The Cloud in Orbit: How Satlyt is Building the ‘Android’ of Space Computing

In the rapidly expanding frontier of the New Space economy, a quiet revolution is taking place—not in the metal hulls of massive rockets or the solar arrays of mega-constellations, but in the invisible, ethereal layers of software. Rama Afullo, a Kenyan American engineer with deep roots in both Google’s cloud division and SpaceX’s Starlink, believes that the future of space exploration isn’t just about putting more satellites in orbit; it’s about making them smarter.

Afullo is the co-founder of Satlyt, a startup based in Sunnyvale, California, and Nairobi, which is aiming to solve a fundamental bottleneck in the orbital economy: the inability of satellites to process data autonomously. By providing an open, hardware-agnostic operating layer for orbital computing, Satlyt is positioning itself to be the "Android" of space, allowing diverse companies to run complex AI models in the vacuum of space without the traditional reliance on ground-based processing.

The Genesis of an Orbital Vision

The path to Satlyt was paved with frustration. During his tenure at both Google and SpaceX, Afullo frequently pitched the idea of orbital data centers—a concept that would move compute-intensive workloads from Earth-bound servers to the edge of space.

"When I was at SpaceX, I tried to pitch this internally. They said no. When I was at Google, I tried to pitch this internally. They said no," Afullo told TechCrunch.

Despite the initial institutional rejection, the industry eventually caught up to his vision. SpaceX has since pivoted to prioritize orbital data processing, and the broader aerospace industry is finally acknowledging that the "download-process-uplink" model—where satellites act as simple "bent pipes" relaying data to Earth—is fundamentally unsustainable as the volume of satellite data explodes.

With $8 million in fresh seed funding led by Houston-based Non Sibi Ventures, Satlyt is now moving from the whiteboard to the stratosphere. The company’s timing is impeccable: this week, Satlyt software will launch on a SpaceX rocket, sharing the flight with the first prototype of Google’s own "Project Suncatcher" space data center initiative.

A New Paradigm: Software-Defined Satellites

To understand Satlyt’s impact, one must understand the current limitations of satellite infrastructure. Today, satellites are largely bespoke, vertically integrated systems. If a company wants to run a specific AI model for earth observation, they must write code specifically for that satellite’s proprietary hardware.

Satlyt is breaking this mold. Much like VMware or Snowflake in the terrestrial cloud-computing market, Satlyt provides an abstraction layer. It decouples the software applications from the underlying hardware.

"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," Afullo explained.

This approach allows Satlyt to offer a universal platform. It doesn’t matter if the satellite is built by a startup in India or a legacy aerospace giant; if it uses Satlyt’s software, it can run the same modular AI workloads. This is a game-changer for mission efficiency. Currently, space missions are plagued by latency. When a satellite encounters an anomaly, it must beam telemetry to ground control, wait for a human-in-the-loop to troubleshoot, and then send a command back. By using AI to process sensor data locally, Satlyt can resolve anomalies in milliseconds, saving massive amounts of operational expenditure.

Chronology: From Concept to Orbit

The trajectory of Satlyt has been marked by rapid iteration:

• 2024: Afullo departs SpaceX, having recognized that the industry was approaching a "compute in space" inflection point. Satlyt is formally founded.
• Early 2024: The company successfully deploys Google DeepMind’s Gemma AI model on a satellite operated by Momentus. The results were stark: the model reduced the size of error-reporting transmissions by over 60%.
• Mid-2024: Satlyt secures $8 million in seed funding, backed by Non Sibi Ventures.
• Late 2024 (Current Week): Satlyt launches its software on a mission with TakeMe2Space, an Indian hardware provider. The mission includes three distinct stakeholders: NASA (testing cloud protocols), Stellerian (testing space-based image processing), and TakeMe2Space (demonstrating third-party software hosting).
• 2025 (Projected): Satlyt plans its first "shared compute" demonstration, where two different satellites will act as a unified, distributed cloud system.

The Economic Implications of "Space-as-a-Service"

The financial logic behind Satlyt is grounded in reality rather than hype. Bernard Harris, a former NASA astronaut and current partner at Non Sibi Ventures, noted that the firm’s investment was driven by the practical utility of Satlyt’s software.

"We don’t need data centers in space for Rama to be wildly successful," said Kent Lucas, another partner at Non Sibi. "It can just be driven by the number of satellites going up."

The economic argument is simple: satellites are expensive. A satellite that can process its own data is more efficient, more reliable, and ultimately, more profitable. By turning a satellite into a "revenue-generating managed service," Satlyt allows operators to rent out their onboard computing capacity to third parties, much like AWS rents out server time.

For NASA, the interest lies in standardized protocols. If satellites can share workloads, the agency could theoretically create a distributed "constellation compute" network, where one satellite passes a data-heavy task to another that has idle processing power, effectively creating a peer-to-peer cloud in orbit.

The Road Ahead: Overcoming the GPU Shortage

The primary challenge facing Satlyt is the current lack of high-powered hardware in orbit. Space-hardened processors are historically slow, designed for radiation resistance rather than teraflops of performance.

However, this is changing. As industry demand for real-time AI processing grows, chip manufacturers are increasingly producing radiation-tolerant GPUs. Afullo believes we are approaching a tipping point.

"If you’re putting up a satellite without putting up a GPU on it, at the very least, you’re doing yourself a disservice," Afullo stated.

Satlyt’s roadmap is aggressive. By the end of the decade, the company aims to be running on 20% of all orbiting spacecraft. Their next hurdle is the 2025 demonstration of inter-satellite computing. If they can prove that two independent satellites can handshake and share processing loads, they will have effectively built the world’s first distributed orbital data center.

Conclusion: An Ecosystem, Not an Island

As we look toward the end of the decade, the space industry is shifting from a focus on access (launching things) to utility (doing things). In this new era, companies that can provide the "plumbing" for the space-based internet will likely become as foundational as the companies that build the satellites themselves.

Satlyt is not trying to build the next SpaceX. Instead, it is building the infrastructure that will allow the next generation of space companies to thrive. By focusing on software, integration, and efficiency, Afullo’s team is ensuring that when the orbital data centers of the future are finally fully operational, there will already be an "Android-like" ecosystem ready to power them.

For the private space industry, the message is clear: the future of intelligence is decentralized, and for those who embrace the move to orbital computing, the sky is no longer the limit—it’s just the beginning.