Pre-IPO

Here’s a question I have about equity capital markets. We have talked a few times about pre-IPO perpetual futures: Some offshore trading platforms, most notably Hyperliquid/Trade[XYZ],^[1] list perpetual futures contracts on big private companies that are approaching their initial public offerings. These contracts are bets on the future price of the stock, and it is not at all obvious that the futures price would closely track the actual stock value: The perps do not represent any claim on the underlying stock, they are in theory off-limits to investors in the US, existing shareholders of the underlying companies are probably not allowed to sell the perps to hedge their exposure, big institutional investors who might want to buy shares in the IPO are probably not set up to buy the perps to anticipate their demand, etc. The perps are in a sense just a recreational prediction market on the price of the stock, not a derivative instrument tied to the price of the stock by arbitrage mechanisms.

But a really important thing we have learned about finance in the last few years is that prediction markets are pretty accurate. That is: If you have a platform where people can make real-money bets on some outcome, their bets will tend to give a pretty accurate and well-calibrated estimate of the actual outcome. Kalshi event contracts are not tied to, like, election outcomes or football scores by arbitrage mechanisms, and in the abstract Kalshi could just be wrong about everything. But in fact Kalshi publishes research showing that its markets are good at predicting stuff. And on the Money Stuff podcast, Susquehanna International Group prediction market trader Jeremy Maletz made the same point, noting that prediction market odds are well calibrated even when the volumes are not enormous. The wisdom of the crowd largely works.

And so in fact there is a track record of pre-IPO perpetual futures on Hyperliquid being quite accurate, that is, predicting the opening prices of IPOs. It’s quite a small sample — the main US examples are Cerebras and SpaceX — and, as I have written, kind of unsettling. But so far it seems to be empirically true, and it is consistent with the broader theme that prediction markets work.

Let’s say this track record continues and expands. Let’s say that by the time OpenAI goes public next year, the Hyperliquid perpetual futures will have been proved out as accurate predictors of trading prices. The price of OpenAI on Hyperliquid on the afternoon of the IPO^[2] will predict the opening price of OpenAI on the stock exchange the day after the IPO about as well as the closing price of Tesla stock on Nasdaq today will predict the opening price of Tesla on Nasdaq tomorrow: not perfectly (news happens, etc.), but in a reasonably continuous way. And let’s say that everyone knows this: The pre-IPO perpetual futures get lots of news coverage, they are routinely referenced by analysts and investors and bankers, they are part of everyone’s understanding of the IPO process, etc. 

And then OpenAI goes out and markets its IPO, and it meets with investors and gets feedback. And the pre-IPO perps keep trading. Let’s say the perps say that OpenAI is worth $120 per share, and they are relatively stable in the days leading up to the IPO. Investors put orders in for the IPO, the banks build a book, it’s 10 times oversubscribed or whatever, and on the afternoon of the IPO the bankers come to OpenAI to set the IPO price. The bankers say to OpenAI: “We know, as confidently as anyone can know anything in this business, that your stock will open at $120 per share tomorrow. We are less confident that it will close at $120 per share — markets can be weird — but that’s our best guess. So we think you should price the IPO at $_____.”

What goes in the blank? I think there are two possible answers. One — the traditional answer — is $100. If you were the equity capital markets banker leading a big IPO, and perpetual futures markets did not exist, but an infallible genie appeared to you in a dream the night before the IPO and said “this stocks’s gonna trade at $120,” you would naturally price the IPO at $100. That way, you will get a 20% IPO “pop,” which is perfect. All of the investors in the IPO will have an immediate mark-to-market profit, though most of them will not cash out: You have placed the IPO shares mostly with long-term fundamental investors who believe in the company’s vision, not with fast-money flippers, and you want to reward their belief with a quick profit. You are in a repeated game: You call the same investors for every IPO, and you want them to take your calls and put in the effort to understand the new companies that you are bringing to market, and the way to make sure they do that is by making IPOs generally lucrative for them.

Bill Gurley will complain that you are leaving money on the table. In fact most big IPOs do come with “pops” — often considerably more than 20% — and Gurley is the most prominent proponent of the view that this is bad for companies. The company, in this view, should sell its stock at the price the market will pay for it, which here is basically $120. If it deliberately prices it cheap, it is transferring billions of dollars of value from the company (and its pre-IPO shareholders) to its banks’ hand-picked favorite institutional investors.

