Key Takeaways

  • Non Sibi Ventures led Satlyt’s $8 million seed round to fund its satellite AI software.
  • Satlyt aims to run models across spacecraft from multiple operators instead of building satellites.
  • Onboard processing could reduce downlink demands and deliver Earth-observation insights faster.

Rama Afullo saw orbital data centers taking shape before founding Satlyt. The Kenyan American engineer had worked in Google Cloud and as a SpaceX Starlink product manager, experience that put him close to two industries now converging: distributed computing and commercial space infrastructure.

Satlyt has raised an $8 million seed round led by Non Sibi Ventures, according to TechCrunch. The startup plans to use the financing to develop software that allows artificial intelligence workloads to run directly aboard satellites, reducing the need to transmit every piece of raw data to Earth before it can be analyzed.

That distinction matters. Satlyt is not proposing another spacecraft manufacturing operation. Its software-first model is intended to work across satellite hardware owned by different operators, somewhat like a cloud-computing layer adapted to the constraints of orbit. Those constraints include limited power, uneven connectivity, radiation exposure and processors that may remain in service for years.

The approach has already moved beyond a presentation deck. Reported work includes running Google DeepMind’s Gemma model on a Momentus spacecraft, while future deployments are expected to involve TakeMe2Space. NDTV Profit also reported the financing and Satlyt’s effort to take AI processing into orbit.

Satellites already collect more information than operators can economically send through ground networks. High-resolution imagery and sensor streams compete for limited downlink windows. Processing data onboard can help a spacecraft identify clouds, wildfires, maritime activity or other relevant features before transmitting a smaller, more useful result.

This approach eliminates the need to transmit an entire image when a customer primarily needs a specific alert.

That proposition is particularly relevant to Earth observation, where governments and commercial customers are seeking analytics closer to real time. Onboard AI could reduce latency, conserve bandwidth and make some services less dependent on constant access to ground infrastructure. It could also help operators prioritize transmissions when connectivity is constrained.

The market context is favorable, though hardly simple. The global space economy reached $630 billion in 2023, with commercial activity representing 78% of the total. Meanwhile, the number of active satellites exceeded 10,000 by 2025. That expanding installed base gives interoperable computing software a potentially broad addressable market without requiring Satlyt to finance launches or manufacture spacecraft.

Still, portability is a serious engineering challenge. Satellite processors differ substantially, as do operating environments, mission requirements and risk tolerances. Satlyt will need to demonstrate that workloads can be deployed, monitored and updated across varied systems without compromising mission operations. Compatibility with standards from the Consultative Committee for Space Data Systems and established frameworks such as core Flight System could help, but standards alone do not erase hardware differences.

There is also a sales issue. Spacecraft operators tend to be cautious about introducing new code after launch, for understandable reasons. Satlyt may find its earliest opportunities in newer commercial missions designed around software updates and onboard accelerators, rather than older satellites with tightly constrained computing resources.

Competition is forming as well. Starcloud, Cowboy Space Company and SpaceX are pursuing adjacent parts of the orbital-computing market. Some efforts emphasize data-center infrastructure in space; others focus on vertically integrated satellite networks. Independent Space News tracks a broader commercial space sector in which computing, connectivity and spacecraft operations are increasingly overlapping.

Satlyt’s differentiation rests on being an independent software layer rather than owning the underlying constellation. That could allow operators to retain control of their spacecraft while gaining access to deployable AI tools. Or it could leave Satlyt dependent on partners for hardware access and mission schedules. Both possibilities are real.

The $8 million round gives Satlyt capital to prove that its architecture works across more than one demonstration mission. Its next tests will be practical ones: supporting different processors, managing models remotely, securing orbital workloads and showing customers that faster analysis produces economic value. If the software layer proves viable across multiple constellations, satellite data may increasingly arrive on Earth already filtered, classified and ready for action.