Satlyt raises $8M to run AI models on satellites
Original: Satlyt, founded by a former Google and SpaceX product manager, raises $8M to run AI on satellites
Why This Matters
On-orbit AI processing is emerging as a real cost-reduction lever as satellite fleets scale — software-layer plays like Satlyt could capture value across competing hardware providers.
Satlyt, a Sunnyvale/Nairobi-based startup founded by ex-Google and SpaceX PM Rama Afullo, closed an $8M seed round to build software enabling AI inference across multiple satellites. Its software launches Thursday on a SpaceX rocket alongside Google's Project Suncatcher prototype.
Rama Afullo pitched the idea of orbital data centers at both Google and SpaceX — and was turned down at both. He left SpaceX in 2024 and co-founded Satlyt, which now has offices in Sunnyvale and Nairobi. The company's pitch is to be the software layer — think VMware or Snowflake — for running AI workloads across satellites owned by different operators, rather than building any hardware itself. Afullo frames it as 'Android to SpaceX's iPhone.'
Satlyt has already completed two demonstration flights. In the most notable, it ran Google DeepMind's Gemma model aboard a Momentus spacecraft, compressing onboard error-log transmissions by over 60% — savings Afullo says can translate to hundreds of thousands of dollars per satellite annually. The logic: sending data back to Earth (downlink) is slow and costly, so processing it onboard cuts both latency and expense.
This week's SpaceX launch carries Satlyt's software on a TakeMe2Space satellite, with three customers on board: NASA testing space cloud computing protocols, and two others unnamed in available text. The $8M seed round was disclosed exclusively to TechCrunch.