I haven’t seen anything as stupid as the WeWork IPO document in a very long time — that is, until Elon Musk filed to take SpaceX public. WeWork was a joke. SpaceX is a threat. And if Musk and his bankers have their way, you are going to be their bagholder. Lots of the top-line details leaked long before the S-1 filing itself became public. There’s the rumored valuation of more than $1 trillion. That’s despite the nearly $5 billion in losses last year. The total addressable market (TAM) for SpaceX — the amount of revenue SpaceX thinks it could make if won over what it thinks is its entire customer base — was listed as $28.5 trillion. By way of comparison, the gross domestic product of the US as a whole was a hair over $24 trillion, according to the St. Louis Fed. I guess I could believe that Musk is the Lord and Savior of a bunch of weird polygons. This is absurd nonsense, but it might not matter. Musk is the original financial influencer, and his struggling electric car company, Tesla, trades at more than 300 times earnings. Ford and Toyota both trade at about 11 times earnings. Even Nvidia, a company that is arguably printing money, trades at 33 times earnings. Tesla is a meme stock, and SpaceX is poised to be the next one. Never mind that it is basically a space company plus an AI company plus a social network — a meme stock doesn’t have to make sense. So where do I start? I guess we’re all supposed to pretend it’s 2015 and that Musk cares about humanity, and especially about sending humanity to space. Musk is trying to sell a big company with a big story: his messianic mission to “extend the light of consciousness to the stars,” a phrase that occurs seven times in the S-1. (“Light of consciousness” without its astral accompaniment occurs an additional three times.) There is an artist’s illustration of “Life on Mars.” The people who live there appear to be composed of polygons. I guess I could believe that Musk is the Lord and Savior of a bunch of weird polygons; I’ve seen the Cybertruck. (WeWork guru Rebekah Neumann must be eating her heart out right now — “the power of We” is so quaint by comparison.) Musk knows that his strength is the cult of losers who worship him. That’s why 30 percent of the IPO is reserved for retail investors. As for the grown-ups, well, there’s a Keynesian beauty contest in play; if you know that the loser cultists will buy whatever he’s selling, and that Nasdaq rule changes may get it fast-tracked onto the index, it might make sense to buy into the IPO. You’ll be watching Number Go Up regardless of the underlying value, and you’d be an idiot to leave that on the table. And the more people think that way, the more Number Go Up. Say it with me now, my Keynesians: The market can stay irrational longer than you can stay solvent. In some ways, this isn’t just SpaceX’s IPO. It’s the IPO of financial nihilism writ large. Robinhood profited off financial nihilism, but it wasn’t itself a meme stock. SpaceX is different. It’s worse. And I don’t know how normal people can avoid having it stuffed down their throats. SpaceX: An AI company. This company is called SpaceX, and it’s known for building rockets. The filing is peppered with references to the Moon (74), Mars (63) and “and beyond” (13), as in “Earth’s orbit and beyond” or “the Moon, Mars, and beyond.” But looking at the numbers from its IPO document, this is, by SpaceX’s own admission, an AI company. $26.5 trillion of its $28.5 trillion TAM is AI applications. If that seems awfully optimistic to you, don’t worry: SpaceX excluded the Russian and Chinese markets from its estimates. About $13 billion, or roughly two-thirds, of SpaceX’s capital spending in 2025 went to AI buildout. How did that go? Well, the AI arm of SpaceX lost $6 billion in operations and had revenue of just $3.2 billion. Meanwhile, Anthropic is going to be turning an operating profit of $559 million in the second quarter of this year. Yeah, you read that right: Profit. In the quarter. Is Grok one of the most advanced frontier models? Well, it’s distilled from them, anyway! Of course, you could figure the vibes on that out just by knowing that SpaceX leased out its massive cloud computing operation to Anthropic for what the S-1 reveals to be $15 billion a year. Incidentally, xAI’s government contracts aren’t going so hot, which is possibly a problem for the company’s planned IPO. So let’s talk about Grok, which Verge readers may also know as MechaHitler, “a truth-seeking AI model ... which has emerged as one of the world’s most advanced frontier models,” according to the S-1. This is quite a turnaround from Musk saying in March that “xAI was not built right first time around, so is being rebuilt from the foundations up.” Is it one of the most advanced frontier models? Well, it’s distilled from them, anyway! The details on xAI’s recent deal with Cursor make this look even worse. You may recall that in April, SpaceX announced it made a commitment to maybe buy the AI coding