Satlyt, a fresh startup co-founded by engineer Rama Afullo—formerly with Google Cloud and SpaceX—has secured $8 million in seed funding to enable artificial intelligence operations directly from satellites. With dual bases in Sunnyvale, California, and Nairobi, Kenya, the company aims to make satellites smarter and more autonomous. Rather than building spacecraft, Satlyt is focused on software platforms that can run complex AI workloads in orbit. This approach positions it alongside cloud-platforms like VMware or Snowflake—but for space.
What Satlyt is Tackling
Satlyt’s core mission is to reduce how much spacecraft depend on ground control centers. Traditionally, satellites send large amounts of raw sensor data to Earth for processing, which can be slow, expensive, and inefficient. Satlyt wants more of that processing done onboard, using AI to detect issues, process images, and even handle anomalies without waiting for ground-based responses. This on-orbit intelligence has already been trialed: earlier this year, Satlyt deployed DeepMind’s Gemma model aboard a spacecraft operated by Momentus. That mission achieved a cut of more than 60% in transmission size for error reports—translating to substantial cost savings per satellite annually.
Satlyt is also participating in a launch this week. Its software will fly on a satellite built by TakeMe2Space, an Indian hardware startup. The mission carries payloads for three clients: NASA (testing cloud computing protocols in orbit), Stellerian (image processing for orbital surveillance), and TakeMe2Space itself (validating that different software applications can coexist on the same satellite). These demonstrations reflect Satlyt’s strategy to serve multiple customers via a shared platform, rather than building and selling its own satellites.
Vision & Market Positioning
While giants like SpaceX are building orbital data centers, and Google is working on its own space computing efforts, Satlyt’s concept is to build software infrastructure compatible across many spacecraft providers. Afullo envisions Satlyt as the “Android equivalent” in space, enabling open ecosystems of satellite hardware using its AI software.
The funding round was led by Non Sibi Ventures, a Houston-based firm with former NASA astronaut Bernard Harris among its leadership. The investor’s reasoning includes the reality that Satlyt doesn’t require full-scale space data centers to succeed; success can grow alongside the rise in satellites with onboard GPUs and advanced processors. Satlyt targets being deployed on roughly 20% of satellites by the end of the decade—a goal tied to the broader trend of integrating higher compute capabilities into spacecraft.
As of now, there are relatively few satellites in orbit that carry high-powered GPUs. Satlyt’s near-term focus is on workloads that demand agility—like anomaly detection, image processing, error correction—rather than training large generative AI models up in space. Looking ahead to next year, the company plans to build a shared computing structure spanning at least two satellites, bringing the idea of a third-party orbital compute cloud closer to reality.
Satlyt was founded after Afullo’s efforts at Google and SpaceX to pursue in-space compute went unheeded. He launched the company to realize the concept of “orbital data centers” in a different way: by offering software tools that let satellites themselves become active computing nodes. With this week’s rocket launch and the newly raised capital, Satlyt begins operationalizing that vision.
Satlyt’s entry comes amid growing interest in on-orbit intelligence from both governments and commercial entities. Advances in miniaturizing compute hardware, falling costs for launching payloads, and the rising demand for real-time data from space all feed into the market opportunity. Yet technical challenges like power constraints, radiation-hardening, and maintaining reliability in space remain significant.
Analysis: Satlyt’s strategy reflects a turning point in space infrastructure. Rather than compete on hardware or volume of satellites, it’s building the software layer that turns fast-growing satellite constellations into intelligent clouds—running AI at the edge of Earth’s orbit. If successful, this could reshape how real-time space data is processed, reduce costs dramatically, and redefine what kinds of missions are practical. Key things to watch: whether its shared cloud across satellites can match reliability of ground-based systems, and how many satellite makers adopt Satlyt’s software by 2030.
