Satlyt Secures $8M Funding to Propel AI Development in Satellite Technology, Founded by Ex-Google and SpaceX Product Leader

Rama Afullo anticipates a future dominated by orbital data centers.

The Kenyan American engineer, who previously held positions at Google Cloud and briefly worked for SpaceX’s Starlink in 2024, believes there is significant potential for computational capabilities in space.

“During my time at SpaceX, I proposed this internally, but they declined. The same happened at Google,” Afullo shared.

This led him to leave SpaceX and co-establish Satlyt, headquartered in Sunnyvale, California, and Nairobi. The company aims to develop software for satellites that will enable them to collectively handle extensive computing tasks in space. Afullo disclosed that they have successfully secured $8 million in seed funding.

The market conditions are favorable for Afullo. SpaceX has invested heavily in orbital data centers, and Satlyt’s software is poised to launch on a SpaceX rocket together with the initial prototype of Google’s space data center initiative, Project Suncatcher.

Unlike companies such as SpaceX, Google, or other startups like Starcloud and Cowboy Space Company, Satlyt is not designing its own spacecraft. Instead, Afullo’s team is focused on creating software that operates AI models on satellites, likening their approach to VMWare and Snowflake by providing platforms that allow users to run complex software without managing the underlying systems.

“While SpaceX represents the iPhone of orbital data centers, we aim to build the Android—a horizontally integrated, open ecosystem,” he explained, emphasizing Satlyt’s intent to deliver software that functions across various companies’ satellites.

Satlyt has already tested its software on two demonstration missions. Their primary goal is to enhance operational efficiencies for spacecraft, which often rely on flight controllers on the ground to address issues. Transmitting data back to Earth can be costly and slow, so utilizing AI for on-board decisions can lead to considerable savings, whether in managing anomalies or processing sensor data locally.

Recently, Satlyt implemented Google DeepMind’s Gemma AI model on a spacecraft managed by Momentus. This model achieved a reduction in the size of communications linked to onboard software errors by over 60%, an efficiency that could save each satellite hundreds of thousands of dollars annually, according to Afullo.

The upcoming launch will see Satlyt’s software deployed on a spacecraft created by the Indian startup TakeMe2Space, known for manufacturing satellite computing hardware. This mission will cater to three clients: NASA, which is funding experiments related to cloud computing in space; Stellerian, a startup engaged in space surveillance looking to test image processing tasks; and TakeMe2Space, showcasing their satellite’s software hosting capabilities.

Currently, the availability of high-power GPUs in orbit is limited. Nonetheless, Satlyt is concentrating on the types of workloads they’re equipped to handle. Their next objective is to demonstrate a shared computing system—a cloud—across two different satellites, which they plan to initiate next year. Successful implementation would enable them to create a true compute cloud in orbital space available to spacecraft manufacturers.

“We can transform your satellite into a managed service that generates revenue,” Afullo stated.

Satlyt’s seed funding round was spearheaded by Non Sibi Ventures from Houston. Partner Bernard Harris, a former NASA astronaut, reinforced the firm’s belief in this innovative space-based venture.

Their investment rationale stemmed from the fact that Satlyt does not rely heavily on a luxurious vision of space data centers, which currently poses economic challenges and a potential launch schedule constraint. “Rama can achieve significant success without needing data centers in space,” remarked Non Sibi partner Kent Lucas.

By the end of the decade, Afullo aims for Satlyt’s software to be integrated into 20% of satellites, anticipating that virtually all satellite manufacturers will begin equipping their spacecraft with advanced processing units.

“If you’re launching a satellite without incorporating a GPU, you’re likely missing an opportunity,” he remarked.

Satlyt is among numerous startups showcased at a recent tech event in San Francisco as part of a competitive battlefield segment.