company, which would give SpaceX a way to compete with enterprise AI products made by Anthropic and OpenAI. Well, if the deal goes through, existing shareholders will be diluted to the tune of $60 billion. If it does not go through, SpaceX pays Cursor $1.5 billion and also lets Cursor use more than $8 billion of compute. This does not suggest a strong negotiating position for SpaceX. Greg Brockman and Ilya Sutskever, two OpenAI cofounders, felt in 2018 that Musk “really hasn’t done his homework [on] AI / AGI.” The filing notes all the places SpaceX is now under investigation for Grok’s production of nonconsensual sexualized images, including those of children. Three lawsuits are called out specifically in the filing, two of which are attempting to achieve class-action status. The timing of this filing is a little bit funny. Just last week, Musk lost his suit against Sam Altman and OpenAI, which accomplished basically nothing except revealing how bad Musk is at AI. Greg Brockman and Ilya Sutskever, two OpenAI cofounders, felt in 2018 that Musk “ really hasn’t done his homework [on] AI / AGI.” That appears to still be true today. SpaceX said that the AI unit containing X and xAI generated only $818 million in revenue in the first quarter of 2026. By way of comparison, Twitter alone made $1.2 billion in the first quarter of 2022, or about 30 percent more, before Musk bought it. A remarkable business mind, truly. Rocket to the crypt. Doubtless you are thinking, well, but what about space? There’s plenty of AI bullshit there, and we’ll get to it later, but in this case it all hinges on Starship, which has so far been prone to unexpected explosions. Starship is clutch for launching the heavier versions of Starlink satellites, some of which are currently sitting around gathering dust as they wait for their rides into orbit. Some NASA (and other) government contracts hinge on it, too. When the filing dropped, the Starship prototypes launched had barely carried more than the Falcon 9, the lowest-end rocket SpaceX has. You may be thinking that’s the kind of thing that would be disclosed in an IPO filing, but you’d be mistaken. “Starship V3 is designed to deliver 100 metric tons to space in a fully reusable configuration while enabling rapid turnaround times.” I am going to skip “fully reusable” except to point out that it will limit capacity if it is even achievable. Instead of disclosure about what SpaceX had actually achieved to date with Starship, the language we see is this: “We expect to commence deploying our next-generation V3 satellites, designed to offer one Tbps of downlink capacity per satellite, using Starship in the second half of 2026. We expect that a single Starship launch will be capable of deploying up to 60 V3 satellites.” “It’s Musk math.” Okay, so the V3 did launch two days after the filing, on May 22nd, with a sort-of successful mission. I say “sort-of successful” in that it did not blow up on launch and did manage to reach space. The problem was, Starship couldn’t keep all its candles lit — one of its engines failed. Its booster also exploded on return, but I am less concerned about this. Just getting the satellites in space is the main problem. This test flight deployed 20 dummy satellites. I don’t know how many metric tons that is, and it seems SpaceX didn’t say, but it’s still shy of the goal of 60 satellites. Again, this is the sort of thing it feels like you should tell investors about? You know, how your key rocket that is required for all your big plans is actually working? Speaking of investors, there is an irritating detail I noticed in the S-1. Assume the timeline in the S-1 is correct — we have no reason to assume this, by the way; SpaceX’s delays are legendary — and let’s do some math. The V3 satellites Musk is talking about weigh 2,000kg, according to an FCC filing. Sixty of them add up to 120,000kg, or 120 metric tons. If Starship works as advertised, delivering 100 metric tons, that is still not enough for all 60 V3 satellites. “It’s Musk math,” says Chris Quilty, the cofounder and co-CEO of Quilty Space, an analytics firm. He notes that the V3 satellites may not weigh exactly 2,000kg, and that if SpaceX stacks them so there’s no wasted volume, 60 satellites may be possible. Sure, maybe. Hard to know without details in the filing, though. If everything hinges on this ship, I expect more fucking details on the ship, you know? There is also some outright fantasy in this section: the vision of Starship as a point-to-point transportation system. Revenue from the launch business decreased by more than a quarter in the first three months of this year, primarily because there were fewer customer launches. One possible theory of SpaceX was that if someone made rocketry much cheaper, many more people would want to shoot shit into space. The paperwork suggests that SpaceX’s biggest customer, when it comes to rockets, is SpaceX. There is also some outright fantasy in this section: the vision of