Author: Steed’s scarf, source: Sina Weibo

On May 21, SpaceX handed over its ledger.

This company has been run like a black box for more than two decades, and its financial data is top secret. But in order to prepare for a listing as early as June 12, it must submit a nearly 400-page prospectus to the U.S. Securities and Exchange Commission. Sothe outside world saw the aerospace giant’s complete accounts for the first time.

First let’s talk about the conclusion:SpaceX’s revenue in 2025 is US$18.7 billion, which is 4.6 billion more than the previous year. But instead of making money, it lost 4.9 billion. The profit that was just turned around in 2024 was swallowed up by the investment in artificial intelligence in 2025, and a large chunk was added to it.

When it was founded in 2002, SpaceX's goal was simple: Build a small rocket called "Falcon 1" and see if it could put things into orbit. Twenty-three years later, it puts 80% of the world’s weight into space every year, and operates more satellites than all other countries and companies in the world combined. The public quotation for a Falcon 9 launch is US$74 million, and industry estimates indicate that the internal cost is only about US$15 million. This cost advantage makes it almost unrivaled in the launch market.

But the launch business is just the tip of the iceberg. The things in the prospectus that really make people stop and take a closer look are hidden in the back.

SpaceX has made an estimate of the future market size of its business. It said the figure was $28.5 trillion.

What is the concept of 28.5 trillion? The global GDP is about over 100 trillion. SpaceX is effectively saying that the market it targets is more than a quarter of today’s global economy. Of the 28.5 trillion, rocket launches and Starlink communications combined are only 2 trillion. The remaining 26.5 trillion is all bet on artificial intelligence.

There is an unabashed sentence on page 171 of the prospectus: "We believe that we have identified the largest accessible market in human history."

A rocket company bets 93% of its future on artificial intelligence.

This number was not taken casually. SpaceX said its AI market estimates are based in part on the RAND Corporation's global data center computing power demand forecast, plus its own internal deduction of a series of operational parameters such as AI workload proportion, power consumption, utilization, and pricing. You can question whether this number is reliable, but you can't question whether SpaceX is serious about this.

How does it plan to make this money? The answer is:Move data centers into space.

SpaceX’s plan is to use its low-cost and high-frequency launch capabilities to send a large number of AI computing satellites into orbit to form a computing network around the earth. The prospectus makes it very clear that the goal is to one day deploy computing facilities with an operating power consumption of 100 gigawatts to space every year. 100 gigawatts is roughly equivalent to the power generated by 100 large nuclear power plants, and this electricity will be used to drive AI chips in orbit.

The first batch of orbiting AI computing satellites are expected to be deployed as early as 2028.

Why must we go to heaven? The expansion of ground-level data centers is hitting a physical ceiling. Difficulty in site selection, high power consumption, and heat dissipation bottlenecks, these problems can be partially circumvented in space (at least in SpaceX’s narrative). There is an unlimited supply of solar energy in orbit, and radiant panels can be directed toward deep space in a vacuum to dissipate heat. Once a super-heavy rocket like Starship matures, it can send more than a hundred tons of payload into orbit in a single launch, and the cost per kilogram is targeted to be below $185. If this goal is achieved, the economics of large-scale deployment of space data centers will be justified.

There is actually a closed loop hidden here. Starships send AI satellites up, and the AI ​​business makes money to support the research and development and mass production of starships. As the cost of starships continues to fall, AI satellites can be deployed more densely. Two seemingly unrelated businesses bite each other in terms of cost structure.

Of course, the prospectus also provided some basic information. The starship is currently still in the testing phase and must complete a series of test flights before it can start launching the V3 version of Starlink satellites. To turn it into a fully reusable and mature system that can carry people to and from the moon and Mars, there are still a lot of technical problems that have not been solved. The original words in the document are: "These systems involve significant technical, engineering and operational challenges, including developing habitable transportation and surface environments and performing complex on-orbit operations. Solving these challenges requires the development of novel or untested solutions and requires significant capital investment."

Picture 1: SpaceX’s Starship rocket and super-heavy booster stand ready for launch, source: SpaceX

This company dares to claim to control the largest market for AI space computing power in human history, but it also admits that the foundation of this grand narrative still rests on the progress of starship testing. If the starship doesn't work, the rest of the story won't be able to be told.

Another detail worth chewing on is people and money. The prospectus discloses management compensation. Musk’s salary in 2025 is $54,080, a figure that is accurately anchored at the minimum wage set by California for tax-exempt employees. That same year, company president and chief operating officer Gwen Shotwell's annual salary was $1.08 million, but including stock awards, her total compensation package was worth $85.8 million.

After the listing, Musk will retain 85.1% of the company's comprehensive voting rights and serve as CEO and chairman of the board. The prospectus did not conceal the fact that it would be extremely difficult to remove him from this position.

The document also contains a confession about political risks. It mentioned that the current political environment in the United States is highly polarized, and changes in Congress or the president may lead to major changes in government spending priorities, regulatory posture, contracts, and resource allocation. Considering that Musk served as an adviser to President Trump, this statement was less a warning of risks than a subtle acknowledgment that a company’s fortunes have become difficult to separate from someone’s political position.

SpaceX’s estimate of the total addressable market, source: SpaceX S-1 filing

Finally, back to the $28.5 trillion figure. For a rocket company, the space business accounts for only 7% of the pie presented to investors, and the remaining 93% is entirely AI. This means thatSpaceX is trying to complete a fundamental identity transformation from an aerospace company to an AI company, and rockets and satellites are just means of implementation.

The prospectus shows the trump card. Investors are about to spend real money to answer a question: When a rocket company says its future lies in AI, is this a vision or a 400-page bet?