21/05/2026
SpaceX Files to Go Public
Inside Elon Musk’s $28 trillion bet on space, satellites, and artificial intelligence
21 May 2026
After more than two decades as one of the most valuable private companies in the world, Space Exploration Technologies Corp. has finally filed the paperwork to take itself public. The S-1 registration statement, lodged with the U.S. Securities and Exchange Commission on 20 May 2026, lays out — in 277 pages of dense legal prose, financial tables, and engineering ambition — exactly what investors will be buying into if they purchase shares under the proposed Nasdaq ticker SPCX.
And what they will be buying is no longer just a rocket company.
A company in three parts
The SpaceX of 2026 operates across three distinct businesses, each large enough to be a major listed company in its own right.
The Space segment is the original business and the part most people still associate with the brand. It includes the Falcon 9 and Falcon Heavy rockets, the Dragon crew capsule, and the still-developing Starship — the fully reusable super heavy-lift vehicle on which much of the company’s long-term strategy depends. Since 2023, SpaceX has launched more than 80% of the world’s mass to orbit each year, with a mission success rate above 99% across roughly 650 orbital launches.
The Connectivity segment is Starlink: about 9,600 satellites in Low-Earth Orbit serving roughly 10.3 million subscribers across 164 countries and territories, including Papua New Guinea. A separate satellite-to-mobile constellation of around 650 V1 Mobile satellites now provides direct-to-phone messaging and voice services in about 30 countries. This is the part of SpaceX that actually makes money — and a lot of it.
The AI segment is the newest, and the most controversial. In February 2026, SpaceX absorbed xAI, Musk’s artificial intelligence company, which had itself swallowed X (formerly Twitter) the year before. With one stroke, SpaceX inherited the Grok large language model, the X social platform, the COLOSSUS and COLOSSUS II gigawatt-scale data centres in Memphis and Mississippi, and a combined user base of roughly 1.3 billion accounts.
The money picture
The financial portrait that emerges from the filing is striking, and not entirely flattering.
For the first quarter of 2026, SpaceX reported revenue of US$4.69 billion against an operating loss of US$1.94 billion. For the full year 2025, revenue was US$18.67 billion and the operating loss was US$2.59 billion. The company is, in other words, a very large business that is currently losing very large sums of money.
Look beneath the headline figures, however, and the picture sharpens. Starlink is enormously profitable. The Connectivity segment generated US$11.4 billion in revenue and US$4.4 billion in operating income in 2025, with year-on-year growth of nearly 50%.
The Space segment roughly breaks even, because every spare dollar is being poured into developing Starship. The AI segment, meanwhile, lost US$6.35 billion on US$3.2 billion in revenue last year — a deliberate consequence of the company’s decision to spend heavily on data centres and model training.
Capital expenditure in 2025 alone exceeded US$20 billion. Of that, roughly US$12.7 billion went into AI infrastructure. In the first quarter of 2026, AI capex was US$7.7 billion — more than the entire Space and Connectivity capex combined.
The message to prospective investors is clear: SpaceX is no longer primarily a launch business that happens to run a satellite network. It is becoming a compute and infrastructure company that happens to own the world’s best rockets.
Musk’s grip on the company
Anyone buying SPCX shares should understand one thing clearly: they will not be buying control. They will not even be buying meaningful influence.
The company will list with a dual-class share structure. Class A shares — the ones offered to the public — carry one vote each. Class B shares, which Musk holds, carry ten votes each. The holders of Class B shares are also entitled to elect a majority of the board of directors. The combined effect is that Musk will personally control the outcome of virtually every shareholder vote and the composition of the board, regardless of how many Class A shares are sold.
SpaceX will be classified as a “controlled company” under Nasdaq rules, an exemption that allows it to avoid certain corporate governance requirements that normally apply to listed firms, including the requirement to have a majority-independent board and independent compensation and nominating committees. The prospectus is admirably blunt about this. Mr Musk, it states, “will have the power to control the outcome of matters requiring shareholder approval.”
Investors who are comfortable with founder-led companies will not be troubled by this. Those who prefer the protections of standard corporate governance will need to weigh the trade-off carefully.
The vision being sold
The strategic argument that runs through the entire prospectus is, in essence, this: artificial intelligence is going to require staggering amounts of electricity, and Earth’s power grids cannot keep up. Between 2008 and 2023, U.S. electricity generation grew at a compound annual rate of just 0.1%. AI data centre demand, meanwhile, is exploding. Something has to give.
