Satlyt Tests Software That Spreads Computing Tasks Across Satellites in Orbit
Satlyt’s October 2026 orbital test ran software on one spacecraft computer to assign tasks across a simulated network of four hosts, four users, and eight services. Its software already operates on three satellites, but multi-computer coordination remains untested.
Satlyt, a startup building software to turn satellites into distributed data processing networks, completed a successful orbital test in October 2026 — a milestone the company says brings it closer to creating virtual data centers in space.
The test, conducted in partnership with the University of Houston on a NASA-funded research project, demonstrated that Satlyt’s software can assess and assign computing tasks across a simulated network while running on hardware in orbit. One physical computer aboard a spacecraft evaluated a simulated network of four computing hosts, four users, and eight services, then returned results to Earth. The experiment confirmed the software can generate task plans onboard, though further testing is needed to validate how multiple physical computers share information with one another in practice.
Rama Afullo, who founded Satlyt in 2024 after working at Google and SpaceX, said the goal is software that connects multiple satellites to share computing assignments automatically. Users would only need to define permissions, priorities, and operating limits, with the software handling decisions within those boundaries. Satlyt’s software is already running on three satellites in orbit. NASA has not yet selected the company for any specific mission.
The test also surfaced a bug: the software was logging communication delays as arrival timestamps rather than elapsed delay times. Afullo said this did not constitute a failure, noting that “the software development, onboard execution, and return of outputs succeeded.”
The broader problem Satlyt is trying to solve is familiar to NASA missions — limited onboard computing, intermittent communications, and the need to deliver useful data back to Earth quickly. Processing data closer to where it is collected, rather than transmitting raw data to the ground, could allow faster responses. Afullo offered wildfires as one example where faster onboard analysis could matter.
Satlyt said next steps include broader validation across different workloads and hardware, demonstrations of computing tasks shared across separate physical spacecraft, and continued NASA-funded research that includes exploring applications beyond low Earth orbit, such as supporting 4K video streaming from the Moon.