One standard response to this is that the company is also in a repeated game: It will need to raise money in the future, and it will want public-market investors to think fondly of it. “Oh yeah, those guys, nice IPO pop they had, sure we’ll give them another $1 billion,” that sort of thing. Arguably this response is more true about, you know, OpenAI than it usually was about tech companies over the last decade: The artificial intelligence boom has made US tech companies more capital-intensive, and more likely to raise equity frequently, than they were a few years ago. Keeping capital providers happy is a more important corporate goal than it used to be. If the banker told the company “an infallible genie told me that your stock will trade to $120, so let’s price it at $100,” the company would probably be like “makes sense yeah.” Elon Musk apparently priced SpaceX’s IPO to target a 20% pop. He knows.

But the other possible answer is, of course, $120. The theory here is: Look, we know the stock will trade at $120, so why give a hand-picked group of big investors a giant one-day windfall profit? The traditional reason is “to incentivize them to take risk on an untested IPO,” but that’s no longer relevant: They know the trading price, so they’re not taking that much risk, so they don’t need a 20% one-day gain to compensate them for it.^[3]

Public companies sometimes (more frequently these days) do big stock offerings. They do not, ordinarily, price them at a 20% discount to their trading price.^[4] Meta Platforms Inc. sold $25 billion of common stock last Monday; the stock closed at $630.56 per share that day, and the offering was priced at $625, about a 0.9% discount to the last trade.^[5] Possibly that’s the sort of discount that OpenAI should be targeting.

Giving new IPO investors a 20% discount to your hypothetical guess about the future trading price of the stock is justifiable; giving new investors a 20% discount to the actual current trading price is harder. The perps are not the same as an actual current trading price of the stock, but they kind of are a little bit?

I have been writing for years that “private markets are the new public markets,” but I have usually meant that in a pretty loose way. “Private companies can be bigger and raise more money from more traditionally public sorts of investors and offer them more liquidity,” that sort of thing. Even in that world, though, the transition from private to public is pretty stark. You’re a private company one day, and a public one the next. Your shareholder base and public perception and trading dynamics and liquidity all shift pretty drastically, and your stock price three days after the IPO is hard to predict from the price of your last private funding round. There’s a whole apparatus — of IPO roadshows and investment bankers and IPO pop rules of thumb and Bill Gurley — to try to make that prediction, but it is imperfect, more art than science.

But now maybe that problem has been solved by, you know, the wisdom of the crowd, or by properly incentivized prediction markets, or by offshore financial innovation, or by blockchain or what have you. Maybe, for big private companies, private and public markets are now completely continuous, and your stock can trade liquidly at $120 the day before your IPO (on Hyperliquid) and at $120 the day after (on Nasdaq).

Anyway, those are my musings! Here are three news hooks. First, the Wall Street Journal reports that “institutional investors told [Anthropic PBC] and its advisers that the company’s valuation in its planned initial public offering later this month should be around $1.5 trillion to $1.6 trillion, according to people familiar with the matter, well below figures closer to $2 trillion that had been bandied about in recent months.” Why? To get a 20% IPO pop of course:

Some investors say they trust Anthropic’s chief financial officer Krishna Rao to set a conservative valuation and give priority to delivering IPO investors a first-day “pop” in the stock price. They believe that even if Anthropic debuts at a roughly $1.5 trillion valuation, it could jump to $2 trillion by the end of the first trading day. Elon Musk’s SpaceX jumped 30% during its first day trading in June—before widespread investor jitters around AI had taken hold.

And in fact Anthropic is trading at around a $1.8 trillion valuation on Hyperliquid, so a $1.5 trillion IPO price would deliver an almost perfect 20% pop.^[6] We are arguably in a transitional phase where big institutional investors can go to Anthropic and say “your company is worth $1.8 trillion, so you should price it at $1.5 trillion.” And, look, it makes sense. This week, anyway, we seem to be in a buyer’s market for gazillion-dollar AI financings: OpenAI raised less than its target in its latest private round, and the Wall Street Journal notes that “lenders and investors are getting choosier about where to place their bets and are demanding tougher terms.” “Give us a 20% discount to market value” is a reasonable thing for investors to ask. But I am not sure that that will last forever.

