SpaceX’s third-quarter numbers moved the company from an earnings-preview question into a much more concrete market test: can the returns on its AI infrastructure keep pace with the speed of the spending?
The company reported $14.6 billion of revenue, up 161% year over year, $2.8 billion of operating income, $4.4 billion of net income and $8.8 billion of adjusted EBITDA for the quarter ended September 30.
The AI business is now the center of the story
SpaceX said its Super Intelligence segment added $6.7 billion of incremental solutions and infrastructure revenue from a year earlier, driven by cloud-services agreements, the Cursor acquisition and its AI models and software applications. Nameplate compute capacity reached 1.7 gigawatts by quarter-end.
Reuters reported Super Intelligence revenue of about $7.6 billion for the quarter. That is already larger than the connectivity business, which means SpaceX can no longer be analyzed primarily as a launch company with Starlink attached.
Capex is the key counterweight
Total capital expenditure reached $19.6 billion in Q3, nearly triple the year-earlier level. Reuters reported that roughly $16.4 billion went to the AI division.
That changes the question the market should ask. The issue is not whether SpaceX can find projects to spend on; it is whether the company can preserve attractive returns as the AI buildout becomes the dominant use of capital.
Starlink still matters because it diversifies the funding base
Connectivity revenue grew 69% year over year and operating income grew 99%, according to the company. SpaceX also said its first revenue-generating Starship mission delivered 26 Starlink V3 satellites, with roughly ten times the downlink capacity added by a Falcon 9 Starlink launch.
That combination — recurring connectivity earnings plus a rapidly scaling AI infrastructure business — is what differentiates SpaceX from many neocloud operators. The balance sheet can absorb more experimentation, but the market will still demand evidence that new compute capacity earns an adequate return.
What changed from the pre-earnings thesis?
Before the release, the risk was that investors might punish another large spending commitment in a market already worried about AI financing. The result instead showed enough revenue and earnings acceleration to make capital productivity the more useful framework.
A strong top line does not make $19.6 billion of quarterly capex irrelevant. It raises the hurdle for the next quarter: investors will want to see utilization, margins, backlog conversion and cash generation develop quickly enough to justify a buildout measured in tens of billions of dollars.