Money Stuff: OpenAI Is Almost Public

Today Bloomberg’s Bailey Lipschultz and Edward Ludlow report that SpaceX filed for the largest initial public offering ever:

SpaceX has filed confidentially for an initial public offering, according to people familiar with the matter, bringing billionaire Elon Musk’s rocket, satellite and AI company closer to delivering the biggest-ever listing. … 

The filing puts it on track for a June listing, which would make SpaceX the first of what could be a trio of mega-IPOs, ahead of OpenAI and Anthropic PBC. ...

A listing for SpaceX would raise as much as $75 billion, Bloomberg News has reported. At that size, it would dwarf the current record holder, Saudi Aramco’s $29 billion debut in 2019.

But Bloomberg’s Shirin Ghaffary reports that OpenAI did an even bigger deal:

OpenAI has completed a deal to raise $122 billion from investors at an $852 billion valuation, marking the company’s largest funding round to date by far and bolstering its costly push for more chips, data centers and talent. ...

In a first for the company, OpenAI raised more than $3 billion from individual investors through bank channels. The startup also said it will be included in several exchange-traded funds managed by Cathie Wood’s Ark Invest, with the goal of offering more people exposure to the AI firm.

OpenAI Chief Financial Officer Sarah Friar said the financing “blows out of the water even the largest IPO that’s ever been done.” The deal, she said, is meant to give the company “a lot of flexibility” to invest in computing resources and its AI roadmap at a time of broader uncertainty for the public markets, including from the Iran war.

Private markets, I often say around here, are the new public markets: You can raise more money for your AI company in a private round than you could in the still-hypothetical largest IPO ever. You can sell $3 billion of stock to retail investors, or quasi-retail investors anyway. (Bank private wealth channels, ARK ETFs.) Going public still has some benefits, but if your goal is just to raise $122 billion to build data centers, sure, the private markets can handle that.

Of course everyone assumes that OpenAI is also a near-term IPO candidate. Presumably it would raise more money in that deal, too. The appetite for data centers is insatiable.

Another thing that we have talked about a lot around here is the amount of demand from individual investors for stock in giant private companies like SpaceX and OpenAI. Broadly speaking, if you are a wealthy individual investor, there are a lot of people who will sell you SpaceX or OpenAI stock, in some form, at a 100% or 1,000% markup. Why do people pay those markups? One possibility is confusion. Another possibility is that, when you buy a hot private stock, you are not paying exclusively for future cash flows. There is a certain cachet in owning a stock that most people can’t own, in being on the inside; you can go to parties and say “oh yes I am an early investor in OpenAI.” That is, apparently, worth something, and I guess OpenAI might as well monetize it itself.

The traditional simple arithmetic is that a lot of people want to buy shares of a big company, but when the company is private, not that many people can. When the company goes public, many more people can buy its stock, so it should be worth more: All the retail investors and public-market funds and non-ARK ETFs will finally be able to buy it, and they will, so its stock will go up. Therefore, if you buy shares while the company is still private, you can get a high return when the company is finally able to tap into that demand.

It is possible to over-learn that lesson, though. “Boy I’m going to make a fortune when everyone is finally able to invest in OpenAI,” you say, when your banker sends you the exclusive invitation to invest in OpenAI’s $122 billion funding round. And then you go to parties and brag about being an early investor in OpenAI and find out that everyone else is too. Who are you going to sell to when the company goes public? 

We have talked a few times recently about the push to put more private assets in ordinary investors’ retirement plans and, you know, this. “Private assets have higher returns than public stocks, so we should let everyone own them” has a certain appeal, but if you act on it it stops being true: Arguably the private assets have higher returns because not everyone can own them. If everyone gets access to the exclusive investments, they are no longer exclusive.

Meanwhile, a traditional advantage of staying private is that, if your stock doesn’t trade publicly, it can’t go down. Having a falling stock price is distracting for employees and annoying for executives, and being private mostly lets you avoid that: Your stock doesn’t trade every day, and its price only moves when you do a new fundraising round, which you should try to do only when there’s good news. But when you’re an $852 billion household-name private company, sorry, private markets are the new public markets, and Bloomberg’s Hema Parmar reports: 

OpenAI shares have fallen out of favor on the secondary market — in some cases becoming almost impossible to unload — as investors pivot quickly to Anthropic, its biggest competitor.

Even as OpenAI raced in recent months to raise tens of billions of dollars, Next Round Capital founder Ken Smythe said his secondary marketplace was seeing a drop in demand for the artificial intelligence giant’s shares. About a half-dozen institutional investors — including hedge funds and venture capital firms that hold large stakes — approached his company in recent weeks looking to sell about $600 million of OpenAI shares.

Last year, they would have been snatched up within days. But now, no one’s biting.

