NASA Validates Radiation-Resilient SAKURA-II AI Accelerator for Orbital and Lunar Missions

Estimated reading time: 1 minute

EdgeCortix® Inc., a leading fabless semiconductor company specializing in energy-efficient AI processing at the edge, announced that its leading edge AI co-processor, SAKURA-II, demonstrated high levels of radiation resiliency in NASA’s heavy ion testing. This report, prepared by the NASA Electronic Parts and Packaging Program (NEPP), is the second EdgeCortix AI co-processor to be tested. As previously reported in 2024, SAKURA-I demonstrated greater radiation resiliency compared to other similar AI processors tested, and now SAKURA-II impressively delivered equivalent performance metrics.

NASA heavy ion testing conducted at Texas A&M Cyclotron confirmed no destructive events and relatively few transitory radiation effects, proving SAKURA-II’s suitability for use cases in low Earth orbit, geosynchronous orbit, and lunar operations. The results of this testing, which was supported by the Defense Innovation Unit (DIU), highlight the future opportunity to enable autonomy in space with machine-learning and computer vision applications.

“The completion of this heavy ion testing and NASA’s published report represent a significant milestone in EdgeCortix’s mission to extend intelligent computing beyond Earth,” said Dr. Sakyasingha Dasgupta, Founder and CEO of EdgeCortix. “These results validate SAKURA-II’s exceptional radiation resilience in the most extreme environments and demonstrate that advanced AI processing can be performed reliably directly in orbit and on the lunar surface. As space systems increasingly demand greater autonomy, lower power consumption, and real-time decision-making, EdgeCortix is proud to help enable a new generation of energy-efficient, AI-driven space exploration.”

The NEPP testing initiative was launched to advance the pursuit of autonomy in space. While advancements in machine learning and computer vision have greatly improved sensor processing, their high computational demands often surpass the limits of embedded central processing units (CPUs) while graphics processing units (GPUs) often require excessive power.