How one company built three compounding businesses on a single obsession: making mass-to-orbit cheap.
Every business line SpaceX operates exists because the price of putting a kilogram into orbit fell roughly twentyfold. The cost curve is the business model: each arm either monetizes it or steepens it.
Reusability and vertical integration push the cost of a kilogram to orbit down.
Lower prices unlock demand no one priced before. Cadence compounds.
Every flight returns hardware and telemetry. Fixed costs amortize across the manifest.
Learning curves convert cadence back into cost reduction.
At a low enough $/kg, entirely new businesses pencil: Starlink first, compute next.
Falcon 9 and Falcon Heavy flew 170 missions in 2025 and lifted 2,213 metric tons to orbit, per the company's S-1. Most flights use boosters that have flown before. The fleet leader, B1067, has flown 35 times, a record for any orbital-class booster; SpaceX depreciates each booster over 25 flights, so the record holder is flying past its own accounting life.
NASA, the Department of Defense, and commercial satellite operators pay the bills. The strategic customer is SpaceX itself, and Starship, still in flight test, is where the cost curve goes next.
When launch got cheap, the market for launches did not grow fast enough to absorb it. So SpaceX filled the manifest itself, building a constellation of more than 7,000 satellites and selling connectivity to homes, ships, airlines, and militaries: consumer broadband, maritime, aviation, direct-to-cell, and Starshield for defense.
Starlink passed 10.3 million subscribers in March 2026. The Connectivity segment earned $11.4 billion in 2025, 61% of company revenue, with $4.4 billion of operating profit. Per the S-1, it is the only segment that makes money.
In February 2026, SpaceX acquired xAI in an all-stock deal, folding Grok, X, and the Colossus data centers into a new AI segment. The logic runs through the cost curve: on Earth, energy and cooling are the binding constraints on AI compute; in orbit, solar power is continuous; and Starship's target of under $200 per kilogram is what makes orbital data centers worth modeling.
The segment earned $3.2 billion in 2025, 17% of revenue, and its data-center capacity reached 1.0 gigawatt in early 2026. It also absorbs most of the company's capital: 61% of 2025 capex, rising to 76% in the first quarter of 2026.
The moat. A structurally lower cost per kilogram that no competitor has matched. 22% of 2025 revenue, and the foundation of the other two arms.
The monetizer. 61% of revenue and the only profitable segment, at $4.4B of operating profit in 2025.
17% of revenue and most of the capex. The wager that cheap orbital mass eventually makes cheap orbital compute.
SpaceX listed on Nasdaq on June 12, 2026, pricing at $135 a share and raising roughly $75 billion, the largest IPO ever. The stock closed its first day at $161, up 19%, and joins the Nasdaq-100 on July 7. Musk holds 42% of the shares and 82% of the voting power. The honest ledger: a $4.9 billion net loss in 2025 against $6.6 billion of adjusted EBITDA, with the losses concentrated in the AI buildout.
Cheap mass to orbit isn't the product. It's the platform.