Satlyt, founded by a former Google and SpaceX product manager, raises $8M to run AI on satellites

Max Afullo’s Satlyt is building orbital AI operations software rather than spacecraft, aiming to be a horizontal layer across many companies’ birds. The $8 million seed coincides with a SpaceX launch carrying Satlyt software beside an early Google Project Suncatcher orbital compute prototype.

Satlyt has flown two demos; one deployed DeepMind’s Gemma on a Momentus spacecraft and cut transmission size about onboard software errors by more than 60%. This week’s TakeMe2Space mission includes NASA cloud-protocol tests, Stellerian space-surveillance image workloads, and hosting demos. Next year’s goal is shared compute spanning two satellites.

We’re interested in the unsexy orbital ops story, not the data-center-in-space hype reel.

Downlink is expensive; onboard triage is useful AI with a clear dollar metric. Staying software-only across vendors is the right bet if the industry fragments. Pairing a rideshare with Google’s Suncatcher prototype is great timing theater — Satlyt still has to prove multi-sat “cloud” orchestration next year.

I’m watching radiation-tolerant runtime reality versus Earth SaaS metaphors. Useful space AI will look more like anomaly compression than training GPT in LEO.

We’ll track the NASA and Stellerian workloads for receipts. Horizontal software beats another vertical sat company if the APIs stay open.

TechCrunch, October 1, 2026, Satlyt $8M seed exclusive.