SpaceX is down about twenty percent in three days, and Damodaran just rewrote his SpaceX number now that there's an actual prospectus to read instead of, quote, drabs of data. This is AI IPO Watch. Today, the SpaceX hangover collides with a sleeper disclosure story: what AI token costs are about to do to every S-1 in the pipeline. From Mikhail Tegin at Oninvest:
SpaceX shares plunged more than 16% on Monday, and are now off more than 23% over the last three trading sessions. In premarket trading on Tuesday, the stock had dipped below its opening price on its first day as a public company. According to analysts surveyed by Oninvest, the sharp correction in SpaceX shares looks set to worsen market conditions for other large companies lining up for mega IPOs in the foreseeable future.
Twenty-three percent in three sessions, and premarket Tuesday it dipped below its day-one open. So anyone who bought at the $135 IPO price is already underwater. And the people underwater here are the retail buyers who took the narrative at face value, not the institutions that discounted the roadshow going in. If there was a pop at all, it just pushed the pain downstream onto them. Let's be precise about the baseline, though. A 23% slide from where? From the $135 offering price — the document number. That's the one I trust, not whatever pre-IPO figure was floating around on anonymous sourcing. Oninvest's analysts are forecasting a fundamentals-driven repricing within a year. So I'd ask: what does a 23% drop in three days do to the syndicate that priced this? That's exactly where stabilization mechanisms get tested, and the bank list tells you who's on the hook. And don't sleep on the Monday news — they announced $20 billion in senior unsecured notes the same day the stock cratered. Bloomberg reads it as the start of a borrowing program to fund AI projects. You're refinancing a bridge loan into a falling equity story. From Aswath Damodaran at Musings on Markets:
A few weeks ago, I assessed the value of SpaceX ahead of its initial public offering, with the admission that I was making my estimates with drabs of data, some of it coming from unofficial sources. I also promised to revisit my valuation, when the prospectus came out, and now that it has, I will examine how the information it contains has changed my view of the company and its valuation.
Damodaran put out his post-prospectus revision on SpaceX, and this is the part I actually care about — he ran the pre-IPO number on what he literally calls 'drabs of data' from unofficial sources, and now he's got a 277-page prospectus to mark against it. That gap between his pre-prospectus estimate and the revised one is the cleanest illustration I've seen all week of what a reported valuation actually is. Until a document pins it down, it's a negotiating position. And what gets me is he flags that the prospectus is 277 pages plus a hundred-page addendum — bloated, four-to-five times longer than thirty years ago, and not necessarily more informative. Length is not disclosure. Founders love a thick document for exactly that reason. Right — pair that with the Oninvest piece we just hit, and you've got the document reset on one side and the stock down twenty percent in three days on the other. The market and the analyst are both correcting off the same pre-IPO baseline. Here's one from Hacker News:
SpaceX is Tesla on steroids --- bigger and bolder. If you think Tesla with a P/E of 385 is on the verge of taking over personal transportation, SpaceX is probably right for you. Maybe Musk should try his hand at religion. He already has a substantial following.
'Tesla on steroids' — and the line about Musk trying religion because he's got the following. That's the whole problem in one post: when the thesis is faith in the founder, the prospectus stops mattering. At a P/E of 385, you're halfway to a congregation anyway. Damodaran's whole point is you can run the cash flows once the document exists — the 'following' isn't in any of the 277 pages. Okay, so these AI founders are used to going on podcasts and making giant predictions about the future — what actually changes about what they can say once they start moving toward an IPO? A lot — and it starts earlier than founders usually think. Legal guidance from Barakat and Bossa says the restrictions begin in the pre-filing period, before a registration statement ever reaches the SEC. That's Section 5(c) of the Securities Act of 1933: broadly, you can't offer securities before you've filed. So if an executive is out there conditioning the market — hyping the company, floating a valuation, or hinting at a timeline — regulators can treat that as an illegal offer. Reuters reported this week that the coming SpaceX and Anthropic IPOs are going to put those leaders right in Wall Street compliance culture's line of fire. And there's a famous example: before Google's 2004 IPO, executives gave Playboy an interview during the quiet period, then had to include that interview in the prospectus as corrective disclosure. Expensive, embarrassing, and very public. A working paper published last fall through the Social Science Research Council made a related point: SEC disclosure is the main tool regulators have for holding AI companies accountable for the claims they make. So the keynote-stage voice — the big, confident, sweeping stuff — gets much riskier. And one piece this week on Anthropic put it bluntly: the S-1 is the one place where a founder legally can't talk the way they do on a podcast. Securities law won't let them. So is there any wiggle room — are there situations where companies are actually allowed to go out and talk to investors before everything is officially locked down? Yes — it's called 'testing the waters.' McDermott Will and Emery put out guidance this week on how founders are supposed to handle those pre-IPO investor meetings without inviting SEC scrutiny: what materials can be shared, who can see them, all of that. So as these AI IPOs move along, watch for anyone wandering outside the guardrails. Reuters made the funny-but-real point that even what a founder wears to a roadshow has historically shaped investor confidence. The pressure to perform is still there; the legal room to talk is much smaller. Here's JDSupra:
The fast-changing economics of artificial intelligence (AI) use by public companies will require public companies’ disclosure obligations to change with them. AI model providers, including Anthropic, OpenAI, Microsoft, and Salesforce, are moving away from flat-fee, unlimited-access subscription models toward usage-based pricing measured in tokens. A “token” is the basic unit of AI computing.
Here's the sleeper story of the day. JDSupra's flagging that token-based AI pricing — with Anthropic, OpenAI, Microsoft, and Salesforce moving away from flat fees — may trigger MD&A disclosure under Item 303. And the line that should make every CFO sweat: companies are blowing through their annual AI budgets in three months. CFOs may want it as a footnote; investors should hear it as a gross-margin problem hiding in cost of revenue. What I like about this is the disclosure mechanics and the cost structure are tied together. If token spend becomes a material, volatile line item, the risk factors in every AI S-1 in the pipeline get rewritten. Right now, the roadshow deck shows you a clean recurring-revenue curve. It doesn't show you that your single biggest variable cost is metered and exploding. Item 303 says it has to. Right — and that's the part founders will fight. Disclosing token unit economics means talking about per-query margin they've been keeping deliberately fuzzy. Suddenly you can't wave at 'AI efficiency gains' anymore. Which is the whole game, Eric. A number that's not in any document can't be challenged. Force it into the 10-Q and the volatility becomes the investor's problem to price — which is exactly where it belongs. If AI IPO Watch helps you stay ahead of the market, take a moment to subscribe and leave a review wherever you're listening. It really helps other people find the show.
You'll find links to every story we covered today in the show notes, so if one caught your ear, you can dig into the source material there. That's AI IPO Watch for today. This is a Lantern Podcast.