Starship as a point-to-point transportation system as imagined by Elon Musk in 2017. In-space manufacturing. Space tourism. Asteroid mining. Sure. Whatever. Starship was supposed to have its first crewed flight two years ago. Obviously, that hasn’t happened. But people like fantasy I guess, so it’s there in the S-1. Starlink has also been a cost center, though it hasn’t been as bad as AI. In 2025, the company spent $3 billion; for the first quarter of 2026, $930 million. The rocket is the lynchpin of SpaceX’s moneymaking plans, so it sort of has to work. The business and fantasy of space. There is one viable business in this IPO. It is Starlink, the satellite internet provider, which brought in more than $11 billion in revenue last year. “The reality is that Starlink is the cashflow machine that will fund the xAI and SpaceX Starship business,” says Quilty. Starlink is an actual good business, even if the revenue reported in the S-1 is less than both Morgan Stanley’s projections and what SpaceX told potential investors earlier this year. That may be because it was heavily discounted, and consequently, the revenue per subscriber declined by about 25 percent. I don’t know, and the filing doesn’t say, what that will mean for retaining those subscribers. SpaceX has stapled its unsuccessful AI dreams to the successful business. But Starlink emerged at the right time — as TV was losing dominance, meaning less revenue for the geostationary satellite TV providers. Though it started with consumers, it’s proven itself to enterprise customers as well, Quilty notes. Starlink has a significant lead on all its competitors. Maybe SpaceX is juicing its numbers for an IPO, but that hardly seems necessary. And while some critics note that there’s no cost of deorbiting written into the S-1, Quilty tells me that at the low orbit at which the Starlink satellites are now being sent — they’re moving down to 380km above Earth — those objects are functionally self-cleaning. That’s because the closer an object is to Earth, the less time it takes to decay; above 700km, it may take years, but at 300km it’s weeks to months, he says. SpaceX has stapled its unsuccessful AI dreams to the successful business using Starlink as a guide. The idea is: data centers in space, with Starlink beaming the data back down. The pitch is that it’s easier to get solar energy in space, so that will solve AI’s energy bottleneck. (Presumably also the environment will help with cooling, which is often done, controversially, with water. I suppose you could radiate heat to space like the James Webb Space Telescope does?) “Our goal over time is to launch 100 gigawatts of compute to space each year,” the filing reads. In case you are wondering, some variation of the phrase “The Sun contains approximately 99.8% of the solar system’s energy” occurs four times in this document. I want you to look at how stupid that is. You could say “almost all” or “more than 99 percent” and get the job done, but it wouldn’t make you sound like a real nerd. No, you need to really drive home that the decimal place is still just not quite accurate enough for your engineering needs. Musk has figured out that if you make a bunch of sci-fi bullshit promises, people can’t really examine them carefully for plausibility. This kind of speaks to what makes my hackles go up. A very cynical way of reading this filing is that Musk has figured out that if you make a bunch of sci-fi bullshit promises, people can’t really examine them carefully for plausibility the way they might for, I don’t know, regular data centers on Earth. So let’s talk about those. Musk, who claims to want to save humanity by spreading us out among the stars and whose car company is premised on renewable energy, bought another $2.8 billion of polluting gas turbines to power SpaceX’s data centers, and tried to bury that particular statistic in a footnote. SpaceX is already being sued for its use of turbines. It’s been granted permits for 15; it’s using 46. Those turbines even show up in the risk factor section because if an injunction is granted or the permits are revoked, the existing bad AI business will suffer. Exciting stuff! Quilty, the space analyst, is more upbeat about data centers in space than I am, viewing them as a mid- to long-term opportunity for the company that hinges on Starship’s success. Musk is not the only person hyping them; Jeff Bezos is excited about them, too, and there are several smaller companies making their own attempts. There are good reasons to be skeptical, which are, in brief: launch and repair costs, the difficulty of replacing air and water radiation cooling, the potential dangers of space debris and solar flares, slower data transfer, and the fact that Earth-based facilities will always be more cutting-edge. To me, the excitement about space-based data centers suggests typical Silicon Valley groupthink that gave Musk a chance to execute a hail-mary pass.

Source: The Verge

Read Original Source →