OpenAI and SpaceX IPOs Ask Public Markets to Price Structural Losses at Scale

The largest IPO prospectus ever filed landed on Wednesday, and within twenty-four hours Polymarket had assigned a 56% probability that SpaceX would close its first trading day worth more than $2.2 trillion ^[1]. The second-order question is not whether the deal prices. It is whose private mark gets repriced when it does, and how brutally.

The reference price problem

Every growth-equity GP, every secondary buyer, every late-stage CFO modelling a downround clause has spent the last three years valuing AI and frontier-tech assets against a missing comp. SpaceX's S-1 and OpenAI's forthcoming confidential filing fill that gap. Within ninety days of pricing, those two clearing multiples become the only numbers that matter in private capital markets. The durable consequence is not the exact level at which SpaceX trades; it is that every 409A valuation, every LP mark, every M&A premium gets recalculated against whatever revenue multiple the public market accepts.

The numbers force the issue. SpaceX disclosed $18.7 billion of 2025 revenue against a $4.9 billion net loss; Q1 2026 alone delivered $4.7 billion of revenue and $4.3 billion of loss ^[2]. OpenAI is running at a $25 billion annualised revenue rate as of February and is committed to $600 billion of semiconductor and data-centre spend over five years ^[3]. If the market clears SpaceX at $1.75 trillion, that is roughly 90x trailing revenue ^[4]. If OpenAI clears above $1.4 trillion, where Polymarket puts a 65% probability, that is 56x annualised revenue against HSBC's 2026 forecast of $34 billion ^[5]. Neither number can be defended through any conventional discounted-cash-flow exercise. Both will become the floor under every subsequent AI and space-economy round.

For board-level capital strategy, this is the actionable point. If you are sitting on a late-stage private position, your IR team needs a memo on the bench by mid-July explaining what the SpaceX clearing price implies for your last 409A. If you are evaluating an acquisition of a venture-funded AI target, the seller's lawyer will arrive at the table in October with the OpenAI multiple printed on the first page of the term sheet. Pricing power on those negotiations shifts to founders in any quarter where the public comps hold.

What the underwriters are actually selling

The framework being constructed in Goldman's and Morgan Stanley's bookbuilding rooms is not a multiple. It is a thesis about substitutability. SpaceX's S-1 claims a $28.5 trillion total addressable market and describes itself as having "the largest actionable total addressable market in human history" ^[6]. That language is not bombast. It is the analytical scaffolding for an explicit argument: there is no second supplier of heavy-lift launch, no second supplier of low-earth-orbit broadband at competitive scale, and after the xAI merger no second supplier of integrated frontier compute, satellite and AI capability. The bull case rests entirely on the proposition that these businesses are not substitutable at any price, which converts the loss line from a problem into evidence of investment in a defensible position.

The OpenAI version of the same argument is weaker, and the underwriters know it. The cost of one million LLM tokens has risen 61% since early March, per Silicon Data's index ^[7], which would normally be margin-positive. But Chinese frontier models are reportedly closing the capability gap at a fraction of training cost, and enterprise AI traffic is shifting toward cheaper inference ^[8]. CFO Sarah Friar has already warned internally that sluggish revenue may constrain the data-centre buildout even as the $600 billion commitment runs ^[9]. Anthropic is reportedly tracking toward Q2 2026 profitability at a fraction of OpenAI's revenue scale ^[10]. The substitutability thesis works for SpaceX. It is contestable for OpenAI.

That asymmetry matters for sequencing. SpaceX prices first because its scarcity story is genuinely scarce. OpenAI prices second into whatever liquidity SpaceX leaves behind, against a moat narrative that hostile analysts can credibly attack.

The index inclusion trade

The least-discussed mechanic in the research pack is the most important for institutional positioning. The Wall Street Journal reported that SpaceX has been pressing index providers for accelerated entry into major equity benchmarks ^[11]. If that succeeds, passive funds become forced buyers at whatever the market is clearing, regardless of valuation defensibility. Apollo's Torsten Slok has flagged the consequence: if SpaceX, OpenAI and Anthropic all list in 2026, the ten largest S&P 500 names could account for close to half the index ^[12].

For any allocator running a benchmarked book, this is a problem worth solving before September, not after. Passive flows will inflate the clearing price on issuance and concentrate single-name risk in vehicles marketed as diversified. The hedge is straightforward in concept: underweight the index against a bespoke basket that reweights mega-cap AI and strips index-forced demand premium. Executing it is harder because the active deviation will look wrong on any day the bull thesis runs. Boards should be asking their investment committees this question now, not in the post-mortem.

The Uber counter

The honest counter-argument is that markets have priced loss-making giants before, and the lesson was humbling. Uber listed in May 2019 as the largest U.S. tech IPO of its era, and the unit-economics scepticism that bulls had dismissed at pricing became the consensus view within twelve months. Cerebras's 70% first-day pop on May 14 is being treated as evidence the window has reopened, but Cerebras priced at $95 billion, not at trillion-dollar scale ^[13]. The lesson from 2019 is that "strategic necessity" premiums survive contact with the second earnings report, not the first.

The counter is that the moat is different. Uber faced credible substitutes; the rideshare market was a duopoly with Lyft, and the regulatory overhang on driver classification capped margin expansion. SpaceX faces no near-term launch competitor at remotely comparable cadence or cost. OpenAI faces real competition but enjoys the enterprise contract incumbency that comes from being the default API in three years of corporate AI pilots. The Uber analogy fits the loss profile and the hype dynamic. It fits less well on the moat. The bear case is alive for OpenAI and weak for SpaceX, and any institutional book that treats them as a single trade is missing the asymmetry.

There is also a governance problem the Uber comparison understates. NYT DealBook flagged an "unusual governance structure" in the SpaceX S-1 ^[14] and OpenAI is converting to a public-benefit corporation mid-flight. Reuters Breakingviews has explicitly labelled the OpenAI deal as carrying a "Sam Altman problem": CFO-CEO tension over IPO timing, missed internal targets, and a structure that has not been stress-tested by public-market disclosure ^[15]. Governance discounts are the first thing the market reprices when sentiment turns. Boards considering their own private-to-public path should note that the discount is applied retroactively, not at IPO.

What to watch

SpaceX index inclusion ruling. Watch for an explicit decision from S&P Dow Jones Indices on whether SpaceX qualifies for fast-track inclusion in the S&P 500 on or before its first trading day. A waiver of the standard seasoning requirement confirms the passive-flow thesis and validates a higher clearing price. A standard waiting period weakens it. Decision likely between IPO pricing and first trade.
OpenAI's S-1 disclosed gross margin on inference revenue. When the confidential filing becomes public, the single most diagnostic number will be gross margin on API and enterprise revenue, stripped of training costs. If it prints above 50%, the moat narrative survives the token-cost spike. If it prints below 30%, HSBC's $64 billion 2027 revenue forecast becomes a margin trap rather than a valuation support, and the OpenAI multiple should compress against SpaceX's.
Anthropic's funding round close. Anthropic was reportedly targeting a $900 billion private round and a Q2 2026 first profit ^[16]. If that round closes near target with profitability confirmed before OpenAI prices, OpenAI's underwriters lose their cleanest comp and gain their most dangerous one. If Anthropic delays or marks down, the AI IPO window narrows materially and OpenAI's September timeline slips into a less hospitable tape.

Sources