Second, today the US Commodity Futures Trading Commission approved trading of perpetual futures on US stocks. Not, currently, on pre-IPO stocks, but we’ll see. And, third, Bloomberg’s Ryan Weeks has a profile of Jeff Yan, the founder of Hyperliquid, who visited the US to try to, you know, get Hyperliquid into the US:

He insists that Hyperliquid, widely described as an exchange for digital assets, isn’t an exchange at all but “a layer on which exchanges can be deployed.” He likens it to the internet; it falls to those building things on top of Hyperliquid to figure out the checks and balances. Rather than opining on the future, Yan wants to focus on the code. “I just work all the time,” he explains. “For me, work is all-consuming.” ...

Even though Hyperliquid, which first went live in 2023, isn’t officially accessible in the US, its rise has rattled industry titans, including the world’s largest derivatives exchange, CME Group Inc. The White House has taken notice too. President Donald Trump, an industry advocate who earned more than $1.4 billion through his own crypto ventures last year, has said that regulators are “working very hard” to allow Hyperliquid to serve American traders. So Yan, usually in Singapore, is here in Washington, where he supports the Hyperliquid Policy Center, an arm’s-length lobbying group that’s peppered US agencies with letters about the merits of the platform. …

Yan says the way to break into the world’s largest financial market is for HIP-3 partners, rather than Hyperliquid itself, to submit to US regulation on matters such as know-your-customer checks and market surveillance. (This is where his argument that he merely provides a tech platform, not an exchange, becomes crucial.) In September, Hyperliquid began testing a modified version of the HIP-3 framework that allows for customer identity verification, a basic due diligence step that the platform doesn’t currently offer.

Back in 2015, in the very early days of people putting traditional financial assets on crypto infrastructure, I wrote: “Just because you mumble the word ‘blockchain’ doesn’t make otherwise illegal things legal.” I’m increasingly convinced that I was wrong. Like: “We have the ability to offer any financial product to anyone, immediately” is a really good pitch, and one that is not possible when you are subject to traditional regulatory approvals. Just build a protocol on the blockchain and see what happens!

Long-term-ism

Arguably the core job of a multimanager multistrategy hedge fund is to distinguish skill from luck. There are some people in the world who do some trades that have generated some past series of returns, and the job of the hedge fund is to look at those return series, talk to the people, and figure out which of them will make money in the future. The most obvious sign of future success is, of course, past success, but:

1. Past performance does not guarantee future returns, just as a general matter.
2. Paradigms shift, regimes change, trades get crowded, and you want to hire people who will make money in the next three years, not the last three years.
3. “A high Sharpe ratio strategy is simply one that hasn’t encountered its left tail yet,” and you have to figure out who is picking up pennies in front of a steamroller that is about to arrive.
4. Some sorts of trading offer larger sample sizes than others, and if you’re looking to hire a long-term buy-and-hold stock investor, her good track record on buying and holding 10 stocks in her career could be attributable to luck that will not persist.
5. Conversely, her bad track record on buying and holding 10 stocks in her career could be attributable to bad luck that will not persist, and you can scoop her up cheap. An investor with a bad track record will probably demand less compensation than one with a good track record, so if you can find high-skill-low-luck investors you can get good future performance and a good deal. “That moment of weakness is when they’re most poachable, driven and available at a discount.”
6. (Whereas the high-skill-high-luck investors might not be available to you, or might charge you 110% of their alpha.)
7. A little adversity is character-building. “My philosophy when I used to hire traders was that the optimal number of past blow ups was one,” John Arnold has said.

Much of this depends on the belief that skill is persistent and luck isn’t, which might be wrong. I think sometimes about a hedge-fund investor who said “I do want to invest in lucky people — that’s better than investing with unlucky people.” Not a rigorous model, but there’s something to it!

But if you run a big multistrategy fund you probably believe in skill. So you will have some process for figuring out who has a consistent repeatable source of alpha; their track record, their thought process, their type of trading, etc., will be inputs to this process. Sometimes you will choose a candidate with a pretty good track record over one with a great track record, because your other inputs tell you that the good track record is more repeatable. Occasionally your process will tell you that a candidate with a bad track record, someone who has lost money consistently, is highly skilled and will make money in the future. “Wait, really,” you will ask your process. You will take a second look. But if it checks out, you might make the counterintuitive call.

Anyway Bloomberg’s Nishant Kumar has a profile of D.E. Shaw & Co. that focuses a lot on how much money many of its employees have lost:

Max Stone will never live it down: Once, he nearly blew up D.E. Shaw & Co., one of the first great quant shops on Wall Street.

Stone, barely 30 at the time, had drastically underestimated the financial risks he was taking. When markets turned — disaster. D.E. Shaw lost roughly 30%. Stone thought his head would roll.

