Rama Afullo’s immediate reaction was simple: “Let’s GOOOO.” His company, Satlyt, is developing software that can transform existing satellites into interconnected data-processing networks, bringing the concept of orbital data centers closer to reality. Now, the startup has reached another important milestone toward that goal.
After 13 months of development, Satlyt successfully completed a test earlier this month in collaboration with the University of Houston as part of a NASA-funded research project. The experiment demonstrated how the company’s software can distribute computing tasks among available resources in orbit, taking communication conditions and processing capacity into account.
Satlyt aims to build software that connects multiple satellites and allows them to share computing workloads. Under the proposed system, users would only need to establish permissions, priorities, and operational limits. The software would then make decisions and allocate tasks within those boundaries, Afullo explained.
The company’s technology is already installed on three satellites in orbit, and the latest experiment represents an early step toward its broader vision. Although the research received NASA funding, the agency has not selected Satlyt for a specific mission or chosen its technology for operational deployment.
Afullo said NASA missions face many of the same challenges Satlyt is attempting to solve, including limited onboard computing capacity, unreliable or intermittent communications, and the need to deliver useful information to people on Earth.
For the experiment, Satlyt built a simulated network consisting of four computers, four users, and eight services. Its software evaluated how different tasks should be assigned and executed across the simulated computers, considering the availability of processing resources and communication links.
A physical computer aboard a spacecraft carried out the experiment. It ran the software to evaluate the simulated network of four computing hosts before transmitting the results back to Earth.
The test demonstrated that Satlyt’s software could develop computing plans using hardware operating in orbit. However, additional experiments will be necessary to determine how multiple physical computers can exchange information and coordinate workloads with one another under real operating conditions.
Afullo described the results as revealing a trade-off between different approaches. One method delivered the strongest performance, but it also required more computing resources than the alternatives.
NASA already operates a technology network designed to move information through space despite communication delays. The agency began using the technology earlier this year and has also tested it aboard aircraft.
Afullo, who previously worked at Google and SpaceX, founded Satlyt in 2024 with the ambition of turning spacecraft into virtual data centers. The company’s software is designed to let satellite operators run AI models and other applications directly onboard their spacecraft.
Processing information closer to where it is collected could reduce the need to transmit large volumes of raw data back to Earth through downlinks before performing analysis on ground-based systems.
For example, when a satellite detects a wildfire, onboard processing could help identify important information and transmit relevant findings to people on Earth more quickly.
“The goal is to get useful answers to people sooner and make better use of limited communication windows,” Afullo said. He added that the latest experiment tested one planning component of this broader capability.
Satlyt still needs to evaluate its software across different workloads, hardware configurations, and connectivity conditions. The team also identified an issue involving how the software measured communication delays. According to Afullo, it recorded the arrival time of information rather than the elapsed time required for that information to arrive, which are two different measurements.
However, he emphasized that this issue did not mean the experiment had failed. The software development, execution aboard the spacecraft, and successful return of the results to Earth were all completed.
Despite the progress, Afullo acknowledged that significant engineering and operational challenges remain. These include establishing reliable communication links, securing permission from participating satellite operators, enabling secure data transfers, scheduling computing resources, and recovering when connections between computers become unavailable.
The company is also exploring applications beyond low Earth orbit. Through its NASA-funded research, Satlyt is investigating the possibility of supporting 4K video streaming from the Moon.
The next phase will involve broader testing and integration, followed by demonstrations in which computing workloads are shared across separate spacecraft.
“Our next technical steps are broader validation and integration, followed by demonstrations of computing work shared across separate spacecraft,” Afullo said.
In the longer term, Satlyt wants to make it possible for developers to build applications for space without having to recreate the underlying computing infrastructure for every mission.
“That’s what virtual AI data centers in space mean to us.”



