Space Edge Onboard AI Computer for Satellites

Spiral Blue builds computers, sensors and software for space, operating from Sydney with teams in Santa Ana, California and Harwell, UK. It addresses the bottleneck of downlinking large volumes of raw satellite data by processing it in orbit, so operators send down answers rather than files.

Its Space Edge computer is a 0.25U module of about 100 g. The latest SE-2 uses an NVIDIA Jetson Orin NX 16 GB with up to 157 TOPS, 250 GB to 2 TB of storage and 25 W maximum power, while the flight-proven SE-1 uses a Jetson Xavier NX. It runs the same NVIDIA software stack used on the ground, with ready apps for vessel detection, cloud clipping, canopy mapping, fire severity and water body mapping. Ten Space Edge units have flown since 2021, including on SatRev and Satellogic satellites. The company also develops Sapphire space-to-space and Teal Earth observation LiDAR.

• SE-2 computer: NVIDIA Jetson Orin NX 16 GB with up to 157 TOPS and 25 W maximum power.
• SE-1 computer: flight-proven Jetson Xavier NX unit at 250 g.
• Onboard apps: Vessel Detect, Cloud Clipper, Canopy Mapper, Fire Severity and Water Body Mapper.
• Analytics: ship and fire detection, land use, NDVI and EVI, segmentation and object detection.
• Support package: flight model with 40 hours of engineering support and optional shielding and thermal parts.
• LiDAR products: Sapphire space-to-space and Teal Earth observation LiDAR in development.

• Vietnamese satellite builders adding Space Edge to a micro-satellite to detect vessels over the East Sea in orbit.
• Forestry and disaster agencies receiving fire severity or water body maps directly instead of raw images.
• Research teams testing their own AI models in orbit on an engineering-model-compatible computer.

• Sends results instead of raw imagery, reducing downlink volume and time to insight.
• Up to 157 TOPS of AI compute in a 0.25U, roughly 100 g module.
• Same NVIDIA software stack as on the ground, so models move from prototype to orbit quickly.
• SE-2 lead time of 12 to 16 weeks, with engineering models for ground testing.

Satellite data volumes are growing faster than ground station capacity, pushing operators towards onboard AI that delivers alerts in minutes. In Southeast Asia, fast detection of illegal fishing, forest fires and floods is a priority, and edge processing in orbit shortens the time from capture to action.

Operates from Sydney (Australia), Santa Ana (US) and Harwell (UK); led by founder and CEO Taofiq Huq and co-founder MJ Akter. Won an AUD 3 million CRC-P grant in 2025 for an AI-enabled LiDAR satellite demonstration and is raising equity crowdfunding on Birchal.