Satlyt has raised an $8 million seed round to develop software for running AI on satellites, pursuing a different route into orbital computing from companies building their own spacecraft. The startup wants its software to operate across satellites supplied by multiple manufacturers and eventually let them share computing work. The funding gives that plan backing before the proposed cross-satellite cloud has been demonstrated.
Founder Rama Afullo disclosed the round to TechCrunch. The company has headquarters in Sunnyvale, California, and Nairobi. Afullo previously worked in Google’s cloud business and briefly at SpaceX’s Starlink network in 2024, experiences he says helped convince him there was a market for computing infrastructure in orbit.
A software layer rather than another spacecraft
Satlyt’s approach is to provide the software that lets AI workloads run on spacecraft built by others. It is not planning to manufacture its own satellites. Afullo compares the ambition to computing platforms that allow customers to use infrastructure without managing every underlying component themselves.
The initial work is more modest than a vast orbital data centre. Spacecraft often send information back to Earth so ground teams can identify and resolve problems. Satlyt is exploring whether onboard models can handle some of that work locally, reducing the amount of information that needs to be transmitted.
TechCrunch reports that Satlyt has already flown its software on two demonstration missions. Earlier this year, it deployed Google’s Gemma model on a spacecraft operated by Momentus. The model reduced the size of a transmission describing onboard software errors by more than 60%, according to the report.
Afullo says efficiencies of that kind could save hundreds of thousands of dollars per satellite annually. That savings estimate is his projection, not a financial result established by the transmission experiment. The reported reduction demonstrates a narrower outcome: less information needed to be sent in that particular use case.
Customers will test distinct computing tasks
The October 1 report describes a forthcoming launch carrying Satlyt’s software on a spacecraft built by Indian startup TakeMe2Space. The same SpaceX flight was expected to carry Google’s first Project Suncatcher prototype. Sharing a launch does not mean the two efforts have the same business model or computing architecture.
Satlyt’s mission involves three customers with different objectives. NASA is paying it to test protocols for cloud computing in space. Stellerian wants to test image-processing workloads for tracking objects in orbit. TakeMe2Space itself wants to demonstrate that its satellite can host software supplied by other companies.
Those tests focus on practical workloads that existing spacecraft could support. TechCrunch notes that few powerful GPUs are currently in orbit, making today’s available hardware an important constraint on what can be attempted.
The next step in Satlyt’s plan is more ambitious: a shared computing system spanning two separate satellites. The company expects to attempt that demonstration next year. A successful test would advance its plan to become a third-party orbital cloud provider, but the source does not report that a multi-satellite service is already operating.
Investors back a nearer-term orbital business
Houston-based Non Sibi Ventures led the seed round. Partner Bernard Harris’s experience as a NASA astronaut helped inform the firm’s interest, according to TechCrunch. Another partner, Kent Lucas, said the investment does not depend on the most expansive version of orbital data centres becoming economically viable.
That rationale fits Satlyt’s initial focus on improving individual spacecraft operations. Its opportunity could grow with satellite deployments even if large computing clusters in space remain constrained by launch capacity and uncertain economics. The proposed network would then build on that more immediate software role.
Afullo hopes Satlyt will be operating on 20% of satellites by the end of the decade. That is a company ambition, alongside his expectation that advanced processors will become much more common aboard spacecraft. Neither forecast is demonstrated by the seed funding or the existing missions.
The distinction between achieved results and future plans is especially important here. Satlyt has funding, reported demonstration experience and identified customers for another test. Its broader promise—turning diverse satellites into a managed compute cloud—still depends on showing that separate spacecraft can reliably share useful work.
