The Vision of an Orbital Cloud
As the race to place advanced hardware in space intensifies, startup Satlyt is aiming to solve the software fragmentation currently hindering the industry. Founded by Rama Afullo, an engineer with significant experience at both Google and SpaceX, the company has just announced an $8 million seed round to build a universal software platform for satellites. Unlike major players attempting to build proprietary space data centers, Satlyt operates with an agnostic philosophy—comparable to how VMware or Snowflake provide abstraction layers for terrestrial cloud infrastructure.
Afullo’s goal is to turn satellite fleets into a cohesive, interoperable computing environment. By developing a horizontally integrated ecosystem, the startup aims to provide an 'Android-like' OS for space, allowing various manufacturers to run AI workloads across different hardware platforms without redesigning their systems for every specific satellite architecture. This approach circumvents the need for massive, single-purpose orbital data centers, focusing instead on optimizing the utility of every satellite already headed to orbit.
Operational Efficiency Through Edge AI
The core value proposition for Satlyt lies in reducing the heavy reliance on ground-based flight control centers. Currently, diagnosing anomalies and processing raw sensor data requires slow, costly, and high-bandwidth downlinks to Earth. By deploying AI models directly onto the spacecraft, Satlyt enables real-time decision-making and autonomous troubleshooting. In a previous demonstration, the company successfully utilized Google’s Gemma AI model on a Momentus spacecraft to handle onboard software errors, resulting in a 60% reduction in transmission size and significant potential cost savings per mission.
This week, Satlyt is hitting a major milestone with a new launch featuring hardware from the Indian startup TakeMe2Space. The mission serves as a critical test for cloud computing in space, with participants including NASA and the surveillance-focused startup Stellerian. The tests aim to demonstrate how diverse companies can host and manage their software workloads on shared orbital infrastructure, effectively turning a simple satellite into a revenue-generating, managed service.
Why it Matters
- Decentralization: Satlyt moves away from the 'big data center' model in favor of a distributed cloud that utilizes existing satellite constellations.
- Cost Reduction: Autonomous, AI-driven troubleshooting reduces the need for expensive data downlinks and human-intensive ground operations.
- Standardization: By providing a software layer compatible with various hardware vendors, Satlyt is positioning itself as the primary interface for future space-based AI.
The Path to a Global Compute Fabric
Looking ahead, Satlyt has set its sights on an even more ambitious goal: linking multiple satellites into a unified orbital cluster. The company plans to attempt a shared computing architecture next year that would allow two different spacecraft to operate as a single cloud node. This capability would effectively establish a decentralized compute fabric, allowing space-based customers to request processing power on demand, regardless of which hardware provider owns the satellite.
Backed by investors including Non Sibi Ventures—which counts former NASA astronaut Bernard Harris among its partners—Satlyt is betting that the proliferation of GPUs in space is inevitable. With the ultimate target of being installed on 20% of all satellites by the end of the decade, the company is positioning itself as an essential utility layer for the next generation of space exploration and commercial activity.
