Money Stuff: Musk’s Moonshot Merger

Here’s how it used to work, say a decade ago. Google was a website. (Still is.) You could go to the website and type in a question and Google would give you a good answer. This was useful, so Google had a lot of users. Advertisers would pay a lot to reach them, so the website made a lot of money. This had, for Google, huge economies of scale: It answered the questions and serve the ads algorithmically, so as more people came to the website and asked more questions, its revenue increased rapidly but its costs — just some servers in data centers — did not.

So Google had all of this extra money, and it spent some of that money on what were colloquially called “moonshots,” risky ambitious long-term bets with uncertain payoffs, like building self-driving cars and curing death. (Some of this stuff happened in a division that Google called X.) Investor enthusiasm for moonshots waxed and waned over time, but at a high level the strategy made sense. Google had some very smart people who wanted to make the world better. They built a software product that generated a lot of money with low marginal costs, and they used the extra money from that software to try to solve harder, more capital-intensive problems in the physical world. 

If you had told me in 2016 “my company will build rockets and send them to Mars in order to generate free cash flow to subsidize a capital-intensive money-losing website that will use machine learning algorithms to answer people’s questions,” I would have been like: No no no, the other way. You subsidize the space exploration with money from the website. The website is cheap and lucrative and scalable; the space exploration is the expensive bit with uncertain commercial prospects.

But that was a decade ago. Now the most capital-intensive activity on Earth (in space?) is building large language models that answer people’s questions on websites and apps, and “let’s use our space rocket business to subsidize our website’s data centers” is a perfectly coherent thought to have. And Elon Musk is doing it:

Elon Musk is combining SpaceX and xAI in a deal that values the enlarged entity at $1.25 trillion, as the world’s richest man looks to fuel his increasingly costly ambitions in artificial intelligence and space exploration.

The acquisition of xAI was announced in a statement on SpaceX’s website signed by Musk and confirming a Bloomberg News report earlier Monday.

The deal gives SpaceX a valuation of $1 trillion, and xAI a value of $250 billion, people familiar with the matter said. The combined company’s valuation was announced to employees in a memo on Monday, some of the people said earlier.

The company is still expecting to hold an initial public offering later this year, one of the people said. SpaceX had been planning an IPO that could raise as much as $50 billion, in what would be the biggest initial share sale to date.

Here is the statement: 

SpaceX has acquired xAI to form the most ambitious, vertically-integrated innovation engine on (and off) Earth, with AI, rockets, space-based internet, direct-to-mobile device communications and the world’s foremost real-time information and free speech platform. This marks not just the next chapter, but the next book in SpaceX and xAI's mission: scaling to make a sentient sun to understand the Universe and extend the light of consciousness to the stars!

Sure. “Free speech platform” refers to X, the website that used to be Twitter, which Musk bought in 2022 and merged with xAI last year. I wrote yesterday that “it would be satisfying if humanity’s ambassadors to distant galaxies turn out to be humanoid robots made by a corporate successor to Twitter,” because “actually the greatest modern human achievements are funded and motivated by selling targeted internet advertisements.” But arguably it’s the reverse: We must explore distant galaxies in order to fund a website where you can get into fights about politics. Martin Peers writes:

Musk’s statement didn’t even attempt to explain how xAI would help SpaceX launch data centers into space. Let’s not forget, that ambition was already expected to be the focus of SpaceX’s IPO pitch. What does xAI bring to the party? ...

There’s no question this move is financially motivated. Musk may be the richest man in the world, but he is facing the same financial realities the leaders of other AI startups face: It’s very difficult to compete in AI development with deep-pocketed tech giants like Google and Meta Platforms, which own cash machines in their advertising businesses. But merging SpaceX with xAI won’t solve his problems. SpaceX doesn’t make anywhere near the money xAI will need in the next few years—it reportedly generated around $15 billion in revenue last year (Meta generated $200 billion). …

That leaves Musk to persuade investors of his vision for data centers in space, in hopes of raising enough cash in the IPO to fund xAI. 

That is: The combination here is not so much “use SpaceX’s revenue to fund xAI’s capital needs” as it is “use SpaceX’s fundraising ability to fund xAI’s capital needs.” There are a lot of money-hungry AI companies these days, but only one money-hungry rocket company that also does AI. “For xAI investors, this is something of a bailout,” writes Dan Primack. And the Information reports:

While the arrangement should give xAI more of a financial cushion, it complicates what would have been a fairly straightforward IPO for SpaceX as a stand-alone company. The immediate reaction from some SpaceX investors who spoke to The Information was downbeat.

Michael Sobel, co-founder and president of Scenic Management, which buys secondary stakes in private companies, said he’s gotten calls from SpaceX shareholders who believe the company’s story to investors will require more explaining with the xAI combination.

