SpaceX as Orbital Compute Utility
CoreWeave of space
This is an attempt to move SpaceX from being valued like a transportation company to being valued like the owner of a new compute utility. Launch and Starlink are already large businesses, but both are showing signs of maturing, so the bigger re rating comes from adding a third layer where SpaceX not only sends hardware to orbit and connects it, but also rents out the GPU capacity running there. That is the same jump CoreWeave made on Earth, from hardware access to infrastructure revenue.
-
The reason this matters now is growth mix. SpaceX reached about $15.5B of revenue in 2025, but Starlink growth slowed and launch growth also moderated. Starship V3 and its targeted sub $100/kg cost are what reopen the path to a much larger market, because orbital compute only works economically if launch gets radically cheaper.
-
SpaceX has a more complete stack than any pure play rival. It already owns the rocket, the satellite manufacturing base, and a 9,000 plus satellite Starlink network that can move data between orbiting assets. After the xAI deal, it also has a built in first customer for training and inference workloads, which helps solve the usual cold start problem for new infrastructure.
-
The comparable is less Blue Origin and more CoreWeave. CoreWeave is valuable because customers pay continuously for compute, not just once for a machine or a building. SpaceX is trying to create the same recurring economics in space, while also collecting launch fees and connectivity revenue around that compute job, which makes each workload worth more than a normal cloud contract.
If Starship reaches the cost and flight cadence targets, orbital compute becomes the bridge from two strong businesses into a much larger infrastructure platform. That would make SpaceX look less like a launch leader with a satellite internet cash cow, and more like the default operating layer for AI workloads that need power, bandwidth, and deployment at planetary scale.