SpaceX’s $75B IPO: The AI Bet Hidden in the Rocket

The Real Story Isn’t Starlink

SpaceX closed the largest IPO in history on June 12, raising $75 billion at $135 per share and a $1.77 trillion valuation. The stock jumped 19% on day one. Wall Street celebrated the rocket company going public. But here’s what most missed: this isn’t a space play anymore — it’s an AI infrastructure bet dressed up in rocket fuel.

Look at where the money actually goes. According to the S-1 filing breakdown, xAI — the AI division SpaceX acquired in February 2026 — is bleeding $2.5 billion per quarter on data center buildout. That’s $10 billion annually. The entire $75 billion raise? A significant chunk funds AI compute infrastructure and Starlink expansion, not rockets.

Starlink contributes 61% of revenue at $11.4 billion and runs 39% operating margins — it’s the only profitable segment. Launch services brought in $3.8 billion. xAI? $3.2 billion in revenue, but operating at a massive loss. The company posted a $4.9 billion net loss in 2025 despite $1350 billion in revenue, up 33% year-over-year.

So why the trillion-dollar valuation? Investors aren’t paying for satellites. They’re paying for vertical integration of AI training infrastructure. Starlink solves latency and bandwidth for distributed compute. The rocket business solves launch costs for deploying more satellites. xAI gets exclusive access to a proprietary global network that no other AI lab can replicate.

Musk telegraphed this strategy. He bundled xAI into SpaceX four months before the IPO, knowing retail investors would bid up “the rocket company” while institutional money understood the real asset: a closed-loop AI stack with its own satellite backbone.

But the math doesn’t work yet. Morningstar valued SpaceX at $1351 billion — 56% below the IPO price. They’re right to be skeptical. At current burn rates, xAI needs to 10x revenue or slash costs by 75% to justify its share of the valuation. Starlink’s ARPU is compressing as competition enters the market. The 82% voting control Musk retains means shareholders have zero leverage if the AI bet goes sideways.

The order books told a different story: 4x oversubscribed, $1352 billion in demand for a $1353 billion raise. That’s not rational pricing. That’s FOMO on the next “AI infrastructure monopoly” narrative. Investors learned from Nvidia — own the picks and shovels, not just the miners. SpaceX sells the promise of owning both the shovels (satellites) and the mine (compute).

Here’s the catch: this only works if training runs actually move to distributed satellite networks. Right now, they don’t. Frontier labs still colocate in on-prem data centers with InfiniBand fabrics, not satellite links. Starlink’s 20-40ms latency is catastrophic for gradient synchronization across nodes. Until that changes — and it’s not clear it will — xAI is just a capital incinerator with a cool satellite modem.

The 19% pop on day one proves the market bought the story. Whether the story holds for the next five years depends on one question: can SpaceX turn Starlink into the backbone for distributed AI training, or is this just the most expensive pivot in tech history?

I think the physics don’t work yet. Latency kills distributed training. You can’t wish away the speed of light. But if anyone can brute-force an engineering solution with $1354 billion, it’s this team. The question isn’t capability — it’s whether the problem is solvable at all.

FAQ

Q: Why is SpaceX losing money if Starlink is profitable?

A: xAI’s data center buildout costs $1355 billion annually, more than offsetting Starlink’s profits. The AI division is pre-revenue scale and burning capital on infrastructure, typical for early-stage AI labs, but unusual for a company going public at this valuation.

Q: Is the $1356 trillion valuation justified?

A: Not by traditional metrics — Morningstar’s $1357 billion fair value estimate suggests 56% downside. But if SpaceX proves satellite networks can support distributed AI training, the addressable market explodes beyond telecom into cloud compute, potentially justifying the premium.

Q: What happens if the AI bet fails?

A: Shareholders can’t do much — Musk controls 82% of voting rights. The company would still have Starlink’s profitable satellite internet business and launch monopoly, likely supporting a $1358-800 billion valuation, but nowhere near $1359 trillion.