“It’s very much on investors’ minds,” Sobel said. “If it was a clean SpaceX IPO, everyone is super stoked. If it’s a multimerger combination of companies, they’re curious and eager and mostly trust [Musk], but they’re like ‘Hmmm,’” he said. 

But it’s all relative. If you have investors who are super stoked for a trillion-dollar space rocket IPO, throwing in $250 billion of somewhat confusing AI and free speech still leaves them with mostly a space rocket company. It’s 80% cool rockets and 20% “hmmm,” which might be easier than funding that “hmmm” on a standalone basis.

A couple of other points. One is the valuation: “The deal gives SpaceX a valuation of $1 trillion, and xAI a value of $250 billion.” Does it? Those numbers are sort of in the ballpark of recent data points — “xAI raised funds at a $230 billion valuation in January, while SpaceX was set to go ahead with a share sale in December at a valuation of about $800 billion” — but of course no cash changed hands. The valuation is just “SpaceX shareholders get 80% of the company and xAI shareholders get 20%”; obviously it is in both of their interests if the total valuation is as high as possible. If the deal gave SpaceX a valuation of $2 trillion and xAI a value of $500 billion, it would be exactly the same deal, although then the IPO might end up looking like a down round.

Also: If you are going to merge SpaceX and xAI, it’s really really important to do it before SpaceX goes public. After SpaceX goes public, merging it with another Elon Musk-controlled company would raise all sorts of difficulties. You’d want a special committee of independent SpaceX directors to evaluate the deal, ideally with Musk out of the room; ideally the special committee would have the power to say no and to negotiate on price. The special committee would have bankers and lawyers and get a fairness opinion. You’d have to put the merger to a shareholder vote, you’d need to make disclosures to the shareholders in connection with that vote, and the disclosures would have to be true. Anyone who wants to own public-company shares can, so someone would probably own the shares, dislike the merger, and sue.

That person would sue in Texas, where SpaceX is incorporated, and where the law on this stuff is less settled than than it is in Delaware. (Where Musk was sued when he had Tesla Inc. acquire SolarCity Corp. in what also looked a bit like a rescue.) Probably Musk would win this lawsuit too. Still. You’d need to at least, like, disclose the deck that the board got in approving the deal. Did SpaceX’s board, or xAI’s, get presentations on the deal terms and valuation? Did they consist entirely of one slide with a poop emoji on it? 

But if they’re both private, meh, who cares. Musk apparently controls the vote at both of them, so he can negotiate the deal with himself, set the relative valuations himself, sign the merger agreement, close the deal, and then tell the boards and shareholders about it. The Financial Times reports:

When news of formal negotiations started to leak last week, most shareholders were kept in the dark and were blindsided by the speed with which the deal closed, multiple people told the FT.

Investors were briefed on hurried calls by SpaceX financial chief Bret Johnsen and xAI’s Jared Birchall. After delays and poor audio quality, many struggled to hear the scant details about the $1.25tn tie-up.

Yeah but they’re not going to sue. The SpaceX IPO prospectus will be 25% more confusing than it would otherwise be, but that’s way better than putting all of this in a merger proxy. 

Meanwhile OpenAI is more of a normal tech company, in that (1) it has a website and an app, (2) it is trying to make money off that website and app and (3) sure, if there’s time and money left over, maybe it can do some moonshots:

OpenAI is prioritising the advancement of ChatGPT over more long-term research, prompting the departure of senior staff as the $500bn company adapts to stiff competition from rivals such as Google and Anthropic.

The San Francisco-based start-up has reallocated resources for experimental work in favour of advances to the large language models that power its flagship chatbot, according to 10 current and former employees. ...

“OpenAI is trying to treat language models now as an engineering problem where they’re scaling up compute and scaling up algorithms and data, and they’re eking out really big gains from doing that,” one person familiar with its research ambitions said.

“But if you want to do original blue-sky research, it is quite tough. And if you don’t find yourself in one of the teams in the centre, it becomes increasingly political.”

Right, yes, traditionally the way companies work is that mostly they try to make money, not do “original blue-sky research,” and the people who make the money are valued more than the people who don’t. The modern AI business has largely turned this model on its head, but still, someone has to make money sometime.

Elsewhere, “the real risk is — is software dead?”

For years, investors have funneled hundreds of billions into software businesses, banking on steady growth and resilient, recurring revenues. But the artificial intelligence revolution is now testing that foundation. …

While various business models are threatened, software-as-a-service is particularly vulnerable. AI-native firms can often offer quicker and cheaper solutions, meaning companies that once operated in a defensible sector are now at risk of competition from new players.

Anthropic’s Claude Code and other “vibe-coding” startups are disrupting traditional SaaS by allowing users with no coding experience to build software. That’s dramatically lowering the programming skill barrier and undermining rigid, one-size-fits-all SaaS products.

