Satlyt raises $8 million for orbital AI compute platform

Space software startup Satlyt has raised an $8 million seed round led by Houston-based Non Sibi Ventures to build an operating system for satellite computing. Founded by former Google and SpaceX engineer Rama Afullo, the Sunnyvale- and Nairobi-based firm aims to act as a horizontal software layer, akin to Android or VMware, allowing space hardware from different vendors to execute artificial intelligence tasks in orbit without relying on ground station control.

Satlyt’s software is scheduled to launch aboard a SpaceX rocket alongside Google’s Project Suncatcher prototype. The software will run on hardware designed by TakeMe2Space, an Indian satellite infrastructure startup. The mission features three main workloads: NASA is paying to test space cloud computing protocols, space surveillance startup Stellerian is evaluating real-time image processing, and TakeMe2Space is demonstrating multi-tenant software hosting capabilities.

In prior demonstration missions on a Momentus spacecraft, Satlyt deployed Google DeepMind’s Gemma AI model directly in orbit. By processing telemetry and error diagnostics locally, the system reduced onboard software error transmission sizes by over 60%. Afullo notes that minimizing expensive downlink bandwidth in this manner can save satellite operators hundreds of thousands of dollars per spacecraft each year.

Former NASA astronaut Bernard Harris, now a partner at Non Sibi Ventures, backed the startup alongside partner Kent Lucas, who emphasized that Satlyt’s growth depends simply on rising satellite counts rather than high-risk orbital data center deployments. Next year, Satlyt plans to demonstrate a unified cloud computing network spanning two separate satellites, with a long-term goal of powering 20% of all operational satellites by the end of the decade.