SpaceX’s answer is to move the data centres into space. The Sun, the filing notes, contains roughly 99.8% of the energy in our solar system, and space-based solar arrays can generate more than five times the energy per unit area of equivalent terrestrial installations.
The company proposes to build constellations of AI compute satellites in Sun-synchronous orbit, powered by solar arrays, cooled by radiating heat directly into the vacuum of space, and connected back to Earth via the existing Starlink network. It plans to begin deployment as early as 2028.
The total addressable market the company quotes is US$28.5 trillion. Of that, US$26.5 trillion is in AI — a number that includes US$22.7 trillion in enterprise applications, US$2.4 trillion in AI infrastructure, US$760 billion in consumer subscriptions, and US$600 billion in digital advertising. Whether one accepts that figure or not, it is a useful indicator of the scale of ambition.
Beyond orbital data centres, the filing also discusses lunar mass drivers, asteroid mining, in-orbit manufacturing, point-to-point terrestrial travel by rocket, energy production on the Moon and Mars, and what the company calls “digital human augmentation.” These are listed as “Future Markets,” with the candid acknowledgment that several of them “do not exist today.”
Notable deals
Three transactions disclosed in the recent developments section are worth flagging.
First, in May 2026, SpaceX signed cloud services agreements with Anthropic, the AI safety company, under which Anthropic will pay SpaceX US$1.25 billion per month through May 2029 for access to compute capacity on the COLOSSUS clusters. That is a US$45 billion contract over three years, with 90-day termination provisions on either side. It also confirms that SpaceX intends to operate as a compute provider to third parties, not just to its own AI models.
Second, SpaceX has an option to acquire Cursor, the AI coding company, at an implied equity value of US$60 billion. The option is structured so that if SpaceX walks away, Cursor receives a US$1.5 billion termination fee and an US$8.5 billion deferred services fee. If exercised, the acquisition would be paid for in SpaceX Class A shares.
Third, the company is in the process of acquiring spectrum licences from EchoStar covering the AWS-3, AWS-4, and H-Block bands — a transaction approved by the FCC on 12 May 2026 but still subject to closing conditions. Combined with the Terafab chip manufacturing collaboration with Tesla and Intel announced earlier this year, the picture is of a company aggressively vertically integrating across every layer of the AI stack: chips, compute, connectivity, models, and applications.
The risks, in the company’s own words
Every S-1 carries a Risk Factors section, and SpaceX’s runs from page 26 to page 64. The risks the company itself flags are worth taking seriously.
Starship is not yet operational at scale. The company has completed eleven flight tests and expects commercial payload delivery in the second half of 2026. Most of the growth strategy — orbital compute, next-generation satellites, lunar missions — depends on Starship working as advertised.
Regulatory dependencies are deep. Launch licences from the FAA, spectrum authorisations from the FCC, and equivalent approvals in every country Starlink operates in are all ongoing requirements with no guarantees of renewal.
The AI business is described in the filing as “recently formed, still being integrated… subject to integration, ex*****on, competitive and operational risks.”
Several core initiatives — orbital AI compute, lunar economy, Mars transport, human augmentation — “involve significant technical complexity, unproven technologies, or technologies that do not exist.”
The company carries substantial debt. The SpaceX Credit Facility was amended and expanded in May 2026, and a new bridge loan was put in place in March.
What it means
SpaceX is, by almost any measure, one of the most consequential private companies of the early twenty-first century. It has fundamentally reshaped the economics of access to space, brought broadband internet to communities that no fibre or cellular network was ever going to reach, and is now positioning itself to become one of the dominant providers of artificial intelligence infrastructure on the planet — and, potentially, off it.
It is also a company that loses billions of dollars a quarter, depends on technologies that do not yet exist for much of its long-term thesis, is controlled absolutely by a single founder whose attention is famously divided across multiple ventures, and is asking the public to fund the next phase of an extraordinarily ambitious expansion.
The S-1 makes no attempt to disguise any of this. It is, in fact, unusually candid about the speculative nature of much of the growth story. Whether that candour is reassuring or alarming will depend largely on the investor.
The pricing details — share price, total shares offered, exact percentage of voting power retained by Musk — are still blank in the preliminary prospectus. Those will be filled in closer to the IPO date. When they are, the market will deliver its own verdict on whether the world’s most ambitious company is also a sensible investment.
—
Source: Space Exploration Technologies Corp., Form S-1 Registration Statement, filed with the U.S. Securities and Exchange Commission, 20 May 2026.
