The basic situation is that the US stock market will pay $2 for $1 worth of cryptocurrency. If a small public company has a $100 million pot of Bitcoin or Ethereum or Trumpcoin or what have you, its stock will be worth at least $200 million. This trade is baffling and magical; it was pioneered by MicroStrategy Inc. (which now calls itself Strategy, and which has about a $70 billion stash of Bitcoins and a $138 billion equity market capitalization [1] ), and is now regularly and successfully imitated by all sorts of random small companies.

I make fun of this a lot, but it is worth asking what it means that the stock market will pay $2 for $1 worth of crypto. There are in principle three sorts of answers:

1. A Bitcoin in the hands of a crypto treasury company is really worth more than a Bitcoin in your hands, because the treasury company can do stuff (investor education! lending! leverage! staking! tokenization! stuff!) with the Bitcoin that you could not easily do. Therefore the premium is justified for business reasons.
2. There are vast pools of institutional capital that (1) want to own Bitcoin but (2) can’t own Bitcoin directly, or through futures or exchange-traded funds or other more normal (and lower-premium) mechanisms. Therefore the premium is justified for market-segmentation reasons: Crypto treasury companies are getting paid a durable premium for bringing crypto to institutional investors in investable form.
3. Retail investors are lazy and/or confused and buy hyped crypto treasury stocks without understanding the enormous premium they are paying for crypto exposure. Therefore the premium is justified for meme reasons.

Every crypto treasury company says some version of No. 1 — it will do stuff with crypto, not just hold it — but I have never found it particularly compelling. No. 3 — “lol retail” — makes a lot of intuitive sense, and I’ve written a version of it. (“The obvious appeal of the crypto treasury strategy for most small US public companies is probably along the lines of ‘nobody is paying attention to our tiny company, but if we announce we’re buying a big pot of crypto, retail traders will get excited and overpay for our stock.’”)

But No. 2 is the interesting one. If the story is “giant asset managers want Bitcoin exposure, and Strategy is the only convenient way for them to get Bitcoin exposure, so they are willing to pay a 100% premium to buy Strategy and get Bitcoin exposure,” then that’s … super weird? And yet maybe true? When I look at Strategy’s shareholder list on Bloomberg, the second-biggest holder is Capital Group, the big active mutual fund manager, with 6.99% of the stock. Is this a good holding for Capital? Well, I mean, it’s up about 175% over the last 12 months, versus about 13% for the S&P 500 index. So: yes?

Should Capital just buy Bitcoin, instead of paying a 100% premium to hold Bitcoin through Strategy? (As Jim Chanos would argue?) Maybe, but it’s a moot point: Capital’s biggest holding of Strategy is through its Growth Fund of America, which “invests primarily in common stocks” and “may also invest in other equity-type securities,” but doesn’t seem to have Bitcoin (or Bitcoin ETFs) in its mandate. If you are a long-only equity mutual fund manager and your investment thesis is “I should buy Bitcoin,” then over the last year (1) you were right and (2) you weren’t allowed to buy Bitcoin. So maybe Strategy was the best practical way to implement that thesis.

(Should you be allowed to implement that thesis? Should you just change your fund’s mandate to allow Bitcoin? If you are a long-only equity mutual fund manager, are your clients paying you for Bitcoin exposure? Is it your job to buy them Bitcoin? Uh, I don’t know, I can see arguments either way, but in any case you are not technically giving them Bitcoin exposure: You’re giving them Strategy exposure, and Strategy is a stock.)

So one thesis for Strategy’s premium could be something like “institutional equity investors will come to demand Bitcoin-in-stock-form faster than Strategy can sell it to them.” Another, related but slightly different thesis is something like “index funds will be forced to buy Strategy at whatever the premium is.” Capital is the second-biggest Strategy shareholder, but Bloomberg’s Vildana Hajric reports on the biggest:

Bitcoin is not “appropriate” for long-term investors. Also, digital assets are more a speculation and less an investment. And they’re an “immature asset class” with little history and “no inherent economic value” that can wreak “havoc” on portfolios.

Vanguard Group Inc. executives, channeling the logic of the venerable Jack Bogle, have made their opinions on crypto clear. Yet thanks to the cold logic of index investing, the $10 trillion money-management giant is now the biggest backer of Strategy, the software firm that famously reinvented itself as a proxy for Bitcoin and became a poster child for the industry’s ambitions.

Vanguard owns more than 20 million shares, nearly 8%, of all of Strategy’s outstanding Class A common stock, and likely surpassed Capital Group Cos. for the no. 1 spot sometime in the fourth quarter, according to data compiled by Bloomberg based on regulatory filings. The dozens of Vanguard mutual funds and ETFs that hold the stakes track everything from small- and mid-cap benchmarks to momentum, value and growth gauges, among others.

