Rama Afullo, a Kenyan American engineer and co‑founder of Satlyt, believes there is untapped potential for computing in orbit. Afullo previously worked in Google’s cloud business and had a brief stint at SpaceX’s Starlink in 2024; those experiences convinced him to leave and start a company building software that enables satellites to share large computing jobs.
Satlyt is headquartered in Sunnyvale, California, and Nairobi, Kenya. Unlike SpaceX, Google, or startups that build spacecraft, Satlyt focuses on software: a platform that will let operators run AI models on satellites regardless of the underlying spacecraft. The company told TechCrunch it has closed an $8 million seed round.
Why run AI on satellites?
Satlyt’s initial use cases focus on operational efficiencies. Many spacecraft rely on ground flight controllers to diagnose and resolve problems, and downlinking data to Earth is costly and can be slow. Onboard AI can make decisions in place or pre‑process sensor data, reducing the volume of data that needs to be sent to the ground and speeding responses to anomalies.
Earlier this year, Satlyt deployed Google DeepMind’s Gemma model on a spacecraft operated by Momentus. According to the company, the model reduced the size of transmissions about onboard software errors by more than 60% — a level of efficiency that Afullo says can save hundreds of thousands of dollars per satellite per year.
Upcoming launch and customers
Satlyt’s software is set to fly this week on a SpaceX rocket alongside the first prototype for Google’s Project Suncatcher. The software will run on a spacecraft built by TakeMe2Space, an Indian startup that designs computing hardware for satellites.
That mission carries three customers: NASA, which is paying to test cloud computing protocols in space; Stellerian, a space‑surveillance startup that wants to trial image‑processing workloads; and TakeMe2Space itself, which will demonstrate the satellite’s ability to host third‑party software.
Building a multi‑satellite cloud
Today there are relatively few high‑power GPUs in orbit, so the most immediate market is the kinds of workloads Satlyt is pursuing. The company’s next technical milestone is a demonstration of a shared compute system spanning two different satellites, which it expects to attempt next year. If successful, Satlyt could operate a true compute cloud in orbit as a third‑party provider to spacecraft builders.
"The way to think about it is that we turn your satellite into a revenue‑generating managed service," Afullo said.
Investment thesis and ambitions
Satlyt’s seed round was led by Houston‑based Non Sibi Ventures. Bernard Harris, a partner at Non Sibi and a former NASA astronaut with more than 20 years of experience, helped validate the firm’s conviction in the business model. Non Sibi partner Kent Lucas told TechCrunch the firm valued that Satlyt’s success does not require the most ambitious version of space data centers — a concept with uncertain economics and launch constraints — but can scale simply with the number of satellites launched.
Afullo hopes Satlyt will be live on about 20% of satellites by the end of the decade, anticipating that most spacecraft builders will begin including GPUs and other advanced processors in their designs. "If you’re putting up a satellite without putting up a GPU on it, at the very least, you’re doing yourself a disservice," he said.
Context and events
Satlyt is one of dozens of startups vetted by TechCrunch and appearing at Disrupt as part of the Battlefield competition in downtown San Francisco this week; the event’s program begins on Tuesday.
The reporting above is based on statements from Rama Afullo, Satlyt, and investor Non Sibi Ventures, as reported to TechCrunch.
