SpaceX's $75 billion IPO on June 12 was priced as a rocket company. It now trades as something else entirely. Its S-1 filing with the SEC claims a $28.5 trillion total addressable market.

Of that figure, 93% — $26.5 trillion — is attributed to AI, not launch or satellite services, according to Augment's analysis of the filing. The offering itself set a record, surpassing Saudi Aramco's 2019 raise. Shares closed up 19% on debut before adding another $100 billion in after-hours trading, per CNBC's live IPO coverage. The company capital allocators are underwriting is no longer the one that put Starlink in orbit. It is a compute and media business that happens to own the rockets.

Four implications for the next board meeting:

  • The valuation gap is the story. Morningstar's discounted-cash-flow estimate of $780 billion sits at a third of the market's $2.1 trillion debut valuation, per CNBC's SpaceX stock quote page. The divergence reflects genuine disagreement about whether AI infrastructure deserves hyperscaler multiples or aerospace ones.

  • Capital is already voting. In Q1 2026, 76% of capital deployment went to AI infrastructure and data centers, not rockets. That share equaled $7.72 billion, according to Futurum Group's analysis of the S-1. The satellite business is now the side hustle funding the compute buildout.

  • SpaceX is becoming a landlord. A disclosed $1.25 billion-per-month compute-leasing deal with Anthropic, reported by Futurum Group, puts SpaceX in the data-center rental business at hyperscale.

  • Governance risk is structural, not disclosed as a footnote. Musk holds 82.4% of voting power through super-voting shares, per the company's Form 424B4 prospectus. Public shareholders are funding the strategy, not setting it.

Capital allocation backs up that framing across every disclosed figure. The underwriting model on the table is AI infrastructure, aerospace, or something priced on the assumption that Musk's next venture bails out this one.

The engineering question buried in SpaceX's S-1 isn't whether Starlink generates cash. It's whether launching data centers into orbit is a real capacity plan or a slide that survived diligence because nobody on the deal team could price it.

What's disclosed is thinner than what's implied. The S-1 states SpaceX directed 76% of Q1 2026 capital deployment into AI infrastructure and data centers rather than rockets or satellites, a total of $7.72 billion, per Futurum Group's breakdown of the filing.

KraneShares' summary of the S-1 puts more than 60% of 2025's $20.7 billion capex into AI infrastructure and Colossus clusters specifically. The orbital play is a stated intention. Capital allocation backs up the framing that it lacks a line item with committed capital attached to it.

The $1.25 billion-per-month Anthropic compute lease is the more interesting operational detail. It signals SpaceX has decided to be a landlord before it's decided to be a foundry, per Futurum Group's coverage of the S-1.

The plan to manufacture its own GPUs is the line worth watching for anyone who has shipped custom silicon on a timeline. Musk told investors on the earnings call that Starship could be caught "as soon as the next flight," then walked that back to "a few months" a few weeks later, according to The Register's report on the timeline reversal. Running two unproven hardware programs concurrently asks the balance sheet to absorb two schedule risks that don't share a failure mode. That is a harder ask than either program alone.

The number that should give investors pause is not $28.5 trillion. It's 2023. That is the year SpaceX's Form S-1 filed with the SEC on May 20, 2026 begins treating three separate businesses, rockets and Starlink, xAI's Grok models and Colossus compute clusters, and the social network X, as if they had always been one company. They hadn't. SpaceX acquired xAI in February 2026, three months before filing. The retroactive combination, achieved through what accounting analysis at mostlymetrics.com's S-1 breakdown identifies as common-control accounting, is permitted under GAAP because Elon Musk controlled all three entities throughout the recast period. It is also, per that same analysis, the kind of cross-industry consolidation, aerospace, advertising, and AI compute under one income statement, that auditors are likely to scrutinize closely.

The mechanics matter because they determine what the headline revenue figure actually measures. Futurum Group's analysis of the S-1 puts 2025 revenue at $18.67 billion, against a $2.58 billion operating loss. That figure blends satellite subscription revenue, launch services, X advertising, and Grok subscriptions into a single line.

There is no disclosed segment breakdown in the sources reviewed that lets an investor isolate what the space business earns versus what the AI business earns. The operating loss could be concentrated in one division or spread evenly, and the filing as summarized does not say which.

That ambiguity collides with the filing's central claim. SpaceX's S-1 asserts what Augment's analysis of the filing, dated April 24, 2026, reports as a $28.5 trillion total addressable market, with $26.5 trillion, 93% of that total, attributed to AI rather than space or satellite services. The S-1's own language, per that reporting, calls it "the largest actionable TAM in human history." TAM figures in S-1 filings are aspirational by convention; they describe a universe of theoretically reachable revenue, not near-term bookings.

What makes this one unusual is the ratio: a company whose regulatory pedigree, launch cadence, and Starlink subscriber base are all in the space business is asking to be valued 93% on a market it entered three months before filing.

The capital allocation in the offering backs up that framing.

Futurum's coverage of the filing put the IPO's targeted valuation at $1.75 trillion. The stock priced at $135 a share and closed its first day near $161, valuing the company around $2.1 trillion, with after-hours trading adding another $100 billion, according to CNBC's live coverage of the IPO. A DCF model discounts projected free cash flow; a first-day trading price reflects order-book demand at a fixed share count. The two are not measuring the same thing.

Governance compounds the uncertainty rather than resolving it. SpaceX's dual-class structure gives Class B shares 10 votes each against one vote for the Class A stock sold to the public. The S-1 language for that package, quoted in the mostlymetrics breakdown, is unambiguous about the milestone required for full vesting: a self-sustaining human colony of one million people.