The ex-Googler and SpaceX alum who got rejected by both is now launching software on their rockets, turning "no" into orbital irony.
The Summary
- Satlyt raised $8M to build open software that runs AI on satellites, positioning itself as the Android to SpaceX's iPhone approach in orbital computing
- Founded by a former Google and SpaceX product manager who pitched both companies and got turned down, then built it anyway
- The platform works across multiple satellite manufacturers, creating infrastructure for edge AI 200 miles above Earth
The Signal
Most edge computing happens at ground level. Cell towers, factories, autonomous cars processing data locally instead of sending everything to the cloud. Satlyt is moving that same principle into orbit, running AI models directly on satellites instead of beaming raw data back to Earth for processing.
The founder's story matters because it shows how infrastructure gets built. He worked at Google and SpaceX, saw the opportunity, pitched both. They passed. So he built the thing they said no to, and now it's launching on a SpaceX rocket alongside Google's hardware.
"Both your former employers rejected this idea. Now you're literally sharing their rocket."
The company wants to be the Android of space, an open platform that works on satellites from multiple manufacturers. SpaceX runs a closed system where their satellites, rockets, and software all come from one company. Tight integration, high performance, zero flexibility if you're not in their ecosystem.
Satlyt's bet is different:
- Satellite manufacturers want AI capabilities without building their own
- Customers want to switch providers without rewriting software
- The orbital compute market will look more like Android's fragmented dominance than iPhone's walled garden
The $8M round funds what happens when you run neural networks in space. Satellites already collect massive amounts of imagery and sensor data. Most of it gets downlinked to ground stations, processed in data centers, then acted on hours or days later. If the satellite can run the model itself, it can filter out useless data, spot changes in real time, and only send back what matters. Less bandwidth wasted. Faster decisions. Same principle as your phone processing "Hey Siri" locally instead of streaming your voice to Apple's servers.
The Implication
Infrastructure layer plays always take longer than the applications built on top, but they capture more value. Satlyt is building rails for orbital AI before most people realize orbital AI is a category. If they're right that the satellite market fragments like smartphones did, being the open platform beats being vertically integrated.
Watch whether other satellite manufacturers adopt this or build their own. Google and SpaceX already chose to build. The next tier down might choose to buy.