Satlyt, a startup founded in 2024 by ex-Google and SpaceX engineer Rama Afullo, is closing the gap on transforming satellites into distributed data centers orbiting Earth. The company’s software, already onboard three satellites, is designed to let spacecraft share computing tasks directly in space—reducing reliance on sending raw data back to Earth for analysis.
Testing the future of orbital compute sharing
Earlier this month, Satlyt conducted a key test in collaboration with the University of Houston under a NASA-backed project. In it, a simulated network of four computers, four users, and eight services demonstrated how tasks could be allocated based on communication delays and available computing resources. Only one of the four computers was a physical onboard unit—the rest were virtual—yet the software successfully planned task distribution and executed parts of the workload from orbit, returning the outputs to Earth. This marks a milestone for onboard decision-making in space beyond simple data collection and transmission.
The experiment also surfaced a technical snag: the software mistakenly treated timestamp data as the moment data arrived, instead of calculating actual transit delay. Though not a failure, this issue points to the finer details still to be polished before fully autonomous task-sharing among spacecraft becomes feasible.
What’s next for space data centers
Satlyt’s vision is bold: to enable developers to build applications that run in space by deploying a virtual infrastructure across multiple satellites. Doing so would mean fewer delays, faster disaster response, and more autonomous missions. For instance, a satellite could recognize signs of a wildfire and transmit immediately actionable insights rather than raw images—cutting hours or days from traditional workflows.
The road ahead has several challenges. Beyond technical work under different hardware, connectivity, and workload conditions, Satlyt must address operational issues like securing communication links, gaining permissions from satellite operators, scheduling resources across physically separate units, and gracefully handling dropped or lost connections. Meanwhile, NASA has not yet committed to using this technology for any missions, despite supporting the test.
Satlyt’s recent success builds on broader efforts in space tech to shift intelligence from ground stations to where it matters most—onboard orbiting hardware. But the idea of spaceborne virtual data centers depends not just on clever software, but also on durable, reliable chains of hardware and regulatory approvals.
Though the path is steep, this test marks a foundational step for Satlyt. It has validated its core planning software, proven that onboard execution is possible, and signaled what’s required for sharing real computing work across multiple satellites. Executing that vision at scale will be the real test—and what the next year or two are likely to deliver.
