Exit liquidity or war chest?
Stress-testing Ed Zitron's doom speak.
The latest thesis from the loudest AI bear, Ed Zitron, is that the elites are cashing out, timed because insiders have seen something in the AI math before the rest of us.
It’s a seductive read, three of the biggest IPOs in history — SpaceX, OpenAI, Anthropic — aimed for one six-month window does sound like a bell ringing on a late cycle.
So I wanted to know whether it survives contact with the mechanics. I started by stress-testing Zitron’s broader bear case with Fable, where it holds and where it overreaches, then pushed on the cash-out claim specifically: what would insiders selling actually look like, and is that what's happening?
As a pattern-matching check on the "monetising publicly means they know something" logic, I ran it against an unrelated case I can't stop thinking about — Emma Grede publishing a self-help book while sitting on roughly $400 million of mostly illiquid, privately-marked equity in the Kardashian brand complex — to see whether scale changes the read.
It does, and that became the resolving principle: elites are always selling something, so the signal is never the sale itself but its magnitude relative to what they're keeping, and whether the structure looks like exit or escalation.
Run the SpaceX IPO through that filter and the thesis splits cleanly into three layers — one real observation, a plausible inference, and an unfalsifiable motive attribution. And they have very different evidentiary status. Here's the read:
The market-mechanics spine of what follows draws heavily on Polymath Investor's piece on the three IPOs, which is worth reading in full.
The real observation. The IPO clustering is genuinely striking and Zitron didn’t invent it. Three of the most anticipated IPOs in history — SpaceX, OpenAI, Anthropic — are converging on the public markets at once: SpaceX published its prospectus on May 20 and lists Friday, Anthropic announced its confidential filing on June 1, and OpenAI followed on June 8. The pipeline is worth roughly $3.6 trillion at current marks, and all three are expected to debut at or near $1 trillion-plus. Only the SpaceX date is fixed — Anthropic and OpenAI haven’t set timing, and OpenAI has said it “may be a while” — but the directional rush is unmistakable. A sudden surge of mega-supply from price-insensitive sellers at peak valuations is classic late-cycle behaviour — 1999-2000 and the 2021 SPAC wave both looked like this. You don’t need any conspiracy for that to be a bearish signal; the cluster itself is information. That’s the steelman, and it’s legitimate. Polymathinvestor
Where the “cashing out” framing breaks. Here's where it gets interesting, because the prospectus is already out — we don't have to speculate about the mechanics, we can read them. On the escalation side: the offering is a primary raise of around $75 billion at a $1.75 trillion valuation — capital into the company, not out of it — and the Journal's reporting frames Musk's motivation as using the listing to boost xAI (absorbed into SpaceX in an all-stock deal this February) so it can narrow the gap with OpenAI and Anthropic in the funding race. Musk himself is excluded from every early-release provision and locked up for 366 days. A founder raising a compute war chest while binding his own shares for a year is the opposite of an exit; it's a leveraged double-down. But on the exit side, the S-1 discloses an unusually accelerated lockup for everyone else: insiders can sell up to 20% of restricted shares as soon as the first quarterly earnings land (likely August), another 10% if the stock trades 30% above the IPO price, 7% more at each of five intervals between days 70 and 135, and another 28% after the second earnings report — against a thin initial float, a ~30% retail allocation, and fast-tracked index inclusion forcing passive funds to buy. So the honest read is that the structure contains both stories at once: the principal is escalating while the cap table has been handed an unusually fast door out. The thing to watch now isn't the prospectus — it's the actual insider sell-down once those windows open.
Where the 'they've seen the numbers' part fails. This is the unfalsifiable bit. If insiders genuinely possessed damning private numbers on AI math, an IPO is a strange way to monetise that knowledge — going public is precisely the open-the-books exercise that forces disclosure under securities law, with personal liability attached to misstatements. The rational move for someone sitting on hidden data is quiet secondary sales in private markets, which have existed at enormous scale for all three companies for years — SpaceX is consistently among the most actively traded names on private platforms, and Anthropic ran a $5–6 billion employee tender as recently as February. Choosing maximum scrutiny is weak evidence for the concealment thesis. "They know" claims also can't be tested, which is exactly what makes them rhetorically powerful and analytically useless — the same evidence (big IPOs) is equally consistent with capital hunger, founder ego, competitive racing, or genuine confidence. Polymathinvestor
The solvency-math wrinkle
One more wrinkle: the version of this circulating as “SpaceX is the biggest IPO trap in history” is meme-ified Zitron, amplified through aggregator accounts, and may be cruder than his actual written argument. His current piece is making a more specific, checkable claim — that Anthropic has made $375 billion in compute commitments and must hit $174 billion in annual revenue by 2029 to pay for it, requiring at least another $200 billion raised in the next year. That’s a real solvency-math argument you can stress-test, and it’s much more interesting than the elites-know-something framing. If those commitment-to-revenue ratios are right, you don’t need anyone to “know” anything — the strain shows up in the public filings on its own schedule. Where’s Your Ed At
So the honest scoring
The timing-cluster observation is signal. The cashing-out framing is half-wrong on mechanics — the raise is primary and Musk is locked for a year, but the accelerated insider unlocks are real, so watch the sell-down once the windows open in August. And the insider-knowledge layer is narrative decoration.
The falsifiable version of his thesis lives in the S-1s — one of which is already public, with two more on the way. Which, for what it’s worth, is the best data event the AI-skeptic debate has had since it started.
See also: PoliticStack, the MarketStack Terminal and MarketStack The Edit
MarketStack is free today. But if you value my work, you can pledge for a future subscription. MarketStack is an independent, anonymous publication summarising publicly available commentary and views from across financial media. Nothing here constitutes financial advice or a recommendation to buy, sell or hold any security. All views are a synthesis of public information. Past performance is not a guide to future results. This publication is not authorised or regulated by the Financial Conduct Authority. The author writes anonymously in a personal capacity.