Satlyt raises $8M to run its AI software on satellites other companies build

A satellite throws a software fault. Before anyone can fix it, the spacecraft has to explain the problem to the ground. Spacecraft typically depend on flight controllers on Earth to solve problems, and the downlink that carries the explanation is expensive and often slow ^[7]. In Satlyt's Momentus test, the error report that came down was less than 40% of its original size ^[1]. Afullo says efficiency of that kind can save hundreds of thousands of dollars per satellite each year ^[9]. That figure is his own estimate ^[9].

The pitch goes well beyond error reports. Afullo compares Satlyt to VMware and Snowflake, platforms that let users run complex software without worrying about the infrastructure underneath ^[4]. Of companies like SpaceX that are building orbital data centers, he said, "if they are the iPhone, we'll build Android as a horizontally integrated, open ecosystem" ^[5]. He hopes to be live on 20% of satellites by the end of the decade, when he expects virtually every spacecraft builder to fly GPUs ^[17]. Few high-powered GPUs are in orbit today ^[12].

What customers on Thursday's flight actually buy is a test. NASA's money goes to testing protocols for cloud computing in space, and Stellerian's workload is image processing it wants to try out ^[11]. The report does not say what NASA is paying ^[11]. What the team tells itself customers will eventually buy is a shared cloud in orbit. Its first attempt at a computing system spanning two satellites is planned for next year ^[13].

This week's flight does have the split a software-only market needs. The satellite's builder, TakeMe2Space, makes computing hardware for satellites, and it is flying Satlyt's software partly to prove it can host other companies' code ^[10][11]. The connection to Google's Project Suncatcher is a shared SpaceX rocket ^[2]. Beyond that, the record is two demonstration missions and this week's flight, as reported by TechCrunch ^[6][10].

The lead investor's case is narrower than the founder's. Non Sibi Ventures invested partly because Satlyt does not depend on space data centers, a vision facing uncertain economics and a launch bottleneck ^[15][18]. "We don't need data centers in space for Rama to be wildly successful, right?" partner Kent Lucas told TechCrunch. "It can just be driven by the number of satellites going up." ^[16]

An operator weighing an offer like this can sort it on two axes. The first is whether the workload runs on one satellite or needs several. The second is whether it cuts money already spent on downlink and ground controllers, or promises new revenue from hosting other companies' code. One satellite and existing spend is what Satlyt has flown, and it can be checked against a downlink bill ^[8]. One satellite and new revenue is TakeMe2Space's test this week, and it is the quadrant Afullo means when he says "we turn your satellite into a revenue-generating managed service" ^[11][14]. Both multi-satellite boxes wait on next year's two-satellite attempt ^[13]. I'd buy the first quadrant on measured results and pay for the second as a pilot. The tradeoff, I'd expect, is that early hosting revenue goes to builders willing to fly a stranger's workload before any cloud exists, as TakeMe2Space is doing this week ^[11].