Money Stuff: Who Should Control Anthropic?

Subscribe to Bloomberg.com for unlimited access to all our coverage.

Traditionally there are two main theories of corporate governance:

1. Shareholders are the owners of the firm, and they should have the final say in how it is run. Of course a board of directors and a chief executive officer run the company day to day, but the shareholders have the ultimate power to vote out the board and management if they’re doing a bad job.
2. Actually the shareholders are just suppliers of capital, and really the firm belongs to the visionary founders or their heirs. The company’s value comes from its long-term commitment to its mission, and you can’t trust outside shareholders — index funds, short-termist activist hedge funds, etc. — to be committed to that mission. The only way to preserve what makes the company special is to give the founders the final say over how it is run, even if they do not own most of its economic value. 

Both theories have their points, and my view is that each is appropriate for some companies. Even if you are an outside shareholder, you might rationally prefer that some of your companies be run entirely by their visionary founders, unconstrained by shareholders: The founders might just be better than the shareholders.

Recently, though, there is a third theory, mostly for giant artificial intelligence labs. The third theory is that AI is so powerful, transformative and potentially dangerous that AI firms must be run for the benefit of humanity. They shouldn’t be controlled by shareholders or by founders. They should be controlled by a wise body of philosophers with no economic stakes, whose only goal is maximizing the benefit of AI for humanity. Where you find these philosophers is an interesting question. Some AI theorists. Some effective altruists. Larry Summers. I don’t know.

The third theory was tested back in 2023, when OpenAI’s nonprofit board fired its visionary founder-CEO, Sam Altman, for like two days. That did not work! It turns out that if you are raising tens of billions of dollars from investors for a trillion-dollar business, you cannot actually give the final say over that business to a wise body of philosophers with no economic stake in the business. When I put it like that it seems obvious, but for a while OpenAI was pretty confusing.

Still, a big part of the pitch that the AI labs make to investors and employees is along the lines of “this stuff is too powerful to be left to index funds.” An AI lab that went public with single-class stock and no special control rights — one whose directors were just answerable to shareholders — would be suspicious; nobody would believe it was serious about AI. “AI is too important to be left to shareholders,” everyone thinks, even the shareholders.

The obvious approach — the one taken by SpaceX — is to fall back on founder control: You can’t trust shareholders to develop AI in a responsible way, and you can’t trust disinterested philosophers to maximize value, but you can definitely trust Elon Musk to do both. Arguably there are problems with this theory. 

I suppose the state-of-the-art answer is, like, both founder control and wise-disinterested-philosopher control? Not sure how that works, but the Information reported last week:

Anthropic has been preparing to give CEO Dario Amodei and other co-founders a class of stock with extra voting power to help insulate them from outside shareholder pressure, two people familiar with the matter said.

It would be the first time Anthropic’s leaders have extra voting power, which has become common practice for tech founders. It is all the more relevant for Anthropic because its co-founders hold relatively small ownership stakes in the company compared to other tech founders, another person said. The company also is planning to maintain its existing body of nonshareholder trustees with a special class of stock to elect the majority of members to the company’s board of directors, a more unconventional buffer against outside shareholder influence. …

And Anthropic has given significant but narrow governance powers to its Long-Term Benefit Trust, a group of advisers who aren’t employees or investors. The trust, created in 2023, has one major formal power today: It can elect the majority of Anthropic’s seven-person board of directors. The company granted it those powers through Class T shares that have no economic power.

These trustees include former Federal Reserve Chair Ben Bernanke. The trust is down to just three members, from its usual five, after Mariano-Florentino Cuéllar, a former member of the California Supreme Court, exited this month to become Anthropic’s chief global affairs officer.

Sure, I mean, I have to say, being a disinterested Anthropic philosopher-overlord sounds cool and perhaps crucial for the future of humanity, but being an Anthropic executive sounds more lucrative. Anyway I guess the hierarchy at Anthropic is:

1. The founders have the most control;
2. The disinterested philosophers have some control; and
3. The outside shareholders have no control.

Which seems roughly correct, possibly as a governance matter and certainly as a marketing matter.

Leveraged exchange-traded funds give investors (usually) two times the daily performance of some underlying stock or index. A 2x ETF on Strategy Inc., for instance, might take $100 of investor money, borrow $100 and buy $200 of Strategy stock. If the stock goes up 3% in a day, the ETF will have $206 worth of stock, giving its investors a 6% return on their money ($106 of equity). But now it is only 1.94x levered, [1] and to give investors the same 2x exposure the next day, the ETF will have to buy more stock: It needs to own $212 of stock to give investors 2x the daily return on their $106 of equity. So the ETF will buy another $6 worth of stock, typically at the close of trading. Similarly, if the underlying stock goes down, the ETF will have too much leverage, and will have to sell some stock to maintain its 2x ratio. 

