Analysis
Satlyt, a startup that lets satellites share AI computing jobs in orbit, raised an $8 million seed round led by Non Sibi Ventures, a Houston-based firm whose investing partner on the deal was Bernard Harris, a former NASA astronaut, according to TechCrunch. Founder Rama Afullo, a Kenyan American engineer who previously worked in Google's cloud computing business and briefly at SpaceX's Starlink unit in 2024, says both companies turned down his internal pitches for orbital computing before he left to build Satlyt independently.
A Multi-Vendor Bet Against SpaceX's Closed Approach
Satlyt doesn't build spacecraft -- it builds the software layer that lets multiple satellite operators pool idle onboard compute into something closer to a shared, multi-vendor data center in orbit. Afullo compares the model to VMware and Snowflake: infrastructure that sits on top of hardware other companies already own, rather than another company racing to launch its own satellites. That positioning puts Satlyt directly against SpaceX, Google's Project Suncatcher, Starcloud, and Cowboy Space Company -- all of which are building their own orbital data centers or spacecraft rather than an open layer that works across operators' existing fleets.
The company's early technical proof point: a Google DeepMind Gemma model deployed on a Momentus spacecraft cut transmission size by more than 60%, processing sensor data onboard before it needs to be sent back to Earth. Satlyt has flown two demonstration missions so far, and its software is scheduled to launch this week on a SpaceX rocket aboard a TakeMe2Space spacecraft, with NASA and Stellerian listed as customers.
Why Houston, Not Silicon Valley
Non Sibi Ventures is a smaller, less prominent fund than the usual deep-tech names backing orbital infrastructure, and Harris's astronaut background looks like a deliberate signal -- a credibility stand-in for a round this early and this small relative to the capital other space-AI plays have raised this year. Satlyt's own ambitions aren't modest: Afullo's stated long-term goal is to be running on 20% of all satellites by the end of the decade, with a shared compute cloud spanning two satellites as the next concrete milestone, expected next year.
The competitive risk is straightforward: SpaceX, with its own launch vehicles and Starlink constellation, doesn't need Satlyt's multi-vendor pitch and can simply build a closed version of the same idea with its existing fleet, a structural advantage no amount of software cleverness fully offsets. For VCs scanning space-tech deal flow, Satlyt is a bet that enough non-SpaceX satellite operators -- government and commercial -- will pay for a shared compute layer before SpaceX or Google's own vertically integrated version locks up the market first.