Rama Afullo had a vision that most of his former employers weren't ready to hear. The Kenyan American engineer, who built his career inside some of the most influential technology companies in the world, believed that satellites could do far more than relay signals - they could become genuine computing platforms capable of running artificial intelligence workloads in orbit. Neither Google nor SpaceX wanted to pursue the idea internally when he raised it. So he went ahead and built it himself.

Afullo spent time working within Google's cloud computing division before moving briefly to SpaceX's Starlink operation in 2024. Both experiences sharpened his conviction that there was a meaningful gap in the market - one that existing players were either ignoring or not yet ready to address. After leaving SpaceX, he co-founded Satlyt, a software company with offices in Sunnyvale, California, and Nairobi, Kenya, focused on enabling satellites to share and execute large-scale computing tasks in space.

The company has now raised an $8 million seed round, a milestone that arrives at a moment when interest in orbital computing has shifted from fringe concept to active investment category.

A Software-First Approach to Space Computing

What distinguishes Satlyt from many of the other players now entering the orbital computing space is that the company has no plans to manufacture its own spacecraft. While companies such as SpaceX, Google, Starcloud, and Cowboy Space Company are investing in building physical hardware for space-based data centers, Afullo's team is concentrating entirely on the software layer - the tools that allow AI models to run directly on satellites already in orbit or being prepared for launch.

Afullo draws a direct comparison to enterprise software companies like VMware and Snowflake, whose platforms allow organizations to run complex software workloads without needing to manage or even fully understand the underlying infrastructure. His ambition is for Satlyt to occupy a similar position in the satellite industry - a horizontal software layer that works across spacecraft built by many different manufacturers.

"The folks like SpaceX who are doing orbital data centers - if they are the iPhone, we'll build Android as a horizontally integrated, open ecosystem."

The analogy is deliberate. Rather than locking customers into a single hardware ecosystem, Satlyt is positioning itself as a cross-platform solution that any satellite operator could adopt, regardless of who built their spacecraft or what processors are onboard.

Early Demonstrations Already in Orbit

Satlyt is not operating purely in the theoretical. The company has already flown its software on two demonstration missions, using those flights to validate its core premise: that AI running directly on a satellite can reduce the cost and complexity of managing spacecraft operations.

Traditionally, satellites rely on ground-based flight controllers to identify and resolve problems. Any anomaly detected onboard must be transmitted back to Earth for analysis - a process known as downlink that is both expensive and time-consuming. By deploying AI models directly on the spacecraft, decisions can be made autonomously without waiting for a round-trip communication cycle.

Earlier this year, Satlyt deployed Google DeepMind's Gemma AI model aboard a spacecraft operated by Momentus. The results were notable: the model reduced the size of onboard software error transmissions by more than 60 percent. Afullo estimates that this type of efficiency gain can translate into savings of hundreds of thousands of dollars per satellite annually - a figure that, across a large constellation, becomes a compelling financial argument for adopting the platform.

This Week's Launch and What It Involves

The company's next milestone is a launch taking place this week aboard a SpaceX rocket - the same launch carrying the first prototype for Google's own space data center initiative, known as Project Suncatcher. Satlyt's software will be operating onboard a spacecraft built by TakeMe2Space, an Indian startup that specializes in computing hardware designed for satellite environments.

The mission is notable for the range of customers it serves simultaneously. Three separate organizations are using the flight to test different capabilities:

• NASA is funding tests of protocols for cloud computing operations conducted in space.
• Stellerian, a startup working on space domain awareness - monitoring and tracking objects in orbit - is evaluating how image processing workloads perform when handled onboard rather than on the ground.
• TakeMe2Space itself is using the mission to demonstrate that its satellite hardware can host third-party software from multiple customers simultaneously.

The multi-tenant nature of the mission reflects exactly what Satlyt is trying to enable more broadly: satellites that function less like dedicated single-purpose hardware and more like shared computing infrastructure available to multiple users and applications.

The Road Toward a True Orbital Cloud

For now, the number of high-powered graphics processing units in orbit remains limited, which means the most commercially viable workloads are the kinds of operational efficiency tasks Satlyt has already demonstrated. However, the company's roadmap extends well beyond single-satellite deployments.

The next major technical objective is to establish a shared computing environment - effectively a cloud - that spans two separate satellites simultaneously. Satlyt expects to attempt this multi-satellite demonstration next year. If it succeeds, the company would be positioned to offer genuine cloud computing services in orbit, functioning as a third-party provider to satellite builders who want to monetize spare computing capacity or offer hosted services to other organizations.

"The way to think about it is that we turn your satellite into a revenue-generating managed service," Afullo explained.

The longer-term ambition is significant in scale. Afullo has set a target of having Satlyt's software running on 20 percent of all satellites by the end of the decade. His expectation is that by that point, virtually every new spacecraft entering orbit will be equipped with GPUs or comparable advanced processors - making the absence of such hardware a competitive disadvantage for any satellite operator.

Backing From an Astronaut-Connected Firm

The $8 million seed round was led by Houston-based Non Sibi Ventures. The firm's involvement carries particular weight given the background of one of its partners: Bernard Harris, who spent more than two decades as a NASA astronaut and helped evaluate Satlyt's potential from a perspective grounded in practical spaceflight experience.

Non Sibi partner Kent Lucas described the investment rationale in terms that reflect a measured view of the orbital computing market's development. The firm does not see Satlyt's success as contingent on the most ambitious visions of large-scale space data centers materializing on any particular timeline. Those scenarios face real challenges, including uncertain economics and limited launch capacity.

"We don't need data centers in space for Rama to be wildly successful. It can just be driven by the number of satellites going up."

The argument is straightforward: as the global satellite count continues to grow - driven by commercial broadband constellations, Earth observation programs, defense applications, and scientific missions - the demand for onboard computing and software management tools will grow alongside it. Satlyt's business model is designed to scale with that expansion rather than depend on a single transformative infrastructure shift.

For Afullo, the pitch he couldn't get approved at Google or SpaceX has now become a funded company with hardware in orbit and a launch on the manifest. The question going forward is whether the broader satellite industry will adopt software-defined computing at the pace he is betting on.