Concepts
What is orbital AI infrastructure?
Orbital AI infrastructure is the compute, sensors, connectivity and ground stations used to run AI models in or through space: processing data on a satellite before it is sent down, or running workloads on computers in orbit. It is early: AI inference on satellites has been demonstrated, and general-purpose, self-service orbital compute is not yet broadly available.
Auvryn is provider-neutral execution infrastructure for AI models. It validates ONNX models, runs them as durable ExecutionJobs on infrastructure it does not own, and returns verified results with provenance and telemetry.
Why run AI in orbit
Sending raw sensor data to the ground is slow and limited by downlink capacity. Running a model where the data is captured can return a result instead of the raw data, sooner and with less bandwidth.
What makes it hard
- Access depends on orbital passes and ground-station contact windows.
- Results come back through a downlink, which takes its own time and capacity.
- Power, thermal limits and radiation constrain the hardware.
- Each operator exposes its own interfaces, packaging and schedules.
How Auvryn fits
Auvryn is designed for this: its provider contract can describe scheduled execution, orbital compute and ground infrastructure, and the ExecutionJob lifecycle has room for a downlink phase. The goal is that a model which runs on cloud or edge infrastructure today can run in orbit through the same lifecycle when providers open access.
Status today
- No orbital execution has happened through Auvryn yet.
- An orbital compute integration has been investigated and is waiting for provider API access.
- A ground-station integration reads infrastructure information only; it books and executes nothing.
- Real execution today is on external edge hardware, under internal evaluation.
Related
Facts reviewed 2026-10-07.