The world's most expensive IPO just revealed that Elon Musk's AI bet is burning cash faster than his rockets burn fuel.
The Summary
- SpaceX filed for what could be the largest IPO ever, targeting a $2 trillion valuation with Musk retaining 85.1% voting control through a super-voting share structure
- The rocket company reported a $4.28 billion net loss on $4.69 billion in revenue for Q1 2026, but the real story is xAI: Musk's AI subsidiary burned $6.4 billion in 2025 with no end to the spending in sight
- SpaceX set aside over $500 million for potential litigation losses, partly to cover complaints that Grok's "spicy mode" created sexualized images
- First public look at the financial reality of competing in frontier AI: you need rocket-company money to build rocket-company AI
The Signal
SpaceX's IPO filing just pulled back the curtain on the most expensive AI bet in history, and the numbers are staggering. While the rocket business lost $4.28 billion on $4.69 billion in revenue in Q1 2026, the filing revealed something more significant: xAI, Musk's AI venture, burned through $6.4 billion in 2025 alone. This is the first time the public has seen real numbers on what it costs to compete with OpenAI and Anthropic at the frontier. The answer: more than most companies make in revenue.
The filing makes clear why Musk structured the deal to maintain 85.1% voting control. You don't hand shareholders a say when you're planning to pour billions more into an AI model that hasn't proven it can generate profit. The prospectus was "full of surprises" according to Fortune, but the biggest one is how transparently Musk is using SpaceX's balance sheet to fund his AI ambitions. Starlink's subscriber revenue is subsidizing Grok's compute bills.
"Some investors remain doubtful whether the sum of Musk's aspirations for SpaceX are worth valuing at as much as $2 trillion."
Then there's the litigation risk. SpaceX has reserved over $500 million for potential legal losses, with a portion earmarked for complaints about Grok's controversial "spicy mode" generating sexualized images. This isn't just a PR problem. It's a material business risk significant enough to warrant half a billion dollars in contingency planning. The filing treats AI safety failures the same way it treats rocket explosions: inevitable, expensive, and factored into the cost of doing business.
What TechCrunch noted is that the spending isn't slowing down. The filing outlines plans for "a massive Grok expansion," which means xAI's $6.4 billion 2025 burn rate is a floor, not a ceiling. The AI arms race has no brake pedal. You either keep spending or you fall behind. Musk is betting that SpaceX's IPO can fund the gap between where Grok is now and where it needs to be to compete with GPT and Claude.
Key takeaways from the filing:
- xAI lost more in 2025 than most AI companies have raised in total funding
- SpaceX is cross-subsidizing AI development with satellite internet revenue
- Legal risk from AI model behavior is now a line item in billion-dollar IPO filings
Goldman Sachs and Morgan Stanley are leading the offering, targeting a valuation above $2 trillion. That would make it the largest IPO in history. But the filing reveals a company that's less "profitable rocket business" and more "expensive AI bet with a rocket-launching side hustle." The prospectus essentially asks investors: do you believe Grok can catch up to the leaders before the money runs out?
The Implication
If SpaceX gets its $2 trillion valuation, it sets a new floor for what frontier AI development costs. Other companies will look at xAI's $6.4 billion annual burn and realize they're not even in the same league. Expect more AI startups to seek non-traditional funding structures, more cross-subsidization from profitable businesses, and more founders trying to replicate Musk's playbook of using one company's cash to fund another's moonshot.
For investors, the message is clear: building competitive AI at scale requires cash reserves that only a handful of companies can access. The agent economy won't be built by scrappy startups. It'll be built by the companies rich enough to burn billions while they figure out the business model.