Starcloud compute at source
Starcloud
This wedge matters because it turns orbital compute from a cheaper cloud someday story into a must-have infrastructure layer for satellites that cannot practically send every raw file back to Earth. For an Earth observation or SAR operator, the bottleneck is often moving huge image files through limited downlink windows. Processing data in orbit lets the customer send down alerts, detections, or compressed outputs instead of terabytes of raw sensor data, which solves a workflow problem immediately, even before orbital compute beats terrestrial cloud on price.
-
Starcloud is already positioned around this model. Its near term product is hosted compute and storage for other spacecraft operators, and the company frames its first commercial revenue around in orbit processing before the larger hyperscaler opportunity arrives. That makes space data workloads the practical beachhead, not general purpose cloud replacement.
-
The closest proof point is Kepler, which commissioned distributed on orbit computing across 10 satellites in March 2026 using 40 NVIDIA Jetson Orin modules. Kepler is selling cloud like processing tied to its relay network, which shows customers will pay for faster decisions in space even with far smaller systems than Starcloud ultimately wants to build.
-
NVIDIA is validating the same architecture at the ecosystem level. Its March 16, 2026 space computing launch was aimed at orbital data centers, geospatial intelligence, and autonomous space operations, which is another way of saying the first valuable jobs are sensor processing and spacecraft decision making at the source of the data, not cheap bulk training in orbit.
The market is heading toward a split model. Space native data will be filtered and acted on in orbit, while only the highest value outputs move to Earth for storage, analysis, and customer delivery. If Starcloud becomes the compute layer that sits next to sensors and relay networks, it can win revenue years before launch costs are low enough for orbital compute to compete head on with terrestrial GPU clouds.