Why 90% of Satellite Compute Is Wasted, and How We Fixed It On-Orbit

Modern satellites generate vast amounts of data but rely on constrained downlink windows and expensive ground infrastructure to process it. At the same time, more than 90% of onboard compute capacity often sits idle. This imbalance is becoming one of the defining bottlenecks of the space economy.

This session presents a new approach: treating satellites as distributed compute nodes and processing data directly in orbit. Drawing from recent on-orbit deployments, we will walk through how lightweight, containerized workloads can be uplinked to satellites, enabling real-time data processing, compression, and AI inference before transmission to Earth.

We will share mission-backed results, including reductions in telemetry size by over 90%, deployment of AI models in constrained environments, and lessons learned from running workloads across heterogeneous spacecraft. The session will also explore how intersatellite networking and delay-tolerant architectures enable coordination across fleets, unlocking new capabilities for operators and mission planners.

Attendees will gain a practical understanding of how to reduce downlink costs, improve mission responsiveness, and unlock unused onboard resources without requiring new hardware.

This session is designed for satellite operators, government program managers, and commercial partners looking to implement near-term solutions that improve mission efficiency and scalability.