SpaceX Is Real. Its IPO Price Is Mostly Imagination.

The strangest thing about the SpaceX IPO is that the skeptical case does not require calling SpaceX fake. It requires admitting the opposite: this is one of the most impressive operating companies of the century, and the public market is still being asked to pay as if the next three businesses already work.

An initial public offering, or IPO, is the first broad sale of a company’s shares to public investors. SpaceX priced its deal at $135 a share, sold 555.56 million shares and raised $75 billion at roughly a $1.77 trillion valuation, according to Axios¹. That makes the offering less like a normal financing and more like a new asset class arriving by rocket. Starlink, SpaceX’s low-Earth-orbit satellite internet network, is already a scaled broadband business. The launch arm is no longer a science project. And yet, when I run the numbers, I do not think investors are mainly buying the company SpaceX is today. They are buying a scarce, Elon Musk-controlled infrastructure option wrapped in an AI story.

Start with the disclosed business. SpaceX reported 2025 revenue of $18.674 billion, adjusted EBITDA of $6.584 billion and net operating cash flow of $6.785 billion in its BaFin-approved EU prospectus². Adjusted EBITDA is a company-defined profit measure before interest, taxes, depreciation, amortization and other adjustments; it is useful, but it is not the same thing as free cash flow. At $1.77 trillion, SpaceX is valued at about 95 times 2025 revenue and 269 times adjusted EBITDA, by my math using the IPO valuation and the prospectus figures.

That is the whole argument in miniature. A valuation multiple is the price investors put on a company divided by a financial metric such as sales, earnings or EBITDA. Defense contractors and telecom firms do not trade anywhere near SpaceX’s implied sales multiple. Lockheed Martin’s price-to-sales ratio was about 1.6 in early June, according to ChartsView³, and AT&T’s was about 1.2, according to CompaniesMarketCap⁴. Even Nvidia, the defining AI-infrastructure winner of the public market, was around 18 times trailing sales in early June, according to FinanceCharts⁵. SpaceX is being priced at a multiple far beyond a defense prime, a telecom network and even the leading AI chip company.

The strongest real-business case is Starlink. SpaceX’s prospectus says the connectivity segment, driven primarily by Starlink, generated $11.387 billion of 2025 revenue, $4.423 billion of operating income and $7.168 billion of segment adjusted EBITDA. That is not vaporware. The company reported about 10.3 million Starlink subscribers across 164 countries, territories and other markets as of March 31, 2026, up from about 5.0 million a year earlier, according to the same prospectus². In plain English: SpaceX built the rockets, launched the satellites, sold broadband subscriptions and turned that into a multi-billion-dollar segment profit measure.

But Starlink also shows why the valuation stretches past evidence. SpaceX disclosed that monthly Starlink subscriber ARPU, or average revenue per user, fell from $91 in 2024 to $81 in 2025 and from $86 in the first quarter of 2025 to $66 in the first quarter of 2026, mainly because of international expansion and lower-priced plans, according to the prospectus². Lower ARPU can be rational if volume and cost reductions more than offset it. Still, the unresolved question is not whether Starlink can produce segment EBITDA while growing. It is whether it can produce durable free cash flow after satellite replacement, ground infrastructure, spectrum costs, user terminals and global customer support.

The launch business is the second pillar. SpaceX reported 165 Falcon launches in 2025, of which 43 were customer launches and 122 were internal launches, according to its prospectus². That is exactly why SpaceX is not comparable to a generic telecom operator. It owns the deployment engine for its own network. The launch arm lowers Starlink’s rollout risk, creates internal demand and gives SpaceX a cost curve rivals cannot easily copy.

But this cuts both ways. Internal launch cadence is not the same as external revenue. SpaceX says its Space segment revenue reflects customer launches and customer activities, while the company allocates significant launch capacity to connectivity and expects to allocate significant capacity to AI in the future, according to the prospectus². So launch is best understood as a strategic cost advantage for downstream bets, not a stand-alone business that can justify a trillion-dollar valuation by itself.