“Technology private equity, in its current form, is dead,” Isaac Kim, a partner at venture capital firm Lightspeed who previously led Elliott Investment Management’s tech private equity business, said in a recent LinkedIn post.

And:

“We call it the ‘SaaSpocalypse,’ an apocalypse for software-as-a-service stocks,” said Jeffrey Favuzza, who works on the equity trading desk at Jefferies. “Trading is very much ‘get me out’ style selling.”

And:

Software companies, larded up with debt after leveraged buyout firms viewed their revenue as relatively predictable, have seen their loan prices drop this week. Investors are becoming increasingly worried that advances in artificial intelligence, including the growing coding capabilities of Anthropic’s Claude, will leave many software products and services obsolete.

I suppose there is a creative-destruction, capital-should-flow-to-its-highest-uses story here, in which the market allocated hundreds of billions of dollars of debt and equity funding to SaaS companies and maybe should have allocated all that money to data centers, oh well.

We have talked recently about prediction markets like Kalshi and Polymarket offering contracts on things like Labubus and sneakers and single-family home prices. At one level, this makes a lot of sense: Stock markets are for trading securities, and commodities markets are for trading commodities, and there is a vast array of other stuff that you might want to trade. Labubus and sneakers and houses are probably not securities or commodities, [1] but people might want to trade them, and prediction markets are trying to become a sort of general fallback trading market. Of course they should and do try to find new things to trade.

At another level, though, I have been annoyed by all of these contracts, because they are structured wrong. The contract you want — or the contract I want — would be along the lines of “at maturity, this contract pays the market price of this Labubu/sneaker/house,” or equivalently “I pay you $1 for every $1 that the price of this Labubu/sneaker/house goes up, and you pay me $1 for every $1 that the price goes down.” That is, I want a linear contract that reflects the price of the asset we are trading. If I am, for instance, hedging my home price, I want to sell a contract that goes down 10% when the price of my home goes down 10%; if I am investing in long-term Labubu growth, I want to buy a contract that doubles when the price of Labubus doubles. This is how stock markets and commodity markets work: Gold futures go down 10% when the price of gold goes down 10%; Tesla’s stock price doubles when the value of Tesla doubles. There are other, more esoteric bets — options with nonlinear payoffs — but the normal bet is the linear one.

But the Labubu and sneaker and house-price bets on prediction markets don’t work like that. Instead, they are binary options, yes/no contracts that pay off $1 if some “event” happens and $0 if it doesn’t. The classic heartland of prediction markets is “event contracts” on elections and, let’s be real here, sports, and elections and sports lend themselves to yes/no binary contracts. “Donald Trump will win the election” pays $1 if he does and $0 if he doesn’t; “the Seahawks will win the Super Bowl” pays $1 if they do and $0 if they don’t. 

And so, when prediction markets extend to bets on houses and Labubus, they work the same way. The “event” is something like “the average sale price of houses in this area will exceed $800,000 next year,” or “the average price of this sneaker on a resale website will exceed $200 next month,” or whatever. If the price of the thing, over some measurement period, exceeds the strike price, you get $1; if it doesn’t, you get $0. The underlying fact that is of interest is a linear variable — the price of something — but the prediction market converts it into a binary variable, “the price will exceed $____.” (Or, rather, a list of binary variables: There will be separate contracts for “the price will be above $180,” “the price will be above $190,” “the price will be above $200,” etc., and you can pick which one you want to bet on.)

I have always assumed that this is, essentially, a glitch. The prediction markets are relatively new, they are moving fast into new territory, their regulatory environment is uncertain, and, importantly, their clearing and credit infrastructure is underdeveloped. [2] Surely they know that the sensible way to trade Labubus or houses is with linear price contracts, and they will eventually get around to building those out. And then, sure, there might also be more esoteric things, call options and put options and maybe even binary options, but those will be a minor sideshow, and the main action will be in the more sensible contracts. Real economic actors who want to hedge real exposures will want the linear contracts, because the payoffs of those contracts will offset changes in their real exposures.

But of course I could be wrong. Here is an alternative possibility: “The essential thing that prediction markets are selling is not the ability to hedge or speculate on real-world phenomena like house prices. The thing they are selling is gambling; it is entertainment. And the most normal and entertaining way to gamble is with all-or-nothing bets. We could make a linear bet on the Super Bowl, where I pay you $1 for each point the Seahawks score and you pay me $1 for each point the Patriots score, but empirically almost everyone prefers simpler all-or-nothing bets on who wins. [3] Similarly, when prediction markets offer contracts on real economic variables, converting those variables into binary bets isn’t a bug, it’s a feature. All-or-nothing bets on whether house prices go up are more fun and satisfying than linear bets on what house prices will be, so that’s what they offer.”

