AI software reaches orbit as Satlyt raises $8 million to cut space data transmission costs

Satlyt raised eight million dollars in a seed funding round led by Noon SEVP Ventures, based in Houston, to develop a software system that enables running AI models and processing data directly on board satellites in space. The company, founded by former Google cloud engineer and Starlink network veteran Rama Afolo, with offices in Sunnyvale, California, and Nairobi, is betting that orbital computing is no longer a postponed luxury but a technical necessity to reduce the burdensome operational costs that companies and satellite operators incur to transmit raw data to ground stations.

Satellites traditionally rely on ground control and monitoring centers to resolve software faults or analyze sensor readings, a process that is slow and extremely costly due to the limited downlink bandwidth and high prices.Satlyt does not aim to build its own rockets or orbital hardware, but rather builds an intermediate software layer similar to a VMware model or an Android system.It provides an open, distributed environment that enables running intelligent models across different satellite fleets without needing to redesign their core systems from scratch.

The company has demonstrated the viability of its approach in previous experimental missions, running the open-source Gama model developed by Google DeepMind on a vehicle operated by Momentus. The model succeeded in reducing the size of data packets sent to Earth for software-error diagnosis by more than 60 percent, translating into financial savings of several hundred thousand dollars per spacecraft each year. The platform is preparing to launch again on a SpaceX rocket to run its software on a satellite built by Indian firm TecMeToSpace, alongside testing a prototype of Google’s space-data-center project known as Project Sun Catcher.

The current mission involves three parties paying for the service: the U.S. space agency NASA, which is testing cloud-computing protocols in the space environment; the space-monitoring company Stellarian, which specializes in tracking orbital objects to field-test image processing; and the Indian hardware manufacturer. The company plans a more ambitious step next year, building a shared, cloud-based computing network spanning two separate satellites, paving the way to convert third-party satellites into managed orbital computing resources that generate revenue for their owners. Investors, including former NASA astronaut Bernard Harris, have affirmed that the success of this model does not necessarily depend on realizing the complex visions of giant space data centers, but simply rests on the accelerating pace of commercial satellite launches and their need for direct operational solutions.

This shift in space-computing architecture has direct implications for satellite software and ground-monitoring projects in the Gulf region and Egypt, where multiple entities are expanding satellite constellations to observe the marine environment, track navigation movements, and monitor urban expansion and desertification.Relying on local orbital inference fundamentally changes the operational budget equation.It allows engineering teams to stop paying high downlink fees to download thousands of unnecessary images and data, replacing that with edge-filtered data and sending only immediate alerts. It also requires regional technology project managers to focus their expertise on engineering lightweight models tailored for edge-space computing, and to mandate the inclusion of graphics processors in the specifications of new satellites rather than relying solely on limited-capacity traditional computers.

Satlyt’s leadership expects advanced processors to become a standard component in all satellites before the end of the current decade, aiming to run its software on a fifth of the satellites in orbit. This transition redefines economic value in the space sector: power is no longer confined to owning launch platforms or mechanical structures, but gradually shifts to the smart software layer that manages hardware in orbit and determines how it is exploited for commercial efficiency.