“We literally couldn’t find anyone in our pool of hundreds of institutional investors to take these shares,” said Smythe, whose firm has handled $2.5 billion of transactions. Meanwhile, “buyers have indicated they have $2 billion of cash ready to deploy into Anthropic.”

Other marketplaces are also seeing record demand for Anthropic, including Augment and Hiive. The large gap between OpenAI’s $852 billion valuation and Anthropic’s $380 billion has investors rushing to grab equity in the latter before it rises, according to Augment co-founder Adam Crawley.

“It’s just better risk-reward right now,” he said. “People are betting that Anthropic’s valuation will catch up with OpenAI’s. But if you buy OpenAI shares, it’s less clear what the return will be in the near term.”

See, OpenAI is practically already public, and if you’re a sophisticated institutional investor in private markets you might be better off putting your money in earlier-stage $380 billion companies.

One model of shareholder activism is that shareholders control the company, they elect the directors, and if they don’t like what the directors are doing they can vote them out. In practice this is pretty hard: To vote out the directors, an activist shareholder has to launch a proxy fight and convince a majority of shareholders to vote for her slate, and the board has tools to delay or impede this. But at least in theory, and sometimes in practice, shareholders really can vote out the board and install new management, and this possibility gives them leverage. And so the directors and executives are responsive to shareholder demands, because they operate in the shadow of binding shareholder votes.

Another model of shareholder activism is that directors and executives are responsive to shareholder demands not because of the threat of a proxy fight but because, come on, we are all professionals here and we all want the same thing. It is good for everyone — the executives, the directors, the activist shareholders — if the stock goes up. The executives and directors have fiduciary duties to the company; they are supposed to make the stock go up. They are chosen for their skills and experience and energy, and they are compensated with stock and options to give them the right incentives. If an activist shows up, the board will thoughtfully engage with her, take her ideas seriously, and either implement them or provide some rational explanation, grounded in shareholder value, for why not.

And if the shareholder doesn’t like the explanation, her main tool to push back is not a proxy fight but embarrassment. Directors want to do the right thing for shareholders, and they want to be seen to do the right thing for shareholders. Being a successful director — one who presides over a rising stock price and is beloved by shareholders — is good for your career; doing good work on a good board can get you more board seats and other business opportunities. Plus you’re probably in a golf club with other public-company directors and executives, and you want to be respected at the golf club. If an activist is out publishing nasty letters about how you’re destroying shareholder value, that is motivating on its own. The proxy fight — the theoretical possibility that you will actually be kicked out of your directorship — is a minor footnote.

These theories have a great deal of overlap in most companies: If a board is doing the wrong thing for shareholders, that might both embarrass them at the country club and lead to a proxy fight. But there is an important area of non-overlap: Lots of companies have dual-class shares that concentrate voting power in the hands of controlling shareholders (generally founders, etc.), so that the public shareholders really can’t vote out the board. If the managers and directors are doing a bad job, the shareholders have no formal power to change things. But they can still write letters. They can still try to persuade the board to make changes, and the board still has some fiduciary duty to listen, and if the board ignores the shareholders and the stock keeps sliding, that will be embarrassing at the club.

Snap Inc. is an unusually extreme case of this in that (1) it went public in 2018 with non-voting public stock, so public shareholders have no voting rights at all and (2) uh, it’s Snap; it has not been on a great run over the last few years. Still, worth a shot. The Wall Street Journal reports:

Activist investor Irenic Capital Management has built a roughly 2.5% stake in Snapchat and is asking the company’s leadership to make a series of changes to boost the company’s valuation, including laying off employees and closing or spinning off its Specs business. ...

The activist investor said Snapchat has underperformed the Nasdaq by 444 percentage points since its initial public offering, adding that a dollar invested in Snapchat at the time of the IPO would be worth 23 cents today.

“We are taking the unusual step of writing you a public letter because Snap isn’t a typical public company,” Irenic said. “Shareholders like us do not vote for Snap directors and we can’t tell you what to do,” Irenic added.

“We can only attempt to persuade you to do the right things for your shareholder-partners,” the activist investor said in its letter.

Shares of Snapchat jumped following the release of Irenic’s letter, and were trading up 13%, to $4.55, as of midday Tuesday. The stock has fallen 48% over the past year.

The stock jump suggests that, even without any voting rights, shareholders might have some influence at Snap. Underperforming the Nasdaq by 444 percentage points since IPO might suggest that they don’t.

In the 2020s, meme stocks were invented: Some smallish US public companies’ stocks went up a lot, for no particular fundamental reason, because online retail investors got really into them. This was arguably something new in corporate finance, a new sort of asset that a company could have. If a company became a meme stock, it could go sell a lot of stock at very high prices and use the proceeds to pay down debt or hire executives or do acquisitions. Becoming a meme stock had a real dollar value to the company, and as the meme-stock concept became more entrenched and better understood, companies took advantage of it. AMC bought a gold mine. 

