The SpaceX IPO: Is It a Space or an AI Company? (Part 2: The Premium)

This is Part 2 of a four-part series on the SpaceX IPO. In Part 1, we valued SpaceX’s operating businesses at approximately $1 trillion under the most flattering operational-peer multiples, against an IPO target of $1.75 to $2.0 trillion. The $1T gap between the two can only be filled by businesses SpaceX has not yet built at commercial scale.

Let’s look more carefully at that premium.

Part 2: What the Premium Pays For

AI Segment Expansion

The AI segment is the largest single piece of the premium and the one with the clearest near-term test. SpaceX absorbed xAI in February 2026 at a $250B implied valuation in an all-stock transaction with both parties under Musk’s common control. The FY2025 AI segment generated $3.2B of revenue against an operating loss of $6.4B. Q1 2026 capital expenditures alone, which pulled up SPCX GPU installed capacity from 0.8GW to 1.0GW, were $7.7B in the segment. Annualized, the run rate would be above $30B, and would bring SPCX capacity closer to CRWV’s 3.1GW. The build-out logic implies the segment must grow approximately ten-fold to justify the capex.

The Anthropic compute agreement is the central argument that this growth is achievable. Signed just this month (May 2026), the contract provides for payments of $1.25B/mo through May 2029, with Anthropic taking the entire Colossus 1 capacity. At full ramp, the contract generates $15B of annual run-rate revenue, which is roughly four times the entire FY2025 AI segment. The complication is that either party may terminate on 90 days’ notice. For now, this customer concentration renders the AI growth thesis somewhat brittle, though the risk is roughly comparable to that of peer CRWV due to the latter’s customer concentration in Microsoft, and we anticipate Anthropic to ramp on (or even ahead of) schedule, as compute capacity is their number one bottleneck.

Starship at Commercial Scale

Starship is positioned in the S-1 as the deployment vehicle for Starlink’s next-generation V3 satellites, the launch vehicle for orbital AI compute satellites, and more aspirational ventures such as carrying cargo for lunar missions and eventual transport for Mars. The argument is that Starship at full reusability could reduce the cost to reach orbit so much that businesses that are economically infeasible today (orbital data centers being the most prominent) would become viable.

However, as of the S-1, Starship has flown just 11 test flights with no commercial payload delivered, against R&D spending of approximately $3B per year. Most of the expected near-term Starship use is internal SpaceX work, primarily Starlink V3 deployment and AI compute satellite constellations, while the commercial demand pipeline for Starship is small. The named third-party commercial contracts are the Astrolab FLEX rover, Sky Perfect’s Superbird-9 satellite, and Starlab’s planned single-piece space station launch.

Interestingly, the S-1 only gives qualitative commentary on SPCX’s book of U.S. government business, despite the high probability that NASA and DoD/Space Force contracts qualify them as major customers. Investors should look to Boeing’s Starliner program as a cautionary parallel for such fixed-price space-systems contracts. Starliner has accumulated over $1B in cumulative reach-forward losses on a $4.8B fixed-price contract. SpaceX’s Starship-related NASA contracts (including the Human Landing System for Artemis) are similarly fixed-price. If Starship encounters multi-year delays or fails to reach an economical cost target, the fixed-price exposure could create losses on a similar scale.

Coming in Part 3…

Investors considering the SpaceX premium narrative should take a close look at what the S-1 says and declines to say. Part 3 walks through the disclosures that are missing from the prospectus, the competitor SpaceX does not really talk about, the related-party arrangements with other Musk companies, and the governance features that may materially weaken public-shareholder rights compared to peer mega-cap IPOs.

---

We put together this four-part analysis in just a day using Revelata’s deepKPI. You can do this kind of work yourself by signing up for our free tier. No credit card required.