The 277-page document SpaceX filed with the Securities and Exchange Commission on May 20, 2026 is simultaneously the most anticipated corporate disclosure in the history of the space industry and a document that, read carefully, contains the refutation of its own central investment thesis. This is not a document about an orbital data center company. It is a document about a satellite internet business carrying a cash-burning AI division toward a public market exit at a valuation that can only be justified by a technology programme the same document describes as potentially never achieving commercial viability.
The S-1 establishes for the first time what SpaceX actually earns and spends. Total 2025 revenue came in at $18.67 billion - up 33 percent year on year. The operating loss was $2.59 billion. Net loss was $4.94 billion. In Q1 2026 the company earned $4.69 billion in revenue and lost $4.28 billion net. The losses are accelerating, not narrowing.
The segment breakdown tells the essential story. Starlink - the connectivity business - generated $11.4 billion in 2025 revenue at a 63 percent segment adjusted EBITDA margin, producing $7.17 billion in operating cash. It is one of the most profitable satellite businesses ever built. Everything else is either break-even or burning cash. The launch business posted an operating loss of $662 million on $619 million of Q1 2026 revenue. The AI segment - xAI, Grok, and X - lost $6.4 billion from operations in 2025 on $3.2 billion of revenue, a loss ratio that worsened in Q1 2026 to $2.5 billion operating loss on $818 million of revenue.
The capital expenditure data makes the AI bet explicit. In Q1 2026 alone, SpaceX spent $7.7 billion on AI infrastructure against $1 billion for space and $1.3 billion for Starlink. The balance sheet carries $23.85 billion in servers and networking, $2.97 billion in data centre infrastructure, and $14.05 billion in construction in progress. SpaceX is, by asset base, primarily an AI data centre company that also operates a satellite internet service and a rocket company. The orbital data centre programme does not yet appear in the asset base at all.
The most revealing disclosure in the S-1 is not a financial statement. It is a customer contract. SpaceX disclosed that Anthropic - a direct competitor to xAI in frontier AI model development - is paying SpaceX $1.25 billion per month for compute capacity at the COLOSSUS terrestrial data centre in Memphis. The arrangement runs through May 2029, with capacity ramping in May and June 2026 at a temporarily reduced fee. Either party can terminate on 90 days' notice.
At $1.25 billion per month, the Anthropic contract is worth $15 billion per year - a revenue stream larger than Starlink's entire 2024 revenue. It is being generated by the same terrestrial data centre infrastructure the company claims will be superseded by orbital compute within two to three years. If SpaceX genuinely believed that timeline, it would not be negotiating multi-year compute leases for ground-based capacity running to May 2029. The contract is a market signal. Both parties understand that the terrestrial data centre is the real product. The orbital data centre is the narrative.
There is a second dimension to note. The Valor compute lease, disclosed separately, involves payments totalling approximately $20.2 billion to a firm run by Antonio Gracias - a SpaceX board member. A company preparing for a public offering is simultaneously entering into multi-billion-dollar contracts with entities controlled by or affiliated with its CEO and board, in a governance structure that leaves Musk with more than 50 percent of voting power post-IPO. Investors buying Class A shares have structurally no ability to challenge these arrangements.
The S-1 confirmed what had been reported but not yet officially documented: Musk will be CEO, CTO, and Chairman of the Board simultaneously after the IPO. He currently holds 93.6 percent of SpaceX's Class B shares, which carry ten votes each, giving him 85.1 percent of total voting power. Following the offering, that percentage will decrease as new shares are issued but will remain above 50 percent, triggering controlled company status under Nasdaq rules.
Controlled company status exempts SpaceX from the exchange requirement that a majority of board directors be independent. Combined with Musk's retention of the authority to elect, remove, or fill vacancies among Class B directors, the structure means that no shareholder vote, no board resolution, and no activist investor campaign can constrain any decision Musk makes about orbital data centres, Terafab timelines, the Mars programme, or anything else. The S-1 acknowledges this directly: Musk "will have the power to control the outcome of matters requiring shareholder approval, including election of all our directors."
Institutional investors have noticed. The California Public Employees' Retirement System, the New York State Common Retirement Fund, and the New York City pension funds raised concerns about the proposed governance structure ahead of the filing, with reports describing it as among the most management-favourable arrangements ever proposed at this scale of public offering.
