SpaceX shares began trading on the Nasdaq this month under the ticker SPCX, the headline number was historic. The rocket maker priced its initial public offering at $135 a share and raised roughly $75 billion, before underwriters exercised their over-allotment option and lifted the total to about $85.7 billion, according to TechCrunch. It was the largest IPO ever recorded, several times the size of Saudi Aramco’s 2019 listing. By the close of its first trading day the company was worth more than $2 trillion, and Musk had become the world’s first trillionaire.
The size of the windfall matters less than where the money is going.
A rocket company’s answer to a cash problem
In February, Musk folded his artificial-intelligence company, xAI, into SpaceX. The all-stock deal converted each xAI share into a fraction of SpaceX stock and created a combined entity valued at roughly $1.25 trillion at the time. The stated reason was technological, pairing xAI’s models with SpaceX’s launch capacity to build data centers in orbit. Money was the deeper motive. xAI, whose flagship product is the Grok chatbot, spends cash at a punishing rate, and the merger turned it into a centerpiece of the IPO pitch.
The scale of that spending is striking. A regulatory filing detailed by Data Center Dynamics showed xAI losing about $2.4 billion in the first quarter of 2026, up from roughly $936 million a year earlier. The company had already spent $12.7 billion on AI infrastructure in 2025 and burned through another $7.7 billion in the opening quarter of 2026, a pace approaching $30 billion a year. Its 1-gigawatt Colossus cluster in Memphis, which Musk calls the largest data center on the planet, anchors that outlay.
Strip away the space-age framing and the problem is a shortage of money, which the IPO exists to solve.
Why xAI cannot go it alone
The competitive math explains the urgency. xAI is chasing rivals with far deeper pockets. Microsoft throws off well over $160 billion in annual EBITDA. Alphabet generates comparable cash on top of a market value north of $4 trillion, and Meta has guided to capital spending above $100 billion this year alone. Those companies can fund their AI buildouts internally at a scale xAI cannot match alone, and Musk has said the coming two to three years will decide whether his company secures the capital it needs to compete.
SpaceX brings what xAI lacks. It earns revenue of an estimated $22 billion to $24 billion in 2026, most of it from Starlink, and it now has the public market as a source of capital. The company has said about $20 billion of the IPO proceeds will retire legacy debt tied to X (formerly Twitter) and to xAI, with the rest split between AI computing power and the company’s existing space business.
The offering ran under the internal name Project Apex, backed by a syndicate of 21 banks. Morgan Stanley, Goldman Sachs, JPMorgan, Bank of America and Citigroup led it, and a sizable block of shares was reportedly set aside for retail investors.
The money-go-round
Critics see something more circular at work. Musk used SpaceX to buy xAI from himself, and the wider AI economy increasingly runs on companies financing one another. Nvidia invests in firms that turn around and buy Nvidia chips. Musk’s own empire runs a version of the same loop. SpaceX committed $2 billion to xAI in mid-2025. Tesla added another $2 billion despite unease among its shareholders, and it also sold roughly $430 million in Megapack batteries to the AI company in 2025.
One line in the IPO filing, reported by Data Center Dynamics, captures the dynamic. The rival lab Anthropic is reportedly set to pay SpaceX about $1.25 billion a month for three years to rent space in xAI’s data centers, which would put a direct competitor in the position of helping bankroll the infrastructure it competes against.
Data centers in orbit
The long-term vision sounds like science fiction. In late January, SpaceX asked the Federal Communications Commission for authority to operate a constellation of up to one million orbital data center satellites, roughly 100 times the size of the current Starlink network. The pitch rests on a simple claim: solar power and the cold of space make orbit the cheapest place to run AI computing. Musk has predicted that within about three years, orbit will beat the ground on cost.
Days before the IPO, Musk showed off the first hardware. The AI1 satellite delivers about 150 kilowatts of peak compute power and stretches roughly 70 meters from tip to tip, wider than a Boeing 747. Its chip payload is interchangeable, so the platform is not locked to a single silicon supplier. SpaceX has said prototypes could launch in early 2027, with a target of roughly 1 gigawatt of orbital compute capacity per year by late that year.
To supply the effort, the company is building a 10-gigawatt-per-year solar factory and a semiconductor packaging operation near Austin. It has also floated a chip-fabrication project called Terafab, a roughly $55 billion plan structured as a joint venture with Tesla.
The doubts are as concrete as the ambition. The economics hinge on Starship achieving reliable reusability and a high launch cadence to drive costs down, and running data centers in orbit remains unproven against hard limits in engineering and finance. SpaceX’s own filing concedes the present-day weakness behind the interchangeable-chip design: the company cannot currently secure enough chips.
Grok’s regulatory shadow
Grok’s record hangs over the venture. Through early 2026, xAI’s chatbot drew formal investigations across several countries after users employed it to generate non-consensual sexualized imagery of real people, including, according to regulators and watchdog groups, images that appeared to depict minors. Britain’s Ofcom and Information Commissioner’s Office opened probes, and the European Commission launched its own. Authorities in at least eight countries took action, with some governments moving to restrict the service. xAI responded by limiting Grok’s image-editing features. The episode followed earlier controversy over antisemitic outputs from the chatbot in 2025.
By absorbing xAI, SpaceX took on that legal and regulatory exposure too, what one group of directors-and-officers underwriters has called “AI risk migration” into a company that is also a major government contractor.
A familiar governance fight
For longtime Musk watchers, the deal rhymes with 2016. That year he had Tesla acquire SolarCity, a company in which he held large stakes on both sides, and the move triggered self-dealing litigation that took years to settle in his favor. Governance specialists have raised the same conflict-of-interest questions here. The backdrop does not help: a Delaware court has already rescinded Musk’s multibillion-dollar Tesla pay package over doubts about whether the board acted independently.
xAI’s leadership has been rebuilt around Musk loyalists drawn from his other companies. Michael Nicolls, a longtime Starlink executive, became its president. Others arrived from Musk’s wider orbit. Critics, among them commentators and lawmakers, argue that channeling a public company’s resources into his private ventures raises serious accountability concerns. Defenders counter that combining his launch business with frontier AI is a genuine strategic edge rivals will struggle to copy.
The wider AI listing wave
Musk’s bet lands in the middle of an AI listing wave. OpenAI, valued above $850 billion, and Anthropic, valued around $965 billion after a large recent raise and a confidential filing of its own, are both steering toward public debuts. Together with SpaceX, the three could raise close to $200 billion, more than every US IPO combined since 2022. That concentration has fed fears of an AI bubble, and the Nasdaq wobbled in early June even as the SpaceX shares priced.
For the people who got in early, the payoff is already banked. One analysis of the offering argues that investors buying SpaceX at $135 a share are far less likely to see the gains insiders like Musk have already locked in.