We have talked about this before: Leveraged ETFs, by their design, buy stock when it goes up and sell it when it goes down, creating “volatility drag” for their investors. Here, though, are two points about this daily activity:

1. It is predictable: You can generally see how much money the leveraged ETF has, and you can see how much the underlying stock has gone up or down, so you can predict how much stock the leveraged ETF will need to buy or sell at the close. This means that there is a trade to be done. If you are a hedge fund, and you know at 3 p.m. that a leveraged ETF will need to buy $100 million worth of some stock at the close, you can buy $80 million of that stock now to sell to the ETF at the close. This is, in a sense, a short-term version of the index rebalancing trade: Leveraged ETFs have a lot of predictable demand for a stock at a fixed time (the close), and arbitrageurs can smooth out this demand by buying stock ahead of the close to deliver to the ETFs.
2. It is (predictably) variable: The amount of stock that the leveraged ETF will have to buy (or sell) is a function of how much the stock has gone up (or down). If you are a hedge fund, and you know at 3 p.m. that a leveraged ETF will need to buy $100 million worth of stock at the close, you can buy $80 million of that stock now, which will push up the price, which will mean that the ETF needs to buy, like, $120 million of it. So buy $40 million more! Now the price has gone up more, so you can buy even more, etc. 

There are a number of ways to characterize this trade. One is that it is a sort of liquidity provision trade: You know that there will be a big spike of demand at the close, so you buy stock efficiently during the day to deliver into the demand at the close, smoothing out stock prices just like index arbitrage does. If nobody did this, then there would be big price jumps in the closing auction, as the leveraged ETF would have to buy a ton of stock at the close with no one to sell it, and the ETF would have a harder time achieving its return target. (In fact most leveraged ETFs use swaps or other derivatives, rather than borrowing cash to buy the underlying stock, and the swap counterparty might hedge its own closing price risk by trading during the day. [2] ) 

Another characterization, which we discussed recently, is that it is an “intraday momentum” trade: Hedge funds and trading firms observe that some stocks exhibit a lot of intraday momentum (if they go up in the morning, they tend to go up more in the afternoon), so they buy those stocks when they go up and sell them at the close. This has the effect of buying stock during the day to deliver to ETFs at the close, but perhaps not the intent. Momentum traders might not be thinking about leveraged ETFs at all; they might not know that leveraged ETFs exist. They just know that when some stocks (Strategy, SK Hynix) go up during the day, they tend to close even higher, so they buy those stocks.

A third characterization is that it is, you know, “front-running” or “preying on” the leveraged ETFs: If you know they have to buy a lot at the close, and you know that the higher the stock goes the more they will have to buy, then you can go ahead and buy a lot of stock to push up the price and sell to them. You are transferring value from them to you; you are pushing up the price, knowing that they have to pay it.

Here’s “Preying on Leveraged ETFs,” by Yinhong Zhao of Princeton:

We argue that arbitrageurs preying on the closing rebalances of leveraged exchange-traded funds (LETFs) contributed to the Korean market's extreme volatility in 2026. An LETF's mandated daily rebalance is sized by the day's return, which generates an upward-sloping demand at market close. Arbitrageurs therefore pre-position, enlarge the fund's order, and liquidate into the demand they have induced. Consistent with this mechanism, Korean stocks tracked by LETFs reverse about 60% of their first-day response to pre-open U.S. news by the next close, a phenomenon not exhibited in any control groups. Our quantification implies that self-referential rebalance raised SK Hynix's annualized volatility from 84.8% to 136.7% over nine weeks and transferred 19% of terminal wealth from the products' predominantly retail holders.

From the paper: 

A large mandated purchase at the close implies a predictable closing return, allowing arbitrageurs to profit by buying earlier in the day. Their trades bring forward the price impact and further increase the required rebalance, causing the closing price to overshoot and subsequently reverse. The LETF thus buys high and sells low in each cycle. This mechanism compounds the familiar “volatility decay” of LETFs and transfers the incremental losses to the counterparties whose trading magnifies the price movements.

And:

No market exhibited a comparably large loop gain before Korea introduced single-stock LETFs in 2026. Their listing was the largest in Korean ETF history: combined assets began at KRW 4.3 trillion and reached KRW 14 trillion within three weeks. During the post-launch period, the worldwide mandated rebalance in SK Hynix averaged 22.4% of the stock’s total daily traded value and reached 50.4% on the peak day. The resulting loop gains have no U.S. counterpart. The values for the treated Korean stocks exceed that of MicroStrategy, the most extreme U.S. single-stock complex, by more than a factor of two and exceed those of the U.S. index complexes underlying the benign evidence by an order of magnitude.

