Satlyt raises $8 million to run AI software across many companies' satellites
The Sunnyvale and Nairobi startup builds software, not spacecraft, to let satellites process data in orbit instead of sending it all back to Earth.
By Eric Powlesland 3 min read
Satlyt, a startup with offices in Sunnyvale, California, and Nairobi, has raised an $8 million seed round to build software that runs AI models on satellites. As TechCrunch reported, the round was led by Houston-based Non Sibi Ventures.
The company was co-founded by Rama Afullo, an engineer who worked in Google's cloud business and briefly at SpaceX's Starlink network in 2024. He told TechCrunch he pitched the idea of computing in orbit at both companies and was turned down each time.
Satlyt is not building spacecraft. Companies like SpaceX, Google, Starcloud and Cowboy Space Company are pursuing their own orbital hardware. Satlyt wants to supply the software layer that works on satellites built by many different companies. Afullo compares SpaceX's approach to the iPhone, a closed, all-in-one product, and positions Satlyt as the Android equivalent: an open system that other builders can adopt.
The software has already flown on two demonstration missions. Earlier this year, Satlyt ran Google DeepMind's Gemma AI model on a spacecraft operated by Momentus. According to the company, the model shrank a transmission about onboard software errors by more than 60%. Afullo says that kind of efficiency can save hundreds of thousands of dollars per satellite each year, though that figure is his estimate.
A new mission was set to launch on a SpaceX rocket, on a spacecraft built by Indian startup TakeMe2Space. NASA is paying Satlyt to test cloud computing protocols in space, a space surveillance startup called Stellerian wants to test image processing, and TakeMe2Space wants to show its satellite can host outside software. Google's first prototype for its own space data center effort, Project Suncatcher, was scheduled on the same launch.
What it means for your business
For most small and mid-sized businesses, the direct impact today is close to zero. You are not going to rent computing time on a satellite next quarter.
The underlying idea is still worth understanding, because it is the same one that shows up on the ground. Sending data back to a central location is slow and costly, so it often makes sense to process it where it is collected and send back only the results. For satellites, that means fewer expensive downloads to Earth. For a business, the same logic applies to security cameras, sensors on equipment, or field devices that do some analysis locally before uploading anything.
The second lesson is about platforms. Satlyt is betting that an open layer working across many vendors' hardware can compete with a single company that controls everything. That choice between a closed, integrated system and a more flexible, mixed one is familiar to anyone who has picked business software or phones for a team.
If your business uses satellite imagery, mapping, weather, agricultural or maritime data, this is a trend that could eventually make those services faster or cheaper. When and by how much is not yet clear.
Investors appear to be backing the more modest version of the story. Non Sibi partner Kent Lucas told TechCrunch, "We don't need data centers in space for Rama to be wildly successful, right?" His point was that growth can come simply from the number of satellites being launched.
What to watch
- A shared cloud across two satellites. Satlyt expects to attempt this next year. Success would be the first step toward offering a real computing cloud in orbit.
- Adoption targets. Afullo hopes Satlyt's software will run on 20% of satellites by the end of the decade, when he expects most builders to include GPUs and similar processors.
- Hardware in orbit. Few high-powered GPUs are in space right now, which limits what any of these companies can do in the near term.