One piece of evidence for that view might be the rise of zero-day stock options: Historically the stock market was a place to put your money so it would grow as the economy grew, but if you want a gambling product, “you double your money if the stock market is up today and lose your bet if it’s down” might be more satisfying.

And now the Wall Street Journal reports:

Cboe Global Markets is in discussions with retail brokerages to relaunch “all-or-nothing” options contracts for individual investors that would vie with prediction markets.

Cboe said the talks with brokerages remain at an early stage. The exchange is also discussing its plans for these new contracts—which could include revamped binary options, sometimes called fixed-return contracts—with market makers that would execute those trades. Traders use binary options to place yes-or-no wagers on a given event. 

Much like prediction-market contracts, these derivatives either pay a set cash settlement or nothing at all. For example, binary call options tied to the S&P 500 index closing at 7000 would pay a fixed amount to the holder if the index settles at or above that level when it expires. Likewise, the holder loses all of his initial bet if the S&P 500 closes below that mark. ...

“Think of us entering the event space, using those products to offer a more simplistic event-style contract like you’re seeing on other platforms,” said Rob Hocking, Cboe’s global head of derivatives.

Right, I mean, I see Kalshi offering “simplistic event-style contracts” on asset prices and think “oh no, that’s wrong, in the long run they’re going to have to offer price-based contracts like the real exchanges do,” but in fact the trend is the other way, and the real exchanges are looking to get into the binary business.

Here’s a trade:

1. Start a company. Issue 100 million shares of stock, 90 million to yourself, 9,999,000 to your buddies and 1,000 to the general public. (The shares held by the general public are called the “float.”)
2. You and your buddies trade a bit of stock back and forth with each other and the general public, making sure that the price never falls below $50 per share. This might cost you as much as $50,000 (to buy the 1,000 shares of float at $50).
3. Now you have a $5 billion company: The $50 share price, times 100 million shares, is $5 billion.
4. Now what? How do you turn that $5 billion company into actual cash? If you just sold the 100 million shares, the price would collapse; it’s only at $50 because you’re propping it up, and you can only prop it up because very few shares trade.
5. But if the company’s market capitalization is big enough for long enough, maybe it will get added to a prominent stock index. Then index funds will have to buy the stock, without any price sensitivity, and who will sell it to them? You. Good trade.

That is: If you are really ambitious at pumping and dumping, the ultimate buyer of last resort — the final dumping place for your stock — is an index fund. Pump until you’re in the index, then dump to the index funds.

Everyone knows all of this and tries to stop it. Stock exchanges might not let you list your company if only a tiny fraction of the shares float publicly, regulators might scrutinize your trading with your buddies, and index providers try not to include companies that have very low floats or that are unprofitable or show signs of manipulation. But maybe you’re clever and can make it work. Maybe you issue 20 million shares to yourself and 79.9 million to your buddies, but your buddies are well-disguised and from the outside that looks like a lot of free float. The Financial Times reports:

The volatile swings of some Indonesian shares have earned them the nickname saham gorengan or “deep-fried stocks”, with their restricted floats and concentrated ownership contributing to rallies that can lift their tycoon owners into the ranks of Asia’s richest people almost overnight.

In recent days, index provider MSCI has shone a light on this practice, threatening to downgrade Indonesia to a frontier market unless it cleans up its act by May. 

The announcement — citing “opacity in shareholding structures and concerns about possible co-ordinated trading behaviour that undermines proper price formation” — wiped more than $80bn off the Jakarta Composite index last week and sent financial regulators scrambling to prevent a reclassification that could trigger billions of dollars in capital outflows. …

Investors say the market is split between stocks that are widely held by international investors and covered by analysts, such as banks, and a group of closely-held tycoon-owned companies that have on paper massively outperformed them.

One way to get this pattern is with “co-ordinated trading behaviour,” i.e. intentional manipulation, but that is not the only way. If the effective float is low enough and retail traders are enthusiastic enough, you can get similar results without trying to manipulate the stock.

Silver’s runaway rally becomes ‘death trap’ for Reddit’s retail crowd. Gold Is Behaving More Like a Meme Stock These Days. What Kevin Warsh Learned Working for Stanley Druckenmiller. Builders Push ‘Trump Homes’ to Win Backing for a Million Houses. Investigation Finds Credit Suisse Had Wider Nazi Ties Than Previously Known. The Crypto-Hoarding Strategy Is Unraveling. Vanguard Slashes Fees Again, This Time on a Quarter of Its Funds. Wall Street Laments Departure of the ‘High Priest’ of the S&P 500. ‘I Lost $300,000 to My Husband’s Secret Day Trading.’ Man arrested on charges of impersonating FBI agent in alleged attempt to free Luigi Mangione.

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!