The IPO window and the round number

There is a thing companies do when they want to go public, which is wait for the window to open, and there is a thing windows do, which is close. Both of these things are happening right now, and the second one is the part nobody at the roadshow will mention. SpaceX filed its S-1 on May 20 and means to list on the Nasdaq around June 12 at a valuation of $1.75 trillion. OpenAI filed confidentially within days of that, eyeing September and a number with a T in it. Anthropic, fresh from a $65 billion raise that priced it near $965 billion, has lawyered up and is reported to be looking at October. Three of the largest private companies on earth have decided, more or less simultaneously, that now is the moment to sell stock to the public. That is the news. The interesting part is the word simultaneously.

My father came home from the 2008 crash a quieter man, and the one thing he taught me that has paid for itself is that the market is a machine for converting other people's certainty into your loss. So let me start where I always start, by separating the real thing from the trade built on top of it. These are real companies. SpaceX puts more mass into orbit than every other entity on the planet combined; Starlink is a genuine business that actually made money last quarter. OpenAI and Anthropic have built software hundreds of millions of people use and, increasingly, pay for. The technology is not the question. The technology was never the question. The question is what you are being asked to pay for it, and why everyone is being asked at once.

The window is the tell

Companies do not go public when they need the money. They go public when the buying is good. The IPO window is the market's tell — it opens when public appetite for risk is high enough that founders and early backers can sell into strength, and it slams shut the moment that appetite blinks. A founder who has guarded a private valuation for fifteen years does not suddenly crave the quarterly humiliation of public markets out of principle. He does it because the bid is there now and might not be in eighteen months. Read it that way and a synchronized rush of the three biggest names in the cycle is not three independent decisions. It is one decision about the price of risk, made three times.

This is the rhyme I keep hearing. In late 1999 and early 2000 the calendar filled with offerings because the calendar could; the deals were a symptom, not a cause, and the symptom peaked just before the patient did. I am not saying June 2026 is March 2000 — being early is a way of being wrong, and I have been early before, by about eighteen months, and I refuse to launder that into foresight. I am saying that when the highest-quality names in a cycle all decide to cash in their chips in the same quarter, the cycle is telling you something about itself, and it is not telling you that it is young.

Companies do not go public when they need the money. They go public when the buying is good. The window is the tell.

Three trades, not one basket

The lazy move is to bundle these into a single 'AI-and-space megacap IPO' trade and buy or sell the basket. Don't. They are three different trades with three different economics and three different ways to break, and conflating them is how a believer overpays for the weakest and a skeptic shorts the strongest.

SpaceX is, underneath the valuation, two companies stapled together: a launch business and a satellite-internet utility, plus an AI line that is presently a money furnace. NPR's read of the filing puts the picture starkly — in the first quarter the rocket business ran an operating loss of roughly $662 million, the AI segment lost nearly $2.5 billion, and only Starlink turned a profit, about $1.2 billion. Net loss for the quarter: nearly $4.3 billion on $4.7 billion of revenue. Full-year 2025 revenue was $18.7 billion, up a third on the prior year, which is genuinely impressive growth. At $1.75 trillion you are paying somewhere north of 90 times trailing revenue for it. The S-1 itself, with admirable candor, warns of 'a history of net losses' and that much of the value rests on technologies that are 'novel and untested.' Translation: the satellite utility is real and the price is a bet on the rockets and the robots.

OpenAI is the consumer-and-frontier-model trade. Revenue is climbing fast — reportedly near $6 billion in the first quarter — and the company, in Fortune's phrasing, 'remains deeply unprofitable,' having reportedly missed its own internal revenue and user-growth targets. Sam Altman, to his credit, said the quiet part: 'filing for an IPO is different from being ready to go public.' The trade here is whether the most recognizable name in AI can convert ubiquity into margin before the compute bill comes due.

Anthropic is the enterprise trade — quieter, more boring, possibly the better business. Its run-rate revenue reportedly crossed $47 billion this year, and the Journal has reported the company expects to reach its first operating profit. It also carries a risk the others don't: the Pentagon designated it a supply-chain risk earlier this year, which is the kind of headline that does not show up in a discounted-cash-flow model and absolutely shows up in a stock price. A near-trillion-dollar private mark is not a public price, and the gap between the two is where the next twelve months will be decided.

What a trillion is, and isn't

I am structurally allergic to the round number, and I would ask you to be too, because a trillion is not an analysis — it is a destination chosen for how it sounds. $1.75 trillion, 'past a trillion,' 'near a trillion.' These are not valuations arrived at by adding up cash flows. They are anchors, and the entire purpose of an anchor is to make the next number you hear feel reasonable by comparison. When you find yourself relieved that a company is 'only' worth $965 billion, the anchor has done its job and you should check your wallet.

Here is the question the round number is designed to keep you from asking: what has to be true for this price to be the right price? At 90-times revenue, SpaceX has to win launch, win orbital internet against incumbents and governments, and win at robotics and AI — three frontiers, priced as if all three are already settled in its favor. The internet really did change everything, and the people who bought pets.com still lost their money. A great technology and a sane price are different claims, and a bull market's favorite trick is to let the obvious truth of the first smuggle in the unexamined assumption of the second.

The number nobody wants on the chart

There is one figure from this whole episode I cannot stop turning over. Bank of America's Michael Hartnett estimates that adding OpenAI and Anthropic to the existing AI leaders could push the concentration of the top US stocks from around 40% to roughly 48% of total market value — exceeding the dot-com peak, the Nifty Fifty, Japan's 1980s mania, and the Roaring Twenties. The only era with more concentration was the railroad boom of the 1880s. Sit with that. We are pricing in a future more concentrated in a handful of names than at any point in American markets except the one that ended in a railroad bust your great-great-grandparents may have lived through.

And the plumbing strains too. From 2016 through 2025 the entire US IPO market raised about $469 billion. These three deals, at standard float ratios, could need to raise somewhere between $432 billion and $576 billion — a decade of IPO supply compressed into a single quarter. Even great companies have to find buyers for the shares, and there is only so much money standing at the window at once. Michael Burry, who is congenitally incapable of leaving a bubble unpoked, noted the three could raise, inflation-adjusted, as much capital as roughly 300 tech and telecom IPOs did in 2000. He is not saying they are bad companies. He is saying that is a lot of stock to sell into one mood.

Which trade is which

So which is which, if you're holding the question and not just the FOMO? On my read: Starlink is the closest thing here to a real, profitable utility, wrapped inside a SpaceX price that asks you to pay for two more miracles on top. Anthropic is the least glamorous and possibly the soundest business, carrying a political risk the spreadsheet ignores. OpenAI is the brand, the ubiquity, and the open question of whether attention ever becomes margin. None of these is a bad company. Several of them may be bad prices. Those are different sentences, and the people who lose money in the next two years — in both directions — will be the ones who let the round number blur them together.

I keep a file of every 'this time is different' I have been handed, and I have heard it now in four distinct accents. It is usually half true, which is exactly what makes it expensive. The window is open. The companies are real. The prices are a wager that nothing slips. My father would tell you the most dangerous moment is not when everyone is afraid — it is when everyone sounds reasonable. Right up until they don't. Compared to when, is the only question I ever really ask. This time, compared to the 1880s. Mind the gap, and mind the door.