SpaceX AI Unit Outearns Starlink in Q3 With $7.6B Revenue
SpaceX is now a compute company that happens to own a launch provider, not the reverse. The 1.7 GW of compute at quarter-end, scaling to 2.5–3 GW by December, targets the same cloud-infrastructure market OpenAI, CoreWeave, and Lambda Labs are fighting over—and SpaceX's satellite-to-ground latency advantage is real but not defensible if terrestrial fiber becomes cheaper faster than Starlink can scale. The second move: Starlink's enterprise revenue nearly doubled YoY to $2.1B, but that growth now feeds SpaceXSI's capex budget, not Starlink's own expansion.
Musk is redirecting SpaceX's capital architecture toward compute infrastructure, not launch cadence.
The company projects capex of $35B–$40B in Q4 alone—roughly double Q3 spending—with $16.4B of Q3 capex already flowing to SpaceXSI, and Musk claiming confidence in adding 10 GW of compute capacity in 2027. This reallocation means launch operations (including Starship) become margin-accretive only if compute monetization sustains the $500B–$1T annualized-revenue forecast Musk offered for end-2027. Watch whether Q4 2026 capex actually hits the $35B–$40B guidance or whether capital constraints force a pivot; if guidance slips, the compute-over-launch thesis fractures and Wall Street reprices the entire venture.
Does the $7.6B SpaceXSI revenue include only the Cursor acquisition ($608M stated separately) or is Cursor double-counted in the cloud-deals number? The earnings call should clarify the segment's actual organic growth versus M&A contribution.
Fault Lines
Strategic intelligence, synthesized daily — with a public track record. Every call graded against what actually happened.