Satlyt has raised an $8 million seed round to develop software that lets satellites run AI models and eventually share computing workloads across orbit. Rather than building spacecraft, the startup wants to provide a common software layer that can work across satellites made by different companies.
Founder and CEO Rama Afullo previously worked in Google’s cloud business and at SpaceX’s Starlink network before starting Satlyt, which now has operations in Sunnyvale, California, and Nairobi. Afullo said he had previously pitched the idea of computing in space while at both companies but was turned down.
Satlyt Wants to Build Software Across Different Satellites
Satlyt’s approach differs from companies pursuing their own orbital data centers. Afullo compares the company to software infrastructure businesses such as VMware and Snowflake, with Satlyt providing a layer that lets operators run computing workloads without depending on one specific satellite platform.
The company says its software can process data directly aboard spacecraft, reducing the amount that must be transmitted back to Earth. Satellite downlinks can be costly and constrained by bandwidth and short communication windows, making onboard processing useful for tasks such as fault diagnosis, image analysis, and telemetry processing.
Satlyt has already flown its software on two demonstration missions. Earlier this year, it deployed Google DeepMind’s Gemma 3 1B model aboard a satellite, where the model analyzed software faults and reduced diagnostic payload sizes by about 64% in two benchmark scenarios.
That meant reducing one diagnostic transmission from 1,319 bytes to 469 bytes and another from 1,318 bytes to 464 bytes. Google DeepMind said the system could generate compact explanations of errors onboard rather than sending larger volumes of logs to ground operators.
Next Mission Will Test More In-Orbit Computing
Satlyt’s software is also set to run aboard a spacecraft built by Indian startup TakeMe2Space, which is developing orbital computing hardware. The satellite is designed to process data in orbit using Nvidia Orin NX processors rather than transmitting all raw information back to Earth.
According to the source report, the mission includes work for NASA, space surveillance startup Stellerian, and TakeMe2Space itself. Satlyt’s next major technical goal is to demonstrate a shared computing environment spanning two different satellites, which it expects to attempt next year.
The company’s seed round was led by Houston-based Non Sibi Ventures. Satlyt says its longer-term aim is to turn satellites into part of a distributed computing network, with Afullo targeting deployment on 20% of satellites by the end of the decade.
Satlyt is positioning that software as useful even if large orbital data centers develop more slowly than expected. Its immediate business depends instead on the growing number of satellites carrying more capable processors and needing software that can coordinate computing and data processing in orbit.
Featured image credits: Magnific.com
For more stories like it, click the +Follow button at the top of this page to follow us.