And Strategy isn’t even in the S&P 500. (“Vanguard’s single biggest stake in the company is in its $1.4 trillion Total Stock Market Index Fund (VITSX), which has 5.7 million shares, worth roughly $2.6 billion.”) But it’s working on getting in. Think what fun we will have when that happens.

Also: This is fine? I make fun of this stuff all the time, but what do I know? Yesterday I made fun of a new crypto treasury company whose treasury will be filled with HYPE tokens. “The name here is particularly on-the-nose,” I wrote. But I make fun of regular old public companies sometimes too, and sometimes those stocks go up anyway, and nothing here is investing advice, and my money is mostly in index funds. I have learned that my desire to make fun of some goofy financial thing is completely unrelated to whether that thing’s price will go up. I have no particular insight into which financial assets will go up, so I try to be a price taker as much as possible: I try to invest in the market portfolio and get the market return. A lot of investors are in the same situation, or should be. [2]

In 2005, “the market portfolio” included mostly stocks and bonds. In 2025, it absolutely includes cryptocurrency, which is now a multi-trillion-dollar asset class. There are various ways to get exposure to crypto (you can just buy Bitcoin, or Bitcoin ETFs, etc.), and I am sure that like 10 people are going to email me about their startups that offer convenient ways to get indexed exposure to crypto (you give them $100 and you get $100 of market-cap-weighted exposure to a bunch of big crypto tokens).

But the simplest and laziest way to get sort-of-index-ish exposure to crypto is to own the total US stock market, because the stock market now includes an ever-growing supply of crypto treasury companies. You might not want crypto in your stock index fund — Vanguard doesn’t want crypto in its stock index funds — but the whole point of an index fund is that you don’t want to invest in what you want! (Or in what a fund manager wants.) You don’t trust yourself (or your fund manager) to want the right things. You want to invest in what the market wants, and what the market wants is crypto.

Superior Industries International makes aluminum car wheels, which it sells to the companies that make the cars. [3] It makes the wheels for US cars in Mexico. You see the problem. Making aluminum wheels in Mexico to sell in the US was until recently a decent business; this March, Superior said that “the overall cost for us to manufacture wheels in Mexico is currently lower than in the United States, due to lower labor costs.” And then President Donald Trump announced high tariffs on Mexico and on aluminum, and now it is a worse business.

To be fair, it wasn’t a great business before; Superior’s 2024 highlights include “maintained profitability despite tough industry environment” and “refinanced all existing debt; maturities extended to 2028.” Superior was already bit fragile. But then Trump got going on tariffs and things got worse. In May, Superior added a restructuring expert to its board of directors. By June, the New York Stock Exchange was threatening to delist Superior because its market capitalization was less than $50 million and it had negative shareholders’ equity. It also needed more cash, so in June it struck an agreement with some of its lenders — who had already loaned it $520 million — to borrow an extra $70 million to keep it going. 

If you were one of Superior’s lenders, and you got a call asking for more money in June 2025, how would you have answered? Was this a good investment? Plausibly the answer would be “depends on the tariffs.” Trump announced high across-the-board tariffs in early April, and then “paused” them for 90 days a week later; June 2025 was sort of the middle of the pause, when uncertainty was arguably maximized. Plausibly your credit evaluation would be:

1. If the tariffs are real, we should not be lending another $70 million — on top of our existing $520 million of loans — to a company that makes aluminum car wheels in Mexico to sell in the US.
2. If the tariffs are just a negotiating ploy, it’s fine.

What do you do with that evaluation? One answer might be “you make an informed estimate of the probabilities of each outcome, you figure out the right interest rate for the loan in each scenario, you probability-weight them and charge the company the blended rate.” Other answers are “you figure the risk is too high and say no” or “you confidently figure that the risk is overstated and say yes.” But the answer you really want is:

1. We lend you the money if the tariffs are fake; 
2. We don’t if they’re real.

That is, instead of evaluating and pricing this risk yourself, you put the risk back on the borrower: The borrower gets the money in the no-tariffs state of the world, but not in the yes-tariffs state of the world. 

That is not exactly easy to structure, but it’s not completely impossible either. The key ingredients are:

1. Instead of giving the company all the money now, you parcel it out over time.
2. You write into the contracts something like “if tariffs happen, you have to give back all the money immediately.” Not just the new money, but also the $520 million that you’ve already loaned to Superior. 

