"Coding Is Solved" Meets a Terminal That Can't Stop Flickering
What kind of company is SpaceX? A rocket business? A satellite internet provider? Humanity's ride to Mars? According to its own filing with the SEC, all three answers are wrong. The company that just pulled off the largest IPO in history is, by its own accounting, mostly an AI bet. Here is how a profitable aerospace business turned itself into a cash-burning AI infrastructure play, and why that should make investors look twice.
The Filing Says the Quiet Part Out Loud
When SpaceX filed its S-1 on May 20, the most revealing number was not the valuation. It was the total addressable market. SpaceX pegged its TAM at $28.5 trillion, roughly the entire annual GDP of the United States, and called it the largest actionable market in human history.
Here is the part that should stop you. Of that $28.5 trillion, $26.5 trillion is attributed to AI, with $22.7 trillion in enterprise AI applications alone. Space accounts for just $370 billion. Connectivity, including all of Starlink, is $1.6 trillion. In other words, the company's own prospectus says that more than 90% of its future is artificial intelligence, not rockets and not satellite internet. Analysts have, predictably, described the figure as aggressive marketing designed to ride the AI investment frenzy.
A Profitable Rocket Company Swallowed a Money-Losing AI One
The pivot has a precise origin date. In February 2026, Musk merged SpaceX with his AI venture xAI at a combined $1.25 trillion valuation. Before that, SpaceX was a genuinely healthy business. In 2024 it posted $791 million in net income, remarkable for an expense-heavy rocket company.
Then came the merger, and the accounting tells the story. SpaceX posted a $4.94 billion net loss in 2025 and a $4.28 billion loss in a single quarter in Q1 2026. The xAI fingerprints are on every line. Starlink, the only segment that reliably makes money, generated $11.4 billion in revenue and $4.4 billion in operating profit in 2025, and those profits are now being consumed by AI losses and Starship development. The AI segment alone generated $3.2 billion in revenue but posted a $6.4 billion operating loss.
The Index Committees Split
SpaceX had hoped to fast-track straight into the major indices, which would have forced every passive fund, 401k, and pension tracking those benchmarks to buy the stock automatically. On June 4, S&P Dow Jones Indices refused, declining to waive its requirement that companies post four consecutive profitable quarters. SpaceX does not qualify and will not be eligible for the S&P 500 until at least mid-2027.
That rejection stranded an estimated $14 billion in forced passive buying. The Nasdaq, however, revised its rules to allow fast entry after just 15 trading days, so SpaceX will still get a wave of passive inflows, just a smaller one. Critics, including investor Michael Burry, characterised the fast-track push as a manipulation of the index that strips passive investors of price discovery.
The Compute Deals Have a Strange Smell
If 90% of the company's pitch is AI, the AI business needs to look strong. The deals struck so far raise eyebrows. xAI holds only a tiny sliver of the enterprise AI market, all 11 of xAI's original co-founders have departed, and the infrastructure story has been messy.
The clearest example is the Colossus 1 data center in Memphis. xAI built it with a mix of three different Nvidia chip generations, which created latency problems so severe that real-world GPU utilisation reportedly sat at just 11%, against the 40% or higher that Meta and Google achieve. Unable to use it effectively for training Grok, SpaceX leased the entire facility to Anthropic, a direct competitor, in a $1.25 billion-per-month deal running through May 2029, worth roughly $45 billion. Google separately entered a compute partnership and holds a stake in xAI, the kind of circular financing arrangement that tends to flatter revenue right before an IPO.
Data Centers in Space
The most ambitious piece of the thesis is literally orbital. SpaceX has filed with the FCC to launch up to 1 million satellites functioning as solar-powered AI data centers, with orbital compute targeted to begin around 2028. For scale, the current Starlink constellation is around 9,000 satellites, so this would be more than 100 times larger.
The engineering challenges are formidable. In a vacuum there is no air or water for convection, so dissipating the heat from densely packed GPUs becomes hard. Cosmic rays can flip bits in memory and corrupt calculations, solar particles can fry unshielded chips, and solar flares pose a catastrophic risk, all of which demands heavy radiation shielding that adds mass and cost. There is also a simpler economic problem: GPUs become obsolete every few years, so an orbital data center risks being outdated almost as soon as it is launched, with no easy way to service it.
What Actually Happened on Listing Day
The bet, for now, has paid off spectacularly. SpaceX listed on the Nasdaq under the ticker SPCX on June 12 and rose 19% on its first day, closing around $161 and reaching a valuation above $2 trillion. The company raised roughly $75 billion by selling more than 555 million shares at $135 each, making it the largest IPO in history, eclipsing the previous record held by Saudi Aramco.
Not everyone is convinced. Morningstar values the company at around $780 billion, roughly 55% below the IPO price, citing the tiny public float, index-inclusion mechanics inflating demand, and unproven profitability. CFRA analyst Keith Snyder assigned a sell rating and a $115 price target, calling the growth required in the AI segment to justify the valuation "borderline comical."
The Takeaway
SpaceX was, by any honest measure, a world-class aerospace company with a profitable and effectively uncontested satellite internet business. On the strength of Starlink alone, it could have gone public on a story of steady, defensible growth. Instead, the prospectus asks investors to scroll past eleven pages of rockets and Mars before reaching the financials that reveal the actual plan: floating an AI company inside a rocket company's reputation.
The stock popped on day one, as heavily marketed mega-IPOs tend to. Whether the orbital AI thesis survives contact with thermodynamics, GPU depreciation, and a market where AI compute is rapidly becoming a commodity is the real question, and it is one that listing-day enthusiasm has not answered. None of this is financial advice. But the smell is worth noticing.