The S-1 claims a total addressable market of $28.5 trillion across SpaceX's businesses, excluding China and Russia. This figure exceeds the annual GDP of the United States. TAM figures in S-1 filings are notoriously optimistic - they represent the theoretical maximum revenue available if a company captured 100 percent of every market it addresses, with no competition and no pricing constraint. They are not forecasts.
The $28.5 trillion figure tells you that SpaceX is asking investors to price it as the future operator of the global communications infrastructure, the AI compute infrastructure, the space transport infrastructure, and the energy infrastructure of a multiplanetary civilisation simultaneously. Damodaran's independent DCF analysis, which placed fundamental value at approximately $1.22 trillion, required no assumptions about the $28.5 trillion TAM - it was derived entirely from the demonstrated businesses of launch and Starlink projected forward at plausible growth rates. The gap of more than $500 billion between Damodaran's estimate and the IPO price is the market's assessment of the probability and timing of the orbital compute thesis.
Wedbush analyst Dan Ives has been the most prominent public advocate for the SpaceX bull case, framing the offering as the entry point for an eight-to-ten-year AI revolution in which space becomes the natural home for compute. The bull case requires believing several things simultaneously: that Starlink continues to grow at something approaching its current pace, that the AI segment's losses narrow as xAI gains market share against OpenAI and Anthropic, that Starship achieves commercial payload certification, that Terafab's D3 chip closes a five-to-six-generation process technology gap, and that 100 gigawatts of annual orbital compute capacity begins deployment in 2028 - all of which the S-1 itself describes as uncertain or unproven.
None of these requirements is individually impossible. The conjunction of all of them, on the timeline implied by a $1.75 trillion valuation, is where the arithmetic becomes difficult. Investors are not being asked to bet on one of these things. They are being asked to bet on all of them at once, with 85 percent of voting power retained by the person who set the timeline, and no mechanism for course correction if any assumption proves wrong.
The S-1 contains four disclosures that collectively say what the roadshow will not.
First, the Anthropic contract: $1.25 billion per month to a competitor for terrestrial compute, terminable in 90 days, running to May 2029. If the orbital timeline were real, this contract would not exist.
Second, the AI loss rate: $6.4 billion operating loss in 2025 on $3.2 billion revenue. Q1 2026: $2.5 billion operating loss on $818 million revenue. The losses are accelerating as the company races to keep pace with Anthropic and Google.
Third, the Terafab language: "Neither Tesla nor Intel are obligated to remain a part of the project, and we may not enter into any such definitive agreements." The $20 to $25 billion chip factory that 80 percent of the orbital thesis depends on is currently a framework agreement.
Fourth, the orbital programme disclaimer: "significant technical complexity and unproven technologies... may not achieve commercial viability." This is the legal description of the programme that justifies the $500-plus billion premium over Damodaran's fundamental valuation.
SpaceX's actual operational businesses are significant and not in question. Starlink, with 10.3 million subscribers across 164 countries, is one of the most profitable satellite businesses ever built. SpaceX launches more than 80 percent of all mass to orbit globally. The company has genuine, substantial, and durable value in these businesses.
What is under dispute is the gap between the value of those real businesses and the $1.75 trillion the IPO proposes. That gap exists because the orbital data centre narrative has been embedded in the investment thesis, and because investors are being asked to price speculative future revenue from a programme that does not yet have a working satellite, a thermal management solution for GPU-class compute, a radiation-tolerant AI chip at commercial scale, or a revenue model that survives contact with the physics. The S-1's own risk language says so.
The global space economy will reach $1 trillion within a decade. SpaceX will capture a large share of it. The long-range vision of an orbital power infrastructure for the space economy is coherent and may ultimately prove prescient. All of that is true, and none of it justifies 94 times trailing revenue for a company losing $5 billion per year on the promise of a programme its own lawyers describe as potentially never commercially viable.
The S-1 is 277 pages. The roadshow is 30 slides. The distance between those two documents is where the risk lives.
Related Links| Subscribe Free To Our Daily Newsletters |
| Subscribe Free To Our Daily Newsletters |
