The S-1 Is Out. Now Do the Math.

SpaceX filed its full S-1 with the SEC after markets closed Wednesday, May 20. The document confirms what confidential leaks hinted at — and adds details that Wall Street needs to absorb before betting on a $1.75 trillion valuation.

The company lost $4.9 billion in 2025 on revenue of more than $18 billion, according to TechCrunch. Total losses since inception now exceed $37 billion.

Starlink Is Carrying This Company

Starlink generated over $11 billion — more than half of SpaceX's total 2025 revenue, per TechCrunch. Without it, the rest of the business doesn't pencil out.

Launch services are impressive. The reusable rocket business is genuinely world-class. But the financial engine right now is satellite internet, not rockets or AI.

The xAI Problem

SpaceX directed roughly 60% of its capital spending in 2025 — around $20 billion — to its AI division, according to TechCrunch. That division, which houses the Grok chatbot following SpaceX's merger with Musk's xAI, lost billions and grew revenue by only 22%.

For context, that's far below the growth rates at competing frontier AI labs. OpenAI and Anthropic are posting triple-digit growth. SpaceX is pouring the majority of its capex into a division that is currently trailing the AI race.

The filing also discloses legal costs stemming from battles tied to Musk's other companies — absorbing xAI and its social media baggage — that SpaceX says will likely cost it $530 million, per TechCrunch.

The 1-Million-Satellite Promise

Musk posted a mission statement on SpaceX's website declaring the company will launch "a constellation of a million satellites that operate as orbital data centers" to support "AI-driven applications for billions of people."

He's promising to fire off the next-generation Starship once every hour to make it happen, with launches beginning in 2028.

Starship is still experimental. And the cost math is stark.

Robert Zubrin — aerospace engineer, rocket designer, and the man who designed an early NASA Space Launch System prototype — told Forbes that building and launching the existing 10,000 Starlink satellites cost roughly $2 million per spacecraft. Applying similar economics to one million AI satellites puts the bill at approximately $2 trillion.

That's the entire projected IPO valuation of the company. Spent on satellites.

Zubrin told Forbes that if Musk moves forward with this plan as described, it could trigger a "lightning-speed halt" to his business winning streak. He also noted that orbital data centers would be astronomically more expensive than terrestrial ones — a basic physics problem that doesn't care about ambition.

What the IPO Looks Like on Paper

SpaceX has chosen the ticker SPCX for its Nasdaq listing. The IPO is targeting a raise of somewhere between $75 billion (per TechCrunch) and $80 billion (per NPR), which would make it the largest IPO in history — dwarfing Saudi Aramco's $29 billion raise in 2019.

Musk controls 85% of voting power in the company after the IPO, according to NPR. This is NOT a company where public shareholders get meaningful governance rights.

Musk is already one of the world's wealthiest people. NPR notes this deal could make him the world's first trillionaire.

What Mainstream Coverage Is Getting Wrong

Most outlets are leading with the record-IPO angle and treating xAI's orbital data center vision as an exciting moonshot. NPR and TechCrunch did solid work surfacing the actual financials. But neither spent meaningful space on Zubrin's $2 trillion cost critique — which is a serious, credentialed objection from someone who has actually built rockets, not a random skeptic.

The Hill correctly identified that SpaceX's near-term growth depends on Starlink and AI together. But "depends on AI" glosses over the fact that the AI division is currently underperforming every major competitor while consuming 60% of the company's capital spending.

The story isn't "SpaceX is amazing and going public." The story is: can Starlink revenue sustain a company burning this much cash on a bet that hasn't proven itself yet?

What This Means for Regular People

If you're thinking about buying SPCX when it lists, you're buying a satellite internet business that works, a rocket launch monopoly that's real, an AI division that's losing money fast, and a one-million-orbital-satellite dream that a top rocket scientist says could cost the company's entire valuation to execute.

The upside is legitimate. So is the risk.