If you are a creative financial engineer, though, you might not stop there. That’s just the low-hanging fruit. The deep analysis is that the discovery of meme stocks gave every smallish US public company a valuable new asset. Every company has some probability of becoming a meme stock: There’s always a chance that some small random company will catch fire with online retail investors and its stock will zoom up. And so the trick, if you are are a small public company that is not currently a meme stock, is converting that probability into cash now. You have some asset — your chance of becoming a meme stock — that has a real but uncertain value. Surely some hedge fund would pay you something for it. But how do you structure that deal, and what is the going rate for it?

There’s actually a longstanding, though somewhat disreputable, financial instrument that fits the bill. It’s what you might call a “floating-strike convertible.” Schematically the trade is:

1. A hedge fund gives a company $100 now.
2. In return, the company promises to give the hedge fund $100 worth of stock, whenever the hedge fund asks for it over the next few months or years. (And the hedge fund can split it up: Get $10 today, $20 in a week, $30 more in a month, etc.)
3. But “$100 worth of stock” is calculated in a weird way. For instance, it might be calculated as “$100 divided by the lowest daily stock price over the last 10 trading days.” [1] So if the company’s stock trades at about $2 per share for a week, and then the next week it becomes a meme stock and trades at $5, $10, $15, $69 and $420, then the hedge fund will exercise its option and say “okay I’d like $100 worth of stock, at $2 per share that’s 50 shares.” And then it will sell those 50 shares for $420 each, or $21,000 total.

Usually there are other bells and whistles, but here I want to focus on that lookback. Because it can convert its $100 into stock at the lowest price over some previous period, the hedge fund has a valuable option. In particular that option pays off if the stock goes up a lot, even briefly. [2] The hedge fund hands over its $100, it waits to see if there’s a meme moment, and if there is it cashes out.

If there is not a meme moment, the hedge fund just converts its $100 into stock at the market prices and sells it down over time. This tends to have the effect of depressing the stock price: The hedge fund gets some stock at $2 per share, it sells that stock, the stock goes down to $1.50, the hedge fund gets more stock at $1.50, it sells that, the stock goes down to $1, the hedge fund gets more at $1, etc. These deals are sometimes rudely called “death spiral convertibles.” The bad outcome is a death spiral. The good outcome is that the hedge fund sells the stock into a meme rally.

Again, this technology has existed for a long time, mostly for distressed companies, but in theory you would think that the meme moment would make it better. There is more upside now: If you are a small distressed public company, there is always a chance that your stock will moon and the hedge fund will make a fortune. So in theory a hedge fund should want to assemble a portfolio of these trades, on reasonably fair terms for the companies, and hope that a few of them return 1,000%.

In practice the terms of these things still seem to be pretty unfriendly to the companies, oh well.

In 2023, Bed Bath & Beyond Corp. did this trade with Hudson Bay Capital. I wrote at the time: “What Bed Bath has done here, I think, is that it has sold the right to do meme-stock offerings to some institutional investors.” And in fact Hudson Bay seems to have made an enormous profit on the trade. But Bed Bath’s meme days were mostly behind it, and the stock pretty much death spiraled. 

The Financial Times has a story this week about Algorhythm Holdings, a tiny publicly listed former karaoke company that last month made some bold claims about artificial intelligence for trucking (???), causing (1) trucking stocks to briefly lose about $17 billion of market value and (2) its own stock to briefly go up 450%. (We talked about this at the time, but it didn’t really make any more sense than that.) Conveniently Algorhythm had previously sold some floating-strike convertibles to a guy:

One shareholder well positioned for that surge was Chicago-based John M Fife. The 65-year-old investor and philanthropist has made a habit of backing tiny stocks in hot sectors in a career spanning decades, tens of millions of dollars in profits and multiple brushes with regulators.

Fife’s firm, Streeterville Capital, invested in Algorhythm last summer, in a deal that let it purchase several times the company’s market cap in newly issued shares at a discounted price, over a period of time. As the share price surged, Streeterville offloaded millions of shares, according to Algorhythm’s head of investor relations, in trades that could have earned Fife millions of dollars in profits.

“Clearly, now in hindsight, it’s worked out really well for him,” Atkinson, Algorhythm’s chief executive, told the FT regarding Streeterville’s investment in the firm.

One problem with this sort of thing is that these convertible deals do have a bad reputation, and Fife has irritated regulators a lot over the years. (“In 2020, the SEC charged Fife and five of his companies with unregistered securities dealing, calling him a ‘recidivist violator of the federal securities laws,’” after he “and his companies had purchased such notes from around 135 different microcaps over five years, with the investor targeting hyped sectors such as ‘the marijuana, blockchain, bitcoin, vapour, lithium and gold mining industries.’”)

