Satlyt Raises $8 Million to Build an Open AI Cloud Across Satellites

Summary

Former Google cloud engineer and SpaceX Starlink product manager Rama Afullo co-founded Satlyt after both companies rejected his orbital computing pitch. Headquartered in Sunnyvale, California, and Nairobi, Satlyt raised an $8 million seed round led by Houston-based Non Sibi Ventures. Rather than build spacecraft like SpaceX, Google, Starcloud and Cowboy Space Company, Satlyt is developing an open software layer, comparable to VMware or Snowflake, that runs AI across different operators’ satellites. On Thursday, October 1, 2026, its software is scheduled to launch on a SpaceX rocket aboard TakeMe2Space hardware, alongside Google’s first Project Suncatcher space data center prototype.

Satlyt has completed two demonstrations, including deploying Google DeepMind’s Gemma model on a Momentus spacecraft. Gemma reduced an onboard software-error transmission by more than 60%, which Afullo says could save hundreds of thousands of dollars per satellite annually by limiting slow, expensive downlinks. The TakeMe2Space mission will test NASA cloud protocols, Stellerian surveillance-image processing and third-party software hosting. Satlyt plans a two-satellite cloud demonstration in 2027, despite few powerful GPUs currently operating in orbit. Non Sibi partners Bernard Harris, a NASA astronaut for more than 20 years, and Kent Lucas backed a model that can grow with satellite launches without requiring full orbital data centers, whose economics and launch capacity remain uncertain. Afullo aims to reach 20% of satellites by 2030 and turn spacecraft into revenue-generating managed services.