This is not perfect: If the tariffs are real, the company probably won’t have enough money to pay you back. But at least you catch the problem early, and ideally are first in line to get paid back. Instead of lending the company a lot more money for three years and hoping it makes it through tariffs, you lend the company a little more money for a month and see what happens with the tariffs.

This is roughly what Superior’s lenders did: In exchange for the extra $70 million of commitments ($10 million of which was funded right away), Superior agreed to a tariff clause requiring it to pay back all of its loans if:

There occurs any tariffs on any of the shipments of the Borrower or its Subsidiaries into the U.S. exceeding 20.0% of the product value (the “Tariff Threshold”); provided that, if on or prior to the date that is sixty (60) days after the date such tariffs exceed the Tariff Threshold, the Borrower enters into agreements (or otherwise makes arrangements) such that such tariffs are either paid by or otherwise covered by (including as a result of price increases) the customers of the Borrower and its Subsidiaries, then no Event of Default in respect of this section (12) shall be deemed to have occurred.

If there are no tariffs, or if Superior can pass on the tariffs to its carmaker customers, no problem. If there are 20+% tariffs and Superior can’t pass them on, no money.

Six weeks later, the uncertainty has not resolved exactly, but things have not improved, and last week Superior agreed to hand over the keys to the company to its lenders. (Technically the lenders “will convert up to approximately $550 million of their term loan claims into 96.5% of the common equity” of the company.) Ah well. Superior was viable in the no-tariffs state of the world, not viable in the yes-tariffs state of the world, and the bad state came true. 

A weird theme in modern banking is that, if you ask a bank for a loan, the bank will say “yes we would be happy to help you find that loan, we have lots of specialist bankers who are expert in the loan markets and who can help you connect and negotiate with lenders,” and you might say “well okay sure but you are a bank, why don’t you just lend me the money,” and the bank will say “ahahahahahaha you dear sweet summer child, we can’t lend you money, are you nuts? What if you didn’t pay us back? We are a bank, we can’t take that risk.”

I am exaggerating, but this is roughly the story for large loans to risky corporate borrowers (and particularly private equity buyouts). In modern times, the normal way those loans get made is that a bank structures the loan and then sells it to a group of investors (credit funds, collateralized loan obligations, etc.); this is called the “broadly syndicated loan market.” The bank takes some risk — it commits to backstop the loan, and if it can’t sell the loan it will be stuck with it — but does not, in the ordinary course, expect to lend the company the money itself; its job is to find other people to lend the money. More recently, banks have “gotten into private credit,” which means that they (1) raise dedicated funds from outside investors and then (2) use those funds to make loans, directly, to companies. Here again, though, it’s not the bank’s money; the bank’s job is still to find other people to lend the money. In the broadly syndicated loan market, it negotiates the loan first and then finds the money; in the private credit market, it finds the money first and then negotiates the loan. But either way the bank is not trying to be in the business of using its own balance sheet to make loans.

It is a weird state of affairs, and here is a funny Wall Street Journal story about how JPMorgan Chase & Co. had various unsatisfying experiences with broadly syndicated loans and “getting into private credit” (raising funds to make loans), and then eventually realized “wait, we are literally JPMorgan, we have so much money, we can just use our money to make loans”:

The bank rolled out a private-credit fund roughly modeled on the old Highbridge unit, again under Erdoes and in the asset-management division. The idea was to have it be separate from the core bank so it wouldn’t be subject to punitive regulations. Erdoes installed one of her lieutenants, Meg McClellan, as the global head of private credit in 2020 and hired two senior bankers from Wells Fargo with a background in lending to smaller companies. They hired a half-dozen analysts and got $10 billion from the bank to get the strategy going.

By early 2022, Dimon and Erdoes scrapped the whole strategy. Dimon decided instead to put the bank’s balance sheet to work, allocating some of JPMorgan’s $100 billion in excess capital to the investment bank for private lending. The funds are above the reserves banks are required to hold to protect deposit accounts. 

In the new setup, investment bankers working with companies could offer them loans directly, instead of having to pass over a Chinese Wall to Erdoes’ asset-management division to get approval for a deal. 

It does seem like an advantage? It is striking that JPMorgan’s strategy of “we have money so let’s use it to make loans” is unusual among banks, controversial within JPMorgan, and only came to them after several false starts along the lines of “let’s see if someone else would give us money to make loans.” But it’s a good idea. Bloomberg’s Hannah Levitt reports:

JPMorgan Chase & Co. created a unit in its commercial and investment bank focused on bespoke financing structures across public and private markets.