Another problem is that the trade works best if the company becomes a meme stock, which creates incentives to, you know, make that happen.

Algorhythm is a karaoke company that pivoted to artificial intelligence for trucking. These days, pretty much every sort of company wants to be an AI company, and if a company manifestly isn’t an AI company it might pivot anyway. If I told you “a food science company famous for making monosodium glutamate is pivoting to AI,” you’d be like “sure right whatever, just like the karaoke one.”

But here is a pretty amazing activist presentation that Palliser Capital released this week, arguing that Ajinomoto Corp., which discovered umami and invented MSG, also controls a bottleneck in the AI supply chain:

Ajinomoto’s mastery of complex polymer chemistry — developed in food science — uniquely positioned the company to solve the most difficult materials problem in advanced semiconductor packaging. …

Ajinomoto Build-up Film (ABF) – a high-performance insulating polymer film used in semiconductor substrates, enabling ultra-fine wiring and multi-layer interconnects between chips and circuit boards … has become the industry standard insulating film for high-end package substrates, … with no qualified substitutes at the high end. 

It has been making this stuff since 1999. But, says Palliser, the market doesn’t appreciate it because, you know, food seasoning company:

Despite owning one of the most critical materials technologies underpinning global AI infrastructure, Ajinomoto currently trades at a significant valuation discount to its ABF substrate customers. Resolving the drivers of this valuation disconnect could unlock well over 70% upside to the current share price and support sustained long‑term value creation in the interests of all stakeholders. 

“The rapidly expanding and lucrative Functional Materials (ABF) business is hidden under the Healthcare & Others business segment,” notes the presentation. You don’t even have to pivot! Just talk about the AI more. (Also raise prices: “The fund is calling for a more than 30% increase in prices for Ajinomoto Build-Up Film,” notes Bloomberg’s Alice French.) The world is complicated, and why wouldn’t inventing MSG set you up to build AI chips? 

On the Money Stuff podcast last year, Cliff Asness talked a bit about the evolution of sentiment analysis. In the olden days, you could have a computer scan a document — say, an earnings-call transcript — and try to pick out good words (“profit”) and bad words (“loss”). If there were a lot of good words, the news was good and you’d buy the stock; if there were a lot of bad words, you’d sell. This might work at some rough statistical level, and if you have no better ideas you could give it a try, but it is obviously pretty crude. Sometimes “profit” will be in a sentence like “we had no profit,” and is actually bad. A human, reading the transcript, would be able to pick up on subtleties of sentiment that are not reflected in mere counts of bad words.

Modern machine learning tools are much better at holistically analyzing a document and telling you if the sentiment is good or bad. Pop the transcript into ChatGPT, ask it “how do the executives feel,” and it will give you a nuanced answer. Or you can probably automate that and quickly get a pretty precise sentiment score that can inform your trading decisions. Still, at least some crude rules of thumb work pretty well. If you count 73 swear words in an earnings-call transcript, something has probably gone wrong. You don’t need to read the rest to get a nuanced understanding of sentiment.

Anyway Anthropic accidentally posted some source code of its Claude coding agent, and here is a very funny post from Alex Kim about, among other things, how Claude does sentiment analysis on its users:

userPromptKeywords.ts contains a regex pattern that detects user frustration:

“/\b(wtf | wth | ffs | omfg | … (ty | tiest) | dumbass | horrible | awful | piss(ed | ing)? off | piece of (...) | what the … (broken | useless | terrible | awful | horrible) | ... you | screw (this | you) | so frustrating | this sucks | damn it)\b/”

An LLM company using regexes for sentiment analysis is peak irony, but also: a regex is faster and cheaper than an LLM inference call just to check if someone is swearing at your tool.

The ellipses there are my edits, not the actual regular expression, but you probably get the idea even without the words I deleted. [3] You don’t need that much nuance!

Freed From US Punishment, Wells Fargo Bolsters the Repo Market. Goldman Tells Clients Eager to Short Loans Its Tool Isn’t Ready. Foreign central banks sell US Treasuries in wake of Iran war. KPMG Faces Allegations of Blown Audit in Private Credit Collapse. Blackstone, Ares, Rivals Grilled by Congress Over Private Credit. Record number of megadeals agreed in first quarter of the year. Spice Maker McCormick’s Rise From a Baltimore Cellar to a Global Food Power. Franklin Templeton Agrees to Buy Crypto Spinoff in Digital-Asset Expansion. Modeling Prediction Markets As Exotic Options. Heirs to Bic Pen Fortune Seek Painting They Say Chauffeur Stole. 

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!