Sunnyvale, Calif. — Satlyt Raises $8M to Run AI on Satellites

Rama Afullo pitched the idea of orbital data centers inside Google and inside SpaceX. Both companies passed. "When I was at SpaceX, I tried to pitch this internally. They said no. When I was at Google, I tried to pitch this internally. They said no," Afullo told TechCrunch. So he left SpaceX's Starlink unit and co-founded Satlyt, a Sunnyvale and Nairobi-based company that has now raised an $8 million seed round to build software that runs artificial intelligence models on satellites, Afullo told TechCrunch.

What does Satlyt's software actually do?

Satlyt does not build spacecraft. It builds the software layer that lets AI models run on satellites that other companies manufacture and operate, according to TechCrunch's reporting. Afullo compares the product to VMware and Snowflake — platforms that let customers run complex workloads without managing the underlying hardware. The near-term use case is operational: satellites typically rely on ground-based flight controllers to diagnose problems, and sending diagnostic data to Earth, known as downlink, is slow and expensive. Running AI onboard lets a satellite process sensor data or flag anomalies without waiting on a full download, per TechCrunch.

How does Satlyt differ from SpaceX, Google, and other orbital computing bets?

SpaceX has moved into building its own orbital data centers, and Google's Project Suncatcher is sending a first prototype spacecraft into orbit this week, according to TechCrunch. Startups Starcloud and Cowboy Space Company are also building their own satellites for onboard computing, the outlet reported. Satlyt's pitch is horizontal rather than vertical integration. "The folks like SpaceX who are doing orbital data centers — if they are the iPhone, we'll build Android as a horizontally integrated, open ecosystem," Afullo said, according to TechCrunch. The distinction matters for customers: an iPhone-style approach bundles software to one company's hardware, while an Android-style approach is meant to run across satellites built by multiple manufacturers.

Who paid for the $8 million seed round, and why now?

TechCrunch reported the $8 million figure as a seed round but did not name the investors in the material reviewed for this story. The timing lines up with a broader industry push: SpaceX's pivot toward orbital data centers and Google's Suncatcher prototype are both advancing in the same window as Satlyt's raise and its software's commercial debut, per TechCrunch.

What did the Momentus demonstration show?

Satlyt has already flown its software on two demonstration missions, TechCrunch reported. Earlier this year, the company deployed Google DeepMind's Gemma AI model onboard a spacecraft operated by Momentus. According to figures Afullo gave TechCrunch:

• The Gemma model cut the size of a transmission describing onboard software errors by more than 60%.
• Afullo told TechCrunch that kind of efficiency can save "hundreds of thousands of dollars per satellite each year" in downlink costs.

"The folks like SpaceX who are doing orbital data centers — if they are the iPhone, we'll build Android as a horizontally integrated, open ecosystem." — Rama Afullo, Satlyt co-founder, to TechCrunch

What happens at Thursday's SpaceX launch?

Satlyt's software is scheduled to launch Thursday on a SpaceX rocket, flying alongside the first prototype for Google's Project Suncatcher, according to TechCrunch. This mission runs on hardware built by TakeMe2Space, an Indian startup that makes computing hardware for satellites, rather than on a Satlyt-built spacecraft, consistent with the company's software-only model. TechCrunch reported the mission involves three customers: NASA, which is paying Satlyt to test protocols for cloud computing in space; a company identified as Stellerian; and TakeMe2Space itself. The source material reviewed for this story did not specify Stellerian's role beyond its listing as a mission customer.

For Satlyt, the launch is a test of whether an operating-system approach to orbital AI can attract customers who would otherwise have to build or buy into a single vendor's satellite stack. For NASA and other early customers, the mission is a test of whether AI processing onboard a spacecraft can reliably cut the cost and delay of sending raw data back to Earth — the problem Afullo says both of his former employers declined to chase when he raised it internally.