Satlyt Secures $8M Seed to Run AI on Satellites Without Building Hardware

Sunnyvale and Nairobi-based startup Satlyt has secured an $8 million seed investment, led by non sibi ventures, to expand its software that enables artificial intelligence (AI) models to run directly on satellites. Co-founded by former Google and SpaceX product manager Rama Afullo, the company intends to facilitate multiple spacecraft sharing resource-intensive computing tasks while orbiting Earth.

Unlike competitors like SpaceX, Google, or Starcloud, Satlyt does not produce satellites; rather, it develops a universal operating system to run across various satellite brands. Afullo likens the company's architectural approach to enterprise solutions such as VMware and Snowflake.

Satlyt markets its software as an open ecosystem. While orbital data center innovators resemble the iPhone, Satlyt aims to provide the underlying infrastructure. The company's software has been deployed on two demonstration missions, significantly decreasing the financial delay in transmitting vast raw datasets back to Earth. In January, Satlyt's platform utilized Google DeepMind's Gemma AI model aboard a Momentus spacecraft, resulting in a transmission of software errors reduced by over 60%.

The software will launch alongside Google's Project Suncatcher prototype on a SpaceX rocket this week, supporting NASA's cloud computing protocols, Stellerian image processing, and TakeMe2Space operations. The recently raised funds will be used to construct a shared computing system across two separate satellites next year, enabling Satlyt to provide an in-orbit compute cloud as a third-party service.

This strategy transforms standard satellites into revenue-generating managed services. As one of the top 5% of US enterprise software seed deals by capital raised, the $8 million investment underscores the strong investor interest in software-driven space infrastructure.