Satlyt Raised $8 Million for In-Orbit AI Processing
The startup aims to minimize data transmission by running machine learning models directly on satellites.
Updated on Oct. 13, 2026 in Artificial Intelligence
Live Poll
Is processing data onboard satellites a more efficient way to manage information than sending it back?
Sunnyvale-based startup Satlyt has raised $8 million in seed funding to expand its orbital AI network. Founded in 2024, the company successfully ran the Google DeepMind Gemma model on satellites to process data in space.
Why it matters
By analyzing telemetry and imagery in orbit, the technology significantly reduces the raw data volume sent back to Earth, optimizing bandwidth for satellite operators. The company plans to scale this by connecting spacecraft into a shared computing network.
Satlyt demonstrated the Google DeepMind Gemma model running in orbit at speeds of 22.71 to 25.48 tokens per second. The onboard system performs local analysis of system logs and imagery before transmission to Earth.
The players
Satlyt
An AI startup based in Sunnyvale and Nairobi that integrates edge computing layers into satellite fleets.
Rama Afullo
The founder of Satlyt who established the company in 2024.
Non Sibi Ventures
A venture capital firm that led the $8 million seed funding round for Satlyt.
Google DeepMind
The research division behind the Gemma model, an open-weights machine learning architecture used by Satlyt for orbital testing.
The details
Satlyt uses an abstraction layer that allows AI models to function across heterogeneous satellite fleets. By processing raw telemetry, imagery, and system logs locally, the software filters actionable insights before transmission to Earth. This approach shifts compute duties to the edge, preventing the need to downlink massive datasets that require ground-based analysis.
Timeline
Satlyt was founded in 2024.
The funding and deployment updates were reported on October 13, 2026.
The Tech Race
Satlyt’s development marks a transition from Small Business Technology Transfer project research to commercial implementation. The company is now racing to establish a shared orbital computing standard, competing against traditional ground-heavy satellite processing architectures.
This technology will initially affect aerospace operators and data-intensive satellite fleets by lowering bandwidth costs and latency. General users may eventually experience faster data delivery and more responsive services from satellite-enabled network infrastructure.
The takeaway
Satlyt's successful deployment of the Gemma model in orbit proves that high-level AI analysis is viable in space environments. Investors should watch for the company's progress in connecting multiple satellite fleets into a unified, shared computing network.
Further reading
For more on how machine learning is moving to the edge, visit our Artificial Intelligence section.
Live Poll
Is processing data onboard satellites a more efficient way to manage information than sending it back?
