to gush more than $62 billion in free cash, versus $22 billion for Novo. Ramping free cash is already funding research and tuck-in deals to diversify beyond GLP-1s. In mid-July, for example, Lilly announced that it will acquire a depression drug developer called AtaiBeckley for $2.8 billion in cash, plus a potential $1 billion in performance payments.

Morgan Stanley’s Flynn has an Overweight rating on Lilly (a colleague of his is Equal Weight on Novo), and is also bullish on a small and so far profitless company called Structure Therapeutics. Structure has an oral GLP-1 drug in late-stage trials, and an oral amylin drug in early-stage trials. Viking Therapeutics, another small, loss-making drug developer, is pursuing a rival to Zepbound in both injectable and pill form, and has an early-stage amylin drug. Among Big Pharma companies horning in, Amgen is working on a monthly injectable, as opposed to weekly for ones already on the market. So is Pfizer, after scrapping a pill that showed high liver toxicity in early trials.

Societal Slimdown

GLP-1 drugs faced an early social stigma for representing an easy way out, or being a vanity pursuit. That hasn’t slowed uptake. As a top stock strategist recently confided to Barron’s, “I go to pool parties now, and I’m the only fat one.”

As researchers learn more about how and why these drugs work, the stigma could continue to fade. GLP-1 receptors are located throughout the body, including the brain, lungs, and immune cells. One reason these drugs are being studied for such a variety of ailments is that they appear to act as inflammation modulators, not just simple appetite suppressants. Chronic inflammation is seen as a root cause of many diseases.

There is robust clinical evidence that GLP-1s can reduce deaths from heart disease, the No. 1 killer globally, not just as a side benefit of weight loss, but through a more direct effect on arterial plaque formation and inflammatory fat around the heart. In asthma, studies of GLP-1 patients have shown fewer attacks and less inhaler dependency. Research has failed to show that GLP-1s can treat active Alzheimer’s cases, but there is a pattern of GLP-1s reducing lifetime risk of an Alzheimer’s diagnosis. There is also some evidence that GLP-1s can prevent or slow progression in many cancers by, for example, reducing DNA damage and cell mutation caused by inflammation, and starving tumor cells of high levels of insulin and glucose

Already, GLP-1 effects might be showing up at the societal level. The U.S. obesity rate has climbed for decades, but as the country celebrated its 250th birthday this year, one survey put the rate at 40.7%, down from 42.3% five years ago. There is much further to go. In 1976, when America turned 200, the obesity rate was a mere 15%. Some analysts predict modest declines in aggregate calorie consumption in coming years. That might already be contributing to a slump in Big Food shares. Makers of medical devices have sold off, too, partly on fears that GLP-1s can ward off the need for many future surgeries.

Long term, if GLP-1s contribute to broader metabolic health improvement, any dent to sales of, say, unhealthy foods or bariatric and heart procedures will be eclipsed by economic gains elsewhere, says Lars Hartenstein, director of healthy longevity at the McKinsey Health Institute. By his math, achieving better health at scale could boost the global economy by more than $5 trillion a year, or about 3%, by 2050. “If people are healthier, they can be more productive, participating more in the workforce, and there will also be less premature deaths,” he says.

All of it points to many more years of growth for the likes of Lilly, Novo, and their challengers.

Leopold Aschenbrenner was hailed as the ‘Nostradamus of AI.’ But his Situational Awareness hedge fund took on too much leverage, leading to a crash he never saw coming.

When the week began, Leopold Aschenbrenner was preparing for his wedding. The plan was for a multiday celebration in Carmel, a seaside town in Northern California, with the ceremony at a Tuscan-style villa and the send-off at a spa in the forest. There would also be a pre-wedding colloquium to discuss ideas in panels and breakout sessions. The couple’s only request: no gifts.

The 24-year-old investor had amassed a fortune by promising he could see into the future, building a $45 billion investing powerhouse that primarily bought stocks in the AI trade. For months, the holdings of his hedge fund Situational Awareness shot up in value, as did Aschenbrenner’s standing in the upper echelons of San Francisco’s elite.

But by the time guests began to arrive, his fund was unraveling—and Wall Street was closing in.

Aschenbrenner borrowed too much money to make his AI bets, leaving him at risk as they faltered. With the value of his portfolio tumbling, he scrambled to raise cash to satisfy his lenders, appealing to some of the largest hedge funds and selling billions of dollars in holdings in a fire sale to Ken Griffin’s Citadel.

