Satlyt raises $8M to put AI on satellites, with one model already in orbit

Modern satellites collect more data than they can beam back to Earth. Downlink windows are short, ground stations are scarce, and radio spectrum is shared. The constraint shapes the architecture: anything that needs to be acted on quickly has to be processed on the spacecraft, or it waits. The same logic that pushed AI from central data centers to base stations and phones now pushes it into orbit.

Satlyt, a Sunnyvale- and Nairobi-based startup, is building that on-orbit layer. The company says it has already run Google's open Gemma model on a partner satellite to analyze system logs, software errors, and stack traces without round-tripping to a ground data center. That deployment accompanies a fresh $8 million seed round Satlyt announced, led by non sibi ventures with participation from TLCOM, Antler, Slauson & Co., Launch Africa Ventures, Enza Capital, Askya Investment Partners, Demos, BAG Collective, Gaingels, Axian Investment, and existing backers. The capital funds product development, hiring, and wider deployment of Satlyt's software on third-party satellites.

The Gemma deployment puts Satlyt in a narrow category. Most space-AI announcements describe a future capability, often tied to a next-generation processor that has not flown. Satlyt's claim is operational: a model is on a satellite, reading telemetry and producing decisions the spacecraft can act on locally. The company did not name the partner for that first deployment, and the partner's identity is the verification gap. The $8M figure and non sibi ventures' lead role also rest on a single regional report at the time of writing and have not been independently confirmed.

The cap table mirrors the operating footprint. Askya, Launch Africa, and Enza anchor the African side; non sibi ventures, Slauson & Co., Gaingels, and Axian bring the US and global-angel flank. Satlyt runs a Sunnyvale headquarters and a Nairobi engineering base where most of the company's software is built. The leadership is Kenyan-American, and engineering headcount sits predominantly in Nairobi.

Two further applications are scheduled to fly on a third-party spacecraft, with no launch date yet. One is research software tied to Satlyt's work under NASA's Small Business Technology Transfer (STTR) program with NASA's Glenn Research Center and the University of Houston. The other is a commercial imagery-processing workload aimed at customers who would rather have the analysis than the raw pixels. Both are forward statements, not current operating evidence.

Satlyt also describes a longer-term target: an orbital compute network. The idea is a common software layer that coordinates processing across satellites owned by different operators, so workloads can be routed to whichever spacecraft has spare capacity and the right sensors. The company pitches it as the equivalent of a cloud region in space. That is a stated objective, not a current product, and the round funds the nearer work of getting the software onto more buses first.

If the STTR-linked research payload and the commercial imagery demo both reach orbit on the third-party bus, the category moves from one verified deployment to a small fleet of them. Until then, Satlyt's orbital-edge-AI claim rests on a single named-model run of Gemma in orbit and the team that built it.