“If I had been fired, no part of me would have thought that was unfair,” Stone said in a recent interview. “I did nearly destroy a company.”

Oopsie. Stone is now “first among equals” on D.E. Shaw’s management committee, and not alone:

Executives say they’ve cultivated a culture that prioritizes long-term success over short-term wins. Look at Stone. He and several other executives who are at the top today sometimes lost big on the way up. Once, an executive was promoted even though, all told, he’d lost money during his time at D.E. Shaw. 

“Well, you keep losing money for investors, but we have faith that you’re going to turn it around so we’re promoting you” is a funny thing for a hedge fund to say to its employees, but that is kind of the job! Also it seems to work: “Assets recently breached $100 billion, and returns this year have outpaced major rivals.”

(Oh, disclosure, another point that Kumar makes is that, unlike a lot of its more straightforwardly quant-y competitors, D.E. Shaw likes to hire “English majors, philosophy grads, Broadway actors, the occasional writers-workshop MFA or modernist painter.” Ages ago, as a recent Harvard classics graduate with no finance background, I at some point interviewed and was offered a nebulous job at D.E. Shaw, but ended up in law school and whatever this is. Alternate histories!)

Fake job interview

If I ran a legitimate crypto exchange, and I was interviewing engineers for jobs at my exchange, I would definitely:

1. Ask them to download a coding assignment as part of the interviewing process, and
2. Make that download a virus that takes over their computer.

I would not do this maliciously. It’s just a test. The virus would just be like “you have failed the assignment, goodbye” and then hand their computer back to them. I just feel like the one very most important job of an engineer at a crypto exchange, way more important than building fast matching engines or coding sophisticated products or even keeping an accurate list of customer accounts, is not getting your computer hacked. All day long, at a crypto exchange, people are emailing you attachments and texting you links and handing you cards with QR codes, and like 90% of them are chock-full of malicious code designed to take over your exchange and steal all the crypto. “The fate of all crypto exchanges is to be hacked,” I say constantly, and the way that happens at crypto exchanges and everywhere else is mostly by people clicking on stuff carelessly.

Anyway this is like a double firing offense:

Binance Holdings Ltd. said Tuesday that the hackers behind a $6.3 billion heist used social engineering techniques including fake job interviews, as well as malicious software, in the most detailed report yet on how the incident unfolded.

Binance, which attributed the hack to North Korea, said the breach involved a “sophisticated, multi-stage campaign”. One of the company’s engineers was contacted through a job hunting platform by someone posing as a project director, the exchange said in the report published Tuesday. They were then interviewed by a person claiming to be the chief executive officer of a fictitious company over a social chat app, and asked to download an assignment containing a “malicious code repository”. This then set off a series of unauthorized activities, the report said.

“This was not a single mistake or a single breach. It was a campaign built over many months,” said Richard Teng, co-chief executive officer of Binance. The attackers “combined fake recruitment interviews, malicious software disguised as legitimate developer tools, compromised devices, and AI techniques.”

Back when I was young and idealistic, I thought that there was nothing riskier and more shameful than interviewing for a new job while working at my current job. “What if my boss finds out that I’m not 100% loyal,” I worried. Now I am older and more cynical and think that (1) employees have a right to earn their market value and (2) employers understand that, so I am more blasé about these things. (Though, also, 100% loyal to Bloomberg.) But if you are interviewing for a new job in a way that costs your current employer $6.3 billion, they will probably want to get rid of you.

Things happen

Major Hedge Funds Mostly Lose in ‘Nowhere to Hide’ October. Hedge Fund Elliott Warns of FOGO: Fear of Getting Obliterated in the Market. Ray-Ban Heir Mounts €30 Billion Counterbid for Family Firm. North Korean hackers expected to take months to launder stolen Binance billions. Eager to Rebut Rogue-Nation Image, North Korea Denies Role in Binance Heist. Venezuela’s Oil Rush Brings Private Jets and Soaring Rents to Frontier Town. Energy Transfer Approves $1.6 Billion Gas Pipe as AI Use Soars. Disney Plans Thousands of Layoffs, Targeting 20% Cost Savings. These Associates Had a Golden Ticket to Big Law — Until Weil Gotshal Collapsed. Man caught in car with £14k of drugs as a lamb ate chips in the back seat▸ AI summary avoids jail.

Follow Us

You received this message because you are subscribed to Bloomberg’s Money Stuff newsletter. If a friend forwarded you this message, sign up here to get it in your inbox.

|