AivexaNewsSearch
AI news for builders and product teamsChecked every hour

SpaceX and Google both said no. Now his software is riding a SpaceX rocket next to Google’s

Collected Oct 13, 2026

Satlyt has raised an $8 million seed round to build software that lets satellites run AI models on orbit, founder and CEO Rama Afullo told TechCrunch. The startup, founded after Afullo left SpaceX in 2024, is headquartered in Sunnyvale, California, and Nairobi; Afullo, a Kenyan American engineer, previously worked in Google's cloud computing business and had a brief stint at SpaceX's Starlink communications network.

A pitch he made internally at both companies was declined in each case, which led him to leave and build the company himself. The bet is that there is unused potential for computation aboard spacecraft, and that it can be sold as software rather than as hardware.

Satlyt's approach differs from that of the other well-known names in the emerging orbital data center field. SpaceX, Google and startups such as Starcloud and Cowboy Space Company are pursuing their own spacecraft for compute. Satlyt does not plan to build satellites at all. It writes software that runs AI models onboard, and is designed to work across hardware built by many different companies. Afullo frames the distinction with a consumer analogy: if SpaceX and its peers are building the iPhone, Satlyt intends to be Android, describing a horizontally integrated, open ecosystem. He also compares the business to VMware and Snowflake, platform companies whose customers run complex software without managing the underlying infrastructure.

Two demonstration missions have already flown Satlyt's software. The first commercial priority is operational efficiency. Spacecraft today typically depend on flight controllers on the ground to resolve problems, and getting data back to Earth, known as downlink, is slow and costly. Running inference onboard can cut both the delay and the bill, whether the task is diagnosing an anomaly or processing sensor readings that would otherwise be shipped home. Earlier this year Satlyt ran Google DeepMind's Gemma model onboard a spacecraft operated by Momentus. That deployment reduced the size of a transmission about onboard software errors by more than 60%, an efficiency Afullo says can translate into hundreds of thousands of dollars of savings per satellite each year.

This week's launch is the next step. Satlyt's software is set to fly on a spacecraft built by TakeMe2Space, an Indian startup that makes computing hardware for satellites, on a SpaceX rocket. It launches alongside the first prototype for Google's space data center effort, Project Suncatcher. Three customers are involved in the TakeMe2Space mission: NASA, which is paying to test protocols for cloud computing in space; Stellerian, a space surveillance startup tracking objects in orbit that wants to exercise image processing workloads; and TakeMe2Space itself, which wants to demonstrate that its satellite can host other companies' software. Running several unrelated workloads on one spacecraft is the proof point that a shared, multi-tenant model is workable in orbit.

The next milestone is harder. Satlyt expects to attempt a shared computing system, a cloud, spanning two different satellites next year. Success there would let it position itself as a true third-party compute cloud in orbit, selling capacity to spacecraft builders rather than operating just an efficiency layer on a single bird. Afullo describes the end state as turning a customer's satellite into a revenue-generating managed service. That is a meaningful shift in framing: instead of satellite operators buying software to cut their own costs, they would resell compute they host.

The investment was led by Houston-based Non Sibi Ventures. Partner Bernard Harris, who spent more than 20 years as a NASA astronaut, helped ground the firm's conviction in the space-based business, and partner Kent Lucas told TechCrunch that one attraction was that Satlyt does not depend on the most ambitious version of orbital data centers, which faces uncertain economics and a launch bottleneck. In Lucas's framing, the company does not need space data centers to succeed; growth can simply track the number of satellites going up. Afullo's stated ambition is to be live on 20% of satellites by the end of the decade, by which point he expects virtually every spacecraft builder to be putting GPUs and similar advanced processors onboard.

In practice, the near-term constraint is real: few high-powered GPUs are in orbit today, which means the workloads available to a software layer like Satlyt's are narrow for now. That is why the company's early work centers on inference that pays for itself through downlink savings and faster anomaly handling rather than on training or heavy compute. A likely trade-off is that a horizontal layer wins only if satellite builders standardize enough to make one software stack portable across vendors; a fragmented hardware market, each with its own tooling, would blunt the Android-style pitch. Inter-satellite compute adds further requirements, since a cloud spanning two spacecraft depends on links, orchestration and scheduling that no single operator controls.

Satlyt is among dozens of startups vetted by TechCrunch and appearing at Disrupt through its Battlefield competition, which begins Tuesday in downtown San Francisco.

Why it matters: Satellite operators and the hardware startups supplying them get another option that is software-only, which lowers adoption cost compared with launching custom compute spacecraft. For developers, the interesting question is whether orbital inference becomes a portable target, letting teams deploy models to many vendors' satellites the way they deploy to cloud regions today; Satlyt's two-satellite cloud attempt next year is the test of that idea, and it is an open question whether the economics hold before GPUs are common in orbit.

Read at TechCrunch · AI

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

Satlyt wants to be the Android of orbital computing, offering open software that works on many companies' satellites, versus SpaceX's closed, all-in-one iPhone-style approach.