Satlyt raises $8M to run AI software on satellites

In short: Satlyt raised $8 million to build software that helps satellites run AI without always sending data back to Earth.

What happened

Satlyt is a new company with offices in Sunnyvale, California and Nairobi. It was co-founded by Rama Afullo, a former product manager who worked at Google’s cloud team and later at SpaceX’s Starlink in 2024.

Afullo told TechCrunch he tried to pitch the idea of “computers in space” at both companies, but it did not move forward. He then helped start Satlyt, which focuses on software rather than building its own satellites.

The company says its software can run AI models directly on a satellite. This can reduce “downlink,” which is the process of sending data from space back to Earth. Think of downlink like paying for a slow and expensive international phone call, instead of handling the problem locally.

Satlyt has already tested its software on two demonstration missions. Earlier this year, it deployed Google DeepMind’s Gemma AI model on a spacecraft operated by Momentus, and said it cut the size of an error report transmission by more than 60%.

Satlyt’s software is also scheduled to launch on a SpaceX rocket this week. Customers on that mission include NASA, space surveillance startup Stellerian, and India-based TakeMe2Space.

Why it matters

If satellites can make more decisions on their own, they may need fewer back-and-forth messages with teams on the ground. That can save time and money, and it can make satellites quicker to respond when something goes wrong. Satlyt also says it wants its software to work across many companies’ satellites, similar to how Android works on phones made by different brands.

Source: TechCrunch AI