Satlyt secures $8M to operate AI workloads across satellites
Satlyt, a satellite-software startup founded by former Google Cloud and SpaceX Starlink product manager Rama Afullo, has raised an $8 million seed round to run artificial intelligence workloads in orbit. The company, headquartered in Sunnyvale, California, and Nairobi, is preparing to launch its software this week on a SpaceX rocket aboard a spacecraft built by Indian startup TakeMe2Space.
Rather than designing and launching its own spacecraft, Satlyt is building software intended to let satellite operators run AI models and computing workloads across different satellite platforms. Afullo compares the approach with infrastructure platforms such as VMware and Snowflake: users should be able to run complex software without needing to manage the underlying hardware.
Software-first approach to orbital computing
Satlyt has already flown its software on two demonstration missions. Its initial use case is improving spacecraft operations that often rely on ground-based flight controllers to diagnose anomalies and make decisions. Downlinking data to Earth can be expensive and slow, so processing sensor readings and operational information onboard could reduce the amount of data sent to the ground.
Earlier this year, Satlyt deployed Google DeepMind's Gemma model on a spacecraft operated by Momentus. The company said the model reduced the size of a transmission concerning onboard software errors by more than 60%. Afullo said efficiencies of that kind can save hundreds of thousands of dollars per satellite annually.
Three customers on the upcoming mission
The TakeMe2Space mission will carry three customer workloads. NASA is paying Satlyt to test protocols for cloud computing in space. Space-surveillance startup Stellerian plans to test image-processing workloads, while TakeMe2Space will demonstrate that its satellite can host software from other companies.
Satlyt's seed round was led by Houston-based Non Sibi Ventures. The investor's thesis does not depend on the immediate arrival of large orbital data centers, whose economics and launch capacity remain uncertain. Instead, the company can address operational computing needs as more satellites are launched with advanced processors.
From individual spacecraft to a shared cloud
High-powered GPUs remain scarce in orbit, which limits the near-term scope for intensive AI workloads. Satlyt's next planned project is to create a shared computing system spanning two satellites, an attempt it expects to make next year. Success would move the company closer to offering a third-party computing cloud for spacecraft builders.
For satellite operators, the immediate business case is not a distant orbital data center but lower downlink demand and more autonomous handling of specific onboard tasks. Teams evaluating new satellites can therefore weigh AI-capable processors and portable software against the cost of transmitting and resolving operational data from the ground.