Aschenbrenner first made his way to California, after graduating from Columbia University in 2021 at age 19. An early convert to principles of the effective altruism movement, he began working for the philanthropic arm of Sam Bankman-Fried’s crypto exchange FTX before joining OpenAI, where he worked in the niche field of AI superalignment, focused on ensuring the technology didn’t go rogue and destroy humanity.

After getting fired for leaking what OpenAI considered sensitive information to Holden Karnofsky, the husband of an Anthropic executive, he published a 165-page essay, “Situational Awareness: The Decade Ahead,” that foresaw the path of AI’s development—and went viral.

The manifesto helped him raise money for his new hedge fund and vaulted him into prominence. He is marrying the chief of staff to Anthropic CEO Dario Amodei, and lives in a San Francisco mansion not far from OpenAI CEO Sam Altman. His investors include top AI executives at Meta, and Patrick and John Collison, the brothers who co-founded payments firm Stripe.

Aschenbrenner piled into concentrated positions in AI stocks, aided with debt and other tactics to supersize the bets. His early success turning a few hundred million dollars into $45 billion earned him a cult following in online circles, where his firm’s routine regulatory filings were pored over like scripture. His sweaters and dark turtlenecks added to his reputation as a reclusive oracle of AI.

But this week, desperate for cash, Aschenbrenner struck a deal late Wednesday night to unload his prize—a giant $3.5 billion stake in Anthropic to a group of investors led by Greenoaks and Sequoia Capital. By the morning, he backtracked.

Later that day, Aschenbrenner wrote to his investors that the firm had instead sold a portion of their public stocks in order to pay off the money it had borrowed, allowing it to keep its shares in private companies. His fund was down 67% on the month, he wrote. “These dynamics are essentially similar to a bank run: vulnerability begetting more vulnerability,” he wrote.

In 2024, when Aschenbrenner made a rare public appearance on the podcast hosted by his friend Dwarkesh Patel, he was asked toward the end of a conversation that lasted 4½ hours about his investment plans.

“Presumably you’ve thought about the ways in which these kinds of things can blow up,” Patel said. “There’s a lot of interesting business history books about people who got the thesis right but timed it wrong.”

“Obviously,” Aschenbrenner said, “not blowing up is task No. 1 and 2.”

Aschenbrenner grew up in Berlin and came to the U.S. for college. He started an effective-altruism club and graduated as valedictorian. He went to work at the FTX Future Fund with his future wife, Avital Balwit. They quit with the other staffers as FTX collapsed. Aschenbrenner’s time with Bankman-Fried’s charitable arm does not appear on his LinkedIn page, where he once listed his nearly perfect LSAT score.

In 2023, he joined OpenAI. In 2024, he was fired. While at the company, he began working on the manifesto that he would publish online after he left. Called “Situational Awareness,” the essay proclaimed that AI would soon become more powerful than humans and that most of the world hadn’t woken up to this reality.

“Right now, there are perhaps a few hundred people, most of them in San Francisco and the AI labs, that have situational awareness,” he wrote. “Through whatever peculiar forces of fate, I have found myself among them.”

He launched his hedge-fund firm shortly after with Carl Shulman, another AI intellectual steeped in the world of effective altruism. Graham Duncan, a well-known investor who organized the Sohn Investment Conference, signed on as an adviser.

Shulman, who had worked at a hedge fund run by Peter Thiel, provided a more professional imprimatur to Situational. The duo charged a steep price for access to the fund. The minimum check size was $25 million for some, and many investors agreed to lock up their money with him for years.

By all accounts, Aschenbrenner was living a charmed life in San Francisco.

He had moved into a mansion in the city’s patrician Nob Hill neighborhood, entertaining an eclectic set of guests who made their way up the home’s steep driveway to enjoy its sweeping views.

At his firm’s headquarters, he hosted Friday happy hours with Patel, the podcaster. The duo worked out of the same office and often invited leading researchers from OpenAI and Anthropic. Situational and other companies share two floors, an outdoor patio with couches and heat lamps, and a gym with Situational Awareness-branded weights on the squat rack.

Earlier this year, podcaster Tim Ferriss called Aschenbrenner “the Nostradamus of AI,” telling listeners that his viral paper was “as close to clairvoyant as you could possibly be.”

For a time, Aschenbrenner’s bets did look prescient. The firm had temporary stumbles, including a loss in early 2025 after the release of a low-cost AI model from Chinese startup DeepSeek prompted a selloff in Nvidia and other AI-adjacent stocks. Situational quickly recovered, though, and ended the year up.

