Satlyt has raised an $8 million seed round to build software that runs AI workloads on satellites from different manufacturers. The Sunnyvale- and Nairobi-based startup wants to provide a shared computing layer in orbit rather than design its own spacecraft or depend on one hardware platform.
The company has already flown software on two demonstration missions. Earlier this year, it ran Google DeepMind’s Gemma model aboard a Momentus spacecraft and says the model compressed a report about onboard software errors by more than 60 percent. Processing data before it is sent to Earth can reduce limited and costly downlink use.
A new mission aboard a satellite built by India’s TakeMe2Space will carry workloads for three customers. NASA will test protocols for cloud computing in space, Stellerian will evaluate image processing for tracking orbital objects, and TakeMe2Space will demonstrate that outside software can run on its hardware.
Satlyt plans to connect computing across two separate satellites next year. That would be an early test of its larger claim that spacecraft can share jobs as an orbital cloud. The near-term market is narrower: few satellites carry powerful GPUs, and launch capacity and orbital data-center economics remain uncertain. The funding round was led by Non Sibi Ventures.