Satlyt Raises $8 Million to Run Artificial Intelligence on Satellites

Satlyt has raised $8 million in a seed funding round led by Houston, US-based Noon CB Ventures. Headquartered in Sunnyvale, California, and Nairobi, Kenya, the company develops software that runs AI models on satellites manufactured by other companies instead of building its own constellation. Rama Afulu, a former product manager at Google Cloud and SpaceX's Starlink, serves as the company's founder and CEO.

The company stated that it previously succeeded in reducing the size of a software bug report signal sent from a satellite by over 60% using Google DeepMind's Gemma model aboard a spacecraft operated by Momentus. This means processing takes place on board the satellite itself before transmitting data to Earth, which reduces the volume of sent information, eases pressure on communication channels, and allows for faster response times at lower transmission costs.

The space sector is shifting toward running AI models on board satellites instead of sending all images and data to ground stations for analysis, as transmitting raw data is costly and time-consuming depending on communication window availability. In contrast, this trend imposes constraints on model size, processor capacity, and power availability on board the satellite—limitations that make companies providing lightweight software capable of operating under these conditions a focal point for investors.

This trend benefits Arab markets that are expanding satellite projects and Earth observation services, as running AI on board satellites lowers data transmission costs and reduces the need for massive ground infrastructure. Furthermore, the entry of new investors into this space indicates that the space software market is now attracting venture capital alongside government investments, a shift that opens up opportunities for startups in the region.

What do these terms mean?

Seed round: The initial external funding a startup receives to build its product and test the market. Onboard computing: Running software and AI models inside the satellite itself instead of sending raw data to Earth. Gemma model: A small AI model developed by Google DeepMind that runs efficiently on power-constrained devices. Earth observation: The use of satellite imagery in agriculture, navigation, and early disaster warning. Hosted spacecraft: A satellite owned and operated by a third party that enables software companies to test their applications on it.