One thing that you might take away from this is that moves in leveraged ETFs (and in their underlying stocks) are exaggerated: If the stock goes down during the day, this dynamic will push it down more, it will close below its “correct” value, and it will rally the next day. (“Korean stocks tracked by LETFs reverse about 60% of their first-day response to pre-open U.S. news by the next close.”) This suggests that there’s a lot of intraday momentum but also a lot of next-day reversion, so maybe you should buy the dip. The Financial Times reports:

Investors poured billions of dollars into risky leveraged funds tracking semiconductor companies during July and August, even as those products suffered heavy losses, as they tried to position for a possible rebound in the sector.

The biggest leveraged exchange traded fund tracking chip stocks — called Direxion Daily Semiconductor Bull 3X Shares — attracted almost $7bn of net inflows during July and the first two weeks of August, according to Morningstar data. …

“Buying the dip and selling the rip” has been a popular strategy in many leveraged ETFs, said Ben Snider, US equity strategist at Goldman Sachs, referring to how investors have been buying as the market falls and then selling when stocks rally.

Arguably leveraged ETFs help create the dips and rips, and can be used to reverse them.

Most insider trading cases are dumb, but it’s hard to know what that means. Perhaps it means that most insider traders are dumb: A junior banker texts his buddy about a deal he’s working on, the buddy buys short-dated out-of-the-money call options on the target, the deal is announced, the buddy makes like $50,000, he texts the banker “hey bro thanks for the tip where should i send your bag of cash ;)” and the banker replies “stop texting about this, i do not want to go to prison for this criminal insider trading we are doing lol.” 

Or perhaps it means that most insider traders are smart, and only the dumb ones get caught. If you have never bought options before, and suddenly you put all your money into short-dated out-of-the-money call options on a merger target, probably an alarm bell rings at the US Securities and Exchange Commission, and they show up and ask to look at your phone, which is inevitably full of texts with a banker on the deal. But maybe you don’t do that. Maybe you regularly get insider tips in person and trade on them and make lots of money and don’t get caught because (1) you do a lot of legitimate trading, so your successful insider trades don’t look suspicious and (2) you don’t put the crimes in writing.

Anyway here’s an SEC case from last week against Gavin Wolfe, a retired investment banker who allegedly made $18.5 million insider trading on a merger, and Jason Satsky, his former colleague who worked on the merger and allegedly tipped him. Here is the complaint. What is interesting here is that:

• These guys were not junior analysts: Satsky was global head of energy and utility investment banking at Bank of America, and Wolfe was also a managing director at BofA before he retired to “manag[e] his own investments and operat[e] businesses that he owns or controls.”
• They did large size: Wolfe ran a portfolio “valued at approximately $260 million,” and bought $53 million of stock (not call options!) in the merger target, making $18.5 million of profit when the merger was announced.
• They did not go around emailing about it: The SEC doesn’t quote any messages between them before Wolfe bought the stock, [3] and there is no direct evidence that Satsky told Wolfe anything at all. Instead, the evidence is that “the two men attended a nationally televised [Duke] basketball game together at Madison Square Garden,” [4] and that “Within minutes of the November 9 basketball game ending, after spending approximately three hours with Satsky at Madison Square Garden, Wolfe created a calendar entry for himself at 12:12 a.m. that read ‘SJi and njr,’ using the New York Stock Exchange ticker symbols for South Jersey [Industries Inc., the merger target] and another company. He scheduled the entry for 9:15 a.m. that morning, November 10, 2021.” And then in fact he started buying the target’s stock.

One possibility here is that these are careful experienced professionals who insider traded without creating dumb obvious evidence, and were nonetheless caught by the SEC’s increasingly sophisticated enforcement apparatus. Another possibility is of course that they are just friends who watched some basketball together, and that a former utility investment banker running a $260 million personal portfolio might have had his own reasons for investing in a local utility company. Bloomberg’s Nicola White and Ava Benny-Morrison report:

In a Friday statement, Satsky’s lawyer, Bob Anello, denied the allegations. “The enforcement action brought by the SEC is unfounded,” Anello said. “Jason did not provide Gavin Wolfe, or anyone else, with material nonpublic information regarding South Jersey Industries Inc. He did not breach any duty of confidentiality, and the SEC has no evidence that he did so because it did not happen.”

