Starlink makes money at margins close to 40%. The AI division lost more than $1.2 billion in the quarter. And SpaceX still spent nearly $16 billion on AI infrastructure in those same three months.
Published on August 8, 2026 · by Alex · SpaceX · Artificial intelligence · Earnings
SpaceX reported quarterly results with $7.8 billion in revenue, growth of roughly 92% year over year. Starlink remains the crown jewel: more than 12 million subscribers and margins near 40%.
But the number that settles the argument is none of those. It is this one: the AI division lost more than $1.2 billion in the quarter, and SpaceX spent nearly $16 billion on AI infrastructure over those same three months — more than six times what the division earned.
| Item | Figure |
|---|---|
| Total revenue | $7.8 billion (+92% year over year) |
| Starlink subscribers | More than 12 million |
| Starlink margin | Close to 40% |
| AI division loss | More than $1.2 billion |
| AI infrastructure spending | Nearly $16 billion in the quarter |
One company, two engines pulling in opposite directions. Starlink is a mature business that generates cash. The AI division is a bet that consumes it. Looking only at 92% growth does not tell that story.
Elon Musk's bet is to build the entire infrastructure for the next generation of artificial intelligence before competitors do. Not just data centers: satellites, communications, compute, and possibly the chips themselves.
The logic is the usual one in infrastructure: whoever builds it first controls everyone else's access. If it works, SpaceX could end up operating several industries at once instead of competing inside one.
At the current pace of investment, that division could consume tens of billions of dollars before proving whether it will be profitable. That is what separates the two readings of the same statements.
Nobody has the number that settles this yet, because the number is time: how many quarters the company can sustain this pace before the AI division shows revenue that justifies what it cost.
This week, shares held by employees and early investors began to unlock.
Here is how it works: when a company goes public — SpaceX did on June 12 — there is a lock-up period, usually 180 days, during which early shareholders cannot sell. When it expires, a supply of shares appears that did not exist before.
A lock-up expiry does not mean the price will fall. It means there may be more available sellers than in prior months, and that usually translates into more volatility for a while. It is a change in share supply, not a verdict on the business.
The question, then, is not whether SpaceX goes up or down tomorrow. It is whether we are watching the birth of the largest company ever built, or the riskiest investment of the entire AI era. And there is a detail about a possible SpaceX–Tesla merger that could change the whole story — but that deserves its own piece.
None of this is a buy or sell recommendation. See our legal notice.
$7.8 billion, roughly 92% more than a year earlier. Starlink passed 12 million subscribers with margins near 40%. But the AI division lost more than $1.2 billion over those same three months, and that is the number that settles the argument.
Nearly $16 billion in a single quarter, more than six times what the division earned. Elon Musk is betting on building the infrastructure for the next generation of AI before competitors do: not just data centers, but satellites, communications, compute and possibly the chips themselves.
Time, not revenue. At the current pace, that division could consume tens of billions of dollars before proving whether it will be profitable. If it works, today’s spending will look cheap in hindsight; if it fails, the market will have paid very high multiples for a business that still demands enormous investment.
This article is the written version of the Saturday analysis.
Watch the video on TikTok · Follow the channel
