Satlyt Raises $8 Million to Build an Open AI Cloud Across Satellites
Satlyt raises $8 million to run AI across satellites, launching with NASA workloads as it targets an open orbital cloud spanning multiple spacecraft in 2027.
Summary
On October 13, 2026, Kenyan American engineer Rama Afullo, a former Google cloud employee and SpaceX Starlink product manager in 2024, disclosed an $8 million seed round for Satlyt after both companies rejected his orbital computing pitch. The Sunnyvale and Nairobi startup’s round was led by Houston’s Non Sibi Ventures, where partner Bernard Harris spent more than 20 years as a NASA astronaut. Partner Kent Lucas says Satlyt can grow with satellite launches without relying on orbital data centers, whose economics and launch capacity remain uncertain.
Satlyt is building an open, hardware independent software layer, likened to Android, VMware and Snowflake, rather than spacecraft like SpaceX, Google, Starcloud and Cowboy Space Company. Across two demonstrations, Google DeepMind’s Gemma model aboard a Momentus spacecraft cut an onboard software error transmission by more than 60% earlier in 2026, potentially saving hundreds of thousands of dollars per satellite annually by reducing slow, costly downlinks and ground control work. On Thursday, October 15, its software launches on a SpaceX rocket beside Google’s first Project Suncatcher prototype, aboard TakeMe2Space hardware serving NASA cloud protocol tests, Stellerian orbital surveillance image processing and TakeMe2Space’s software hosting demonstration.
Few high powered GPUs are currently in orbit, making anomaly resolution and onboard sensor processing the near term opportunity. Satlyt plans to link two satellites into a shared cloud in 2027, seeking to become a third party computing provider that turns spacecraft into managed, revenue generating services. Afullo targets deployment on 20% of satellites by the decade’s end, when he expects virtually every spacecraft maker to include GPUs or similar advanced processors.
Positives
- The $8 million seed round gives Satlyt capital to develop software that distributes AI workloads across satellites from different operators.
- Google DeepMind’s Gemma reduced an onboard software error transmission by more than 60% during Satlyt’s Momentus demonstration.
- Reduced downlinks and ground intervention could save hundreds of thousands of dollars per satellite annually.
- NASA, Stellerian and TakeMe2Space will test cloud protocols, image processing and third party software hosting on the October 15 mission.
- Satlyt plans to demonstrate a shared computing cloud spanning two satellites in 2027.
Risks & concerns
- Few high powered GPUs are currently in orbit, limiting the workloads Satlyt can address today.
- Orbital data centers face uncertain economics and a launch bottleneck, according to investor Non Sibi Ventures.
- Satlyt has completed two demonstrations but has not yet proved its planned cloud across separate satellites.
- Downlinking spacecraft data remains expensive and slow, leaving operators dependent on constrained communications and ground controllers.
- Afullo’s goal of reaching 20% of satellites depends on spacecraft makers broadly adopting GPUs and similar processors.