The AI piece is not a footnote. Orbital computing means putting data-center or AI-compute hardware in orbit rather than in buildings on Earth. The pitch is seductive: near-continuous solar power, space-based data processing, lower Earth-side energy and water strain, and eventually compute nodes tied to satellite networks. SpaceX’s own filing says the AI segment had a 2025 operating loss of $6.355 billion and segment adjusted EBITDA of negative $1.237 billion, while AI capital expenditures hit $12.727 billion in 2025 and $7.723 billion in the first quarter of 2026 alone, according to the prospectus². A division spending at that scale is not an “upside optionality” slide tucked in the appendix. It is a capital allocation priority.

The technology is also far from proven. The Government Accountability Office’s April 2026 assessment of space-based data centers said smaller systems that process data generated in space may be closer to maturity, but large data centers for training AI models remain unproven and would need solar arrays larger than any launched and assembled in space as of April 2026; the same GAO report⁶ flagged cooling, communications, radiation, collision risk, frequency coordination and orbital interference as major challenges. Cooling matters because a data center is basically a heat factory, and space is a terrible place to dump heat quickly. Power matters because AI training burns through electricity at industrial scale. Latency matters because some computing tasks are useless if data cannot move fast and cheaply between Earth and orbit.

This is where I part ways with the clean bullish story. If orbital AI is technically unresolved, and if Starlink’s steady-state free cash flow is not yet proven, then the IPO price must be carrying a huge assumption about future markets. That does not make the stock doomed. It makes the valuation more like a venture-capital mark transferred into public markets.

The market-structure story matters too. A sovereign wealth fund is a government-owned investment fund, often managing national savings from commodities, reserves or fiscal surpluses. Public reporting has pointed to intense demand from retail investors and institutions, but the clearest hard number is retail: everyday investors submitted more than $100 billion in purchase orders, according to Axios⁷. Another Axios report said up to a quarter of the shares were earmarked for smaller retail traders and individuals, citing the Financial Times, and framed the deal as a test of Musk’s ability to convert attention into capital for his companies (Axios⁸). That does not prove irrationality. It does prove scarcity was part of the product.

Index inclusion is the other mechanical force. It means a company is added to a benchmark such as the Nasdaq 100 or S&P 500, which can push index funds and exchange-traded funds to buy the shares. S&P Dow Jones Indices decided not to fast-track megacap IPOs and kept its 12-month trading requirement, while Nasdaq had already changed guidelines to expedite the addition of large IPOs to the Nasdaq 100, according to the Associated Press⁹. Reuters reported that S&P 500 inclusion would have brought an estimated $10 billion of passive inflows under J.P. Morgan assumptions, while Russell 1000 and Nasdaq 100 inclusion could draw roughly $4 billion and $4.3 billion, respectively (Reuters via Investing.com¹⁰). Those are not huge next to a $1.77 trillion valuation, but they are very large relative to the initial float.

So my answer is: SpaceX’s IPO is not mainly a space bet. It is not purely an AI bet either. It is a market liquidity event built on a real space-and-connectivity business, with AI supplying the story that lets investors justify paying far beyond current cash flows. The real SpaceX lowers the downside compared with empty space startups. The imagined SpaceX creates most of the upside being priced today.

The next test is specific. Watch the first two post-IPO earnings reports for three numbers: (1) Starlink free cash flow after connectivity capex, not just segment adjusted EBITDA, (2) AI capex as a share of total capex, and (3) Starlink ARPU versus subscriber growth outside North America. If Starlink’s cash generation expands while ARPU keeps falling and AI capex begins to flatten, the bulls will have earned more of this valuation. If AI spending keeps absorbing cash and Starlink’s replacement economics stay murky, the IPO will look less like the birth of a new infrastructure champion and more like the AI boom’s most spectacular liquidity transfer.