Wolfe’s lawyer, Reed Brodsky, also denied the allegations. “The SEC is pursuing this case despite being unable to identify what was allegedly disclosed, or how it was disclosed, while ignoring sworn, immunized testimony and contemporaneous documents confirming that Mr. Wolfe bought South Jersey shares based on an independent investment thesis,” Brodsky said.

Sure sure sure “he bought stock in a merger target the day after sitting with the target’s banker at a basketball game” sounds bad, but it doesn’t prove anything. And there doesn’t seem to be any other proof! Either it’s an innocent coincidence or it’s pretty good insider trading.

One concern about the rise of artificial intelligence is that it will displace entry-level employees in professional services. AI, in this theory, can do the basic tasks of a junior banker or lawyer or consultant, but it does not have the nuanced understanding or personal client relationships of a senior rainmaker. So you could have a viable business made up of rainmaking senior partners who bring in the business and AI agents who actually do it. But these businesses have historically operated on an apprenticeship model, and if there are no junior employees — or if there are fewer junior employees who are not putting in long hours of model-building and presentation-formatting — then where will future senior rainmakers come from?

At this point, though, these worries seem overblown, in part because, what, the senior rainmakers are going to learn how to use AI agents to build financial models? Come on. They’re busy making rain. You still need junior employees, but their job has changed. Instead of building financial models or drafting presentations, their job is to supervise the AI agents, and to teach the senior partners how to supervise the AI agents. Bloomberg’s Nil Codina Martinez and Zainab Haji report:

Banking interns typically spend the summer learning the ropes. But the industry’s push to adopt artificial intelligence means some have even found themselves doing a bit of teaching of their own this year.

A private banking intern at one lender in London said they were given the task of improving their team’s embracing of AI. That included building new agents to streamline to-dos and giving senior bankers demonstrations on how to use the technology in drop-in help sessions — a process they compared to teaching parents how to use a phone.

The role reversal is an extreme sign of how the entry-level banking experience is shifting as banks look to put Gen Z’s digital proficiency to use. Bloomberg News spoke to nearly a dozen summer analysts and interns who said that AI has altered what’s expected of them even in non-tech related programs.

Presumably this is not a stable equilibrium, and if this year’s interns do a good enough job of automating everything and/or teaching senior bankers how to automate everything, there will be no need for interns next year. But there is probably some lag, and AI is moving fast enough that next year’s interns will have all new ways to automate things.

Here is a Wall Street Journal article about how including quirky interests on your résumé is good actually:

In a cutthroat white-collar talent market, where AI looms as a job-security threat, some recruiters are encouraging job seekers to get ahead by emphasizing decidedly human qualities like a canasta addiction or a weakness for pickling vegetables.

Advocates say an offbeat pastime can tempt a glassy-eyed recruiter to reread your application and serve as an icebreaker in interviews. [Legal recruiter Kate] Reder Sheikh said more than half the associates she placed last year listed hobbies—a jump from a few years ago. …

Fred Cibelli, a technology principal at EY in New York, said that, assuming the candidate fulfilled the job requirements, he would want to hear about their regard for olive oil in an interview. “There’s only so many times you can talk about what’s going on with open versus closed models in AI,” he said. He has noticed a small increase in hobbies appearing on LinkedIn profiles and résumés lately and said he would like to see more.

Every once in a while I hear from a friends or former colleagues that they get résumés with “Money Stuff” in the “interests” line, and I want to encourage that. Those people get hired, probably, though this is not career advice.

Juicy Yields Draw Junk Bond ‘Tourists’ to High-Grade AI Debt. Nvidia’s Trillion-Dollar Chip Market Has Friends and Foes Closing In. Anthropic’s best AI model struggles to attract users as cheaper tools thrive. Paramount Prepares to Begin Early Settlement Talks With California Officials. Heirs to the Jack Daniel’s Fortune Are Fending Off a Takeover — and a Rogue Cousin. Hedge fund Saba takes on Baillie Gifford in new board battle. Startup Founders Are Working Harder Than Ever to Keep Up With Their AI Agents. Elite Private Schools Offer Financial Aid to Families Making $500,000. The age of the populist financial scam. PublicSquare “has also shut down a television show that executives hoped would drive conservatives to its marketplace and sold off a diaper brand marketed to antiabortion consumers.” I Can’t Stop Watching This Amazon Delivery Drone Drop Someone’s Package in Their Pool.

If you'd like to get Money Stuff in handy email form, right in your inbox, please subscribe at this link. Or you can subscribe to Money Stuff and other great Bloomberg newsletters here. Thanks!