SpaceX's IPO Filing Decoded: Starlink Prints Money, AI Burns It

SpaceX just filed its S-1. Not a rumor. Not a leaked term sheet. The actual SEC filing. And the numbers tell a more interesting story than the hype cycle would suggest.

I pulled the full filing from EDGAR and read it so you don’t have to. Here’s what matters.

The Structure: Controlled, Dual-Class, Musk-Still-King

SpaceX is going public with a dual-class share structure. Class A shares get 1 vote. Class B shares get 10. Musk holds the Class B. The company explicitly says it will be a “controlled company” under Nasdaq rules.

This is not a surprise. But it’s worth noting: this IPO is not about giving up control. It’s about liquidity, capital access, and the next phase of the build.

The ticker will be SPCX, listed on both Nasdaq and Nasdaq Texas. The pricing and share count are still blank placeholders in the filing, so the valuation speculation is just that - speculation.

The Numbers: $18.7B Revenue, $2.6B Operating Loss

2025 financials from the filing:

• Revenue: $18.674 billion
• Operating loss: $(2.589) billion
• Adjusted EBITDA: $6.584 billion
• Cash (Mar 2026): $15.852 billion
• Total assets: $102.094 billion

Q1 2026:

• Revenue: $4.694 billion
• Operating loss: $(1.943) billion

So SpaceX is massive, growing fast, and still losing money at the operating line. But adjusted EBITDA of $6.5B on $18.7B revenue tells you the underlying business has serious margin potential. The question is where that margin lives and where the cash goes.

The Three Pillars

The filing breaks SpaceX into three segments, and this is where the story gets real.

Starlink (Connectivity): The Cash Machine

• 2025 revenue: $11.387 billion (61% of total revenue)
• 2025 operating income: $4.423 billion
• 2025 adjusted EBITDA: $7.168 billion
• Subscribers: 10.3 million (as of March 31, 2026)
• ARPU: $66/month
• Satellites in orbit: ~9,600 (broadband + mobile)

Starlink is the profit engine. Full stop. It generates more operating income than the rest of SpaceX burns. It has positive unit economics, a growing subscriber base, and a moat built on orbital infrastructure that nobody else can replicate at this scale.

V3 Starlink satellites are expected H2 2026. If those deliver on throughput promises, the margin story gets even better.

Space: The Core Mission, Still Finding Its Margin

• 2025 revenue: $4.086 billion
• 2025 operating loss: $(657) million
• 2025 adjusted EBITDA: $653 million

Launch services, government contracts, satellite deployment for third parties. Revenue is real but margins are thin. The segment is EBITDA-positive but operationally negative, which means heavy capital investment is still running ahead of revenue.

Starship payload delivery is expected H2 2026. That’s the unlock. If Starship works at scale, the unit economics of heavy lift change dramatically. If it doesn’t, Space stays a break-even business funding the national security and exploration missions.

AI (xAI + X): The Burn Furnace

• 2025 revenue: $3.201 billion
• 2025 operating loss: $(6.355) billion
• Q1 2026 capex: $7.723 billion

This is the wild card. xAI was acquired into SpaceX in early 2026 and it’s now part of the AI segment in the filing. X (formerly Twitter) and Grok sit under this umbrella too.

Revenue of $3.2B against an operating loss of $6.4B means the AI segment is burning over $2 for every $1 it brings in. And Q1 2026 capex alone is $7.7B. This is not a side project. This is a massive, sustained capital bet on AI compute and inference infrastructure.

Orbital AI compute satellites are targeted as early as 2028. That’s the long-term vision: move compute off Earth, leverage the launch infrastructure SpaceX already owns, and create an orbital AI platform.

Whether that’s genius or hubris depends entirely on execution.

The Real Story

Strip away the filing language and here’s what you get:

Starlink pays for the rocket company. The rocket company enables the AI ambition. The AI ambition is where the capex goes.

This is not a “space company going public.” This is a vertically integrated infrastructure play where satellite internet funds launch, launch enables orbital compute, and AI is the demand sink for all of it.

The risk is obvious: if AI revenue doesn’t scale to match the burn rate, Starlink margins will be subsidizing losses for years. The upside is that nobody else has this stack. Amazon has Kuiper (years behind). Nobody has reusable heavy lift at Starship scale. Nobody else can plausibly put AI compute in orbit.

What’s Missing

The filing leaves several key blanks:

• Exact pricing and share count (still placeholder)
• Use of proceeds (generic “general corporate purposes” language)
• Detailed risk factors around xAI integration
• Any concrete timeline for AI segment profitability

These will get filled in amended filings before the roadshow. But the structure of the story is already clear.

Why This Matters for AI Operators

If you build with AI or run agent infrastructure, this filing is relevant beyond the finance-porn aspects:

1. Compute is the bottleneck, and SpaceX is betting orbital is the answer. If you’re planning 3-year infrastructure roadmaps, factor in the possibility that the cheapest high-throughput compute might not live in a data center in Virginia.
2. The AI segment’s $6.4B operating loss shows what it costs to compete at the frontier. Grok, Colossus, and the xAI stack are burning cash at a rate that only a company with Starlink margins can sustain.
3. Vertical integration is the moat. Launch + connectivity + compute is a stack nobody else can assemble. The question is whether assembling it is enough, or whether you need to be best-in-class at each layer.

The Bottom Line

SpaceX is going public as a $100B+ asset with real revenue, real losses, and a three-pillar thesis that’s either the most ambitious infrastructure play in history or a very expensive way to subsidize an AI chatbot.

Starlink makes it investable. Starship makes it scalable. xAI makes it a question mark.

The market will price the tension between those three. And the rest of us get to watch.

---

Source: SpaceX S-1 Filing, SEC EDGAR