Satellites do not have to send every piece of information back to Earth before it becomes useful. Satlyt is testing software that lets spacecraft decide where computing tasks should run in orbit, bringing analysis closer to the data itself.
That matters because satellites operate with limited computing power, intermittent communications and narrow windows for sending information home. Satlyt founder Rama Afullo said the company had completed a successful test earlier this month with the University of Houston as part of a NASA-funded research project.
The test was a planning step, not an orbital data center
Satlyt’s software is already running on three satellites, but the latest experiment used one physical computer aboard a spacecraft to evaluate a simulated network of four computing hosts, four users and eight services.
The system assessed how to assign tasks based on available computing resources and communications. It then generated a plan on hardware in orbit and returned the results to Earth. One approach performed best, although it required more computing power than the alternatives.
That distinction is important. The test showed that the software can make computing plans in orbit. It did not yet demonstrate multiple physical spacecraft sharing workloads in practice, which is the larger ambition.
The useful shift is from downlinking data to acting on it
Satlyt is building toward a model in which operators define permissions, priorities and operating limits, while software decides how work should be distributed. In a wildfire scenario, that could mean a satellite identifying relevant information and sending a faster, more useful result instead of transmitting everything for ground-based analysis.
The company still has to test different workloads, hardware and connectivity conditions. It also identified a timing issue in the experiment, where delays were recorded as arrival time rather than elapsed delay. Satlyt says the software execution and return of outputs succeeded, but secure transfers, operator permissions, resource scheduling and recovery from unavailable connections remain unresolved.
The broader consequence is straightforward: if orbital computing becomes reliable, satellites begin to look less like passive sensors and more like distributed computing environments. The strategic value will not come from calling them data centers. It will come from deciding what information deserves to move, when and under whose rules.