The business, dubbed strategic financing solutions, will be a collaboration between banking, markets and sales, according to an internal memo Monday. It will initially focus on structured private solutions, infrastructure finance, strategic asset-backed securities finance, merchant banking and direct lending.

It turns out that a lot of companies want to borrow money in weird ways, and that is a somewhat difficult thing to intermediate: If a client comes to you looking to get a specific sort of weird loan, you are unlikely to be able to go out and find another client who wants to make exactly that sort of weird loan, even if it’s a good investment. But if you just have money, you can make the weird loan yourself. 

There you go:

Artificial intelligence coding startup Cognition AI Inc. has agreed to buy what remains of Windsurf for an undisclosed sum days after Alphabet Inc.’s Google struck a $2.4 billion deal for Windsurf’s top talent and licensing rights. …

The Cognition deal includes Windsurf’s intellectual property as well as its remaining employees, cash, assets and brand. Windsurf has $82 million in annual recurring revenue and its enterprise business has been growing fast, Cognition co-founder and CEO Scott Wu wrote in a note to staff.

All of Windsurf’s employees will participate financially in the deal and see their vesting schedules accelerate, Wu wrote. “Jeff and I worked together to ensure that every single employee is treated with respect and well taken care of in this transaction,” Wu said in the note.

We talked yesterday about Google’s deal to acquire 0% of Windsurf for $2.4 billion, which provided a lot of cash for the top talent and venture-capital investors but left the remaining employees somewhat in the lurch, arguably undermining the implicit startup social contract. Now they are less in the lurch. Ben Thompson notes: “What is noticeably absent is a purchase price; in fact, my strong suspicion is that the purchase price nets out to $0.” The Google deal was the exit for Windsurf’s founders and investors; the Cognition deal is the possibly less exciting exit for the employees. The social contract is a bit more fragile than it was a week ago — what’s to stop Google from hiring the next startup’s founders without giving the VCs or employees anything? — but for now it has held.

The most important modern meta-skill is statistical inference with computers. If you can cause a computer to look at a big pile of data and use it to make predictions, there are many things you can do with that skill. You can, for instance, advance humanity’s understanding of distant galaxies. In some ways that is the paradigmatic use case; people who hire machine learning researchers love to hire astrophysics PhDs. [4] In other ways, it is not the paradigmatic use case, because it is extremely difficult to make money by advancing humanity’s understanding of distant galaxies. Neil deGrasse Tyson seems to do well, but otherwise it’s a pretty small market.

There are other, more lucrative use cases. If you can train a computer on a pile of data about soccer or hockey, and then use it to predict which young prospects will become good soccer or hockey players, that is a skill worth many, many millions of dollars, and so we talk from time to time around here about soccer and hockey astrophysicists. But even that is small potatoes compared to the big three, each of which is worth many hundreds of billions of dollars:

1. Training a computer to predict which advertisements to show to people online.
2. Training a computer to predict which stocks will go up.
3. Training a computer to predict the best next word in an essay, or the next pixel in a picture.

These are normally called “big tech,” “quant finance” and “AI labs,” respectively, though all of those labels are somewhat loose. Traditionally big tech and quant finance were lucrative and desirable employers of machine-learning researchers. But, uh, I don’t know if you’ve noticed, but now top researchers at AI labs are regularly in the news for getting nine-figure pay packages. I am sure that many of the astrophysicists who work at quantitative finance firms do it because they love the people that they work with and the problems that they work on; they love the excitement of the markets and the immediate results they can get from applying math to trading strategies. But probably a few of them are in it for the money, and now the money in AI is soooooooooooooooooo good. As I wrote last year (about DeepSeek, an AI lab that grew out of a quant hedge fund):

An important story in the development of modern finance was “if you are good at building computers that can process natural language, you can pivot to building computers that can pick which stocks will go up, and get much richer.” This is essentially the story of Renaissance Technologies, whose central innovation was to hire computational linguists, because there was not much money in building computers that could talk and lots of money in building computers that could pick stocks.

But that is an old story, and now everyone has a computer that can pick stocks, while there is infinity money in building a computer that can talk. So now the people who got good at building computers that can pick stocks are pivoting to processing natural language.

Anyway, at Business Insider, Alex Morrell and Julia Hornstein have a story about OpenAI hiring all the quants:

Poaching quants for Silicon Valley isn't new. But AI startups awash in cash can not only match but outbid Wall Street pay. Junior and midlevel traders at top high-speed trading firms are now fielding multimillion-dollar packages — up sharply from a year ago, quants and quant recruiters with direct knowledge of the offers told Business Insider.

OpenAI has scooped up researchers, engineers, and senior recruiters from firms like Hudson Riv