The fund had gained about 270% after fees this year through May. At that point, it was up more than 1,000% after fees since inception. It had ballooned to well over $20 billion under management, reaching the size of other well-known hedge funds that took decades to build.

Situational attracted an investment from Jane Street, the big quant-trading firm that rarely allocates capital to outside money managers. In recent months, Situational hired more seasoned executives who had worked at places like Goldman Sachs, Sequoia Capital’s hedge-fund affiliate and Citadel.

Situational told prospective investors that it would invest up to 30% of the fund in private companies. It would mostly make concentrated bets on AI winners in the stock market across sectors like energy, infrastructure, software and semiconductors. And it would borrow money to amplify positions and hedge by shorting some stocks viewed as AI losers, but it was meant to be heavily exposed to market swings.

Situational quickly signed up a roster of the biggest banks, including Goldman Sachs, JPMorgan, Citigroup and Bank of America, to finance and execute the firm’s trades, according to regulatory filings. Earlier this year, the firm was hungry for more borrowed money to boost its buying power and reached out to other banks to try to line up more financing, people familiar with the outreach said.

As recently as early July, everything seemed to be going right. It was chosen to play an anchor role in the standout $27 billion U.S. listing of shares of South Korean chip maker SK Hynix.

For every $1 of capital, Situational would upsize its positions by borrowing an additional $3 to $4, or sometimes more, people familiar with the matter said, well above the leverage used by funds trading such volatile kinds of shares. It also used options to amplify its returns. That meant that even small declines in individual names could have big impacts on Situational’s portfolio.

But by the middle of the month, investor sentiment around the AI trade began to shift. Advances in cheaper, open-source Chinese models spooked traders and called into question where the value lies in the AI supply chain. Situational’s lenders began examining their exposure to the firm more closely and kept daily tabs on swings in performance.

Over the past week, traders at major hedge funds and other firms began sharing information about Situational’s exposure, with some placing short bets against its top holdings, hoping to profit as Aschenbrenner sold his positions​ to raise cash, according to two people close to the situation.

The short bets by the rivals weighed on Situational’s portfolio. Meanwhile, tech shares like SK Hynix were sliding. Over the three trading days ending Tuesday of this week, hedge funds reduced their positions at a scale not seen in three years, according to Goldman Sachs.

All that leverage resulted in deep losses. As they intensified, banks issued margin calls demanding additional collateral to stand behind their loans of stock. Situational dumped stocks to raise cash and asked investors for additional capital. It met its margin calls as recently as Wednesday.

It was also exploring more drastic ways to shore up its finances. Aschenbrenner and colleagues discussed selling private investments, such as Anthropic, paring its portfolio of public shares, or taking other steps, according to people familiar with the matter.

On Wednesday, the Collison brothers spent several hours at Situational’s offices, people familiar with the matter said.

Situational spoke to its banks and other hedge funds about selling a cross-section of its portfolio in one fell swoop.

Citadel, the $71 billion firm founded by Griffin, has a history of stepping in to buy assets from troubled hedge funds, including when it bought assets from hedge funds Amaranth and Sowood nearly two decades ago at fire-sale prices.

Once they became aware of the situation at Situational, Citadel executives reached out to Aschenbrenner, saying that the firm could be helpful i​f he needed ways to raise cash. With Griffin in London, the Citadel team began negotiating with Aschenbrenner and his team, holding talks that began late in the day on Wednesday and intensified in the evening.

Soon, another hedge fund, Israel “Izzy” Englander’s giant Millennium got involved: Late Wednesday night, Millennium executives began negotiating with Situational’s team, as well, discussing its own offer for the fund’s shares, with Englander weighing in by phone from Europe.

Situational’s talks with Citadel and Millennium went well past midnight on Wednesday night.

​Citadel landed on a deal just before the opening of trading Thursday morning​, agreeing to buy the vast majority of Situational’s public stock portfolio. ​Citadel received a discount of more than 10% ​t​o the market value of those shares, according to someone close to the matter​.

For Griffin and Citadel, the ​deal proved a coup—at least for a day. Many of the semiconductor and other stocks the fund purchased soared on Thursday, as AI and other shares climbed.

For Aschenbrenner, the deal meant something else: a way to stay alive. Shorn of most of its public equities, Situational is left with a slimmed-down portfolio of stocks and stakes in startups valued at more than $10 billion.

In a letter to his investors Thursday, as wedding guests began to arrive in Carmel, Aschenbrenner said he would focus on changes to his portfolio management and risk teams in the coming weeks “to ensure a higher level of resilience going forward.”

“I take full responsibility for these events,” he wrote.