EdgeCortix Launches RAIDEN Chiplet Platform for Physical AI Applications
EdgeCortix introduced RAIDEN, a scalable AI chiplet platform that scales from a single die to a four-die flagship, delivering up to 3.36 PFLOPS of FP4 compute and 256 GB of memory. The platform targets physical AI systems deployed outside data centers, with design wins already secured in aerospace, defense, robotics, and edge AI servers. All configurations share the same accelerator architecture and software stack, allowing customers to scale performance without changing environments.
The RAIDEN platform is built around a modular chiplet design, with the X4 flagship combining four compute dies that function as one tightly integrated system rather than separate accelerators. Each die incorporates multi-core host processing and high-speed system connectivity, allowing for self-contained configurations. Power settings are adjustable, giving customers flexibility across different performance and deployment scenarios.
EdgeCortix positions RAIDEN for the "thick edge" market, meaning high-performance AI systems that operate outside centralized data centers, closer to sensors and operational equipment. The company has already secured customer design wins in aerospace, defense, robotics, and edge AI servers. Specifications for the smaller X1 and X2 configurations will be released in subsequent phases of the launch program.
The RAIDEN platform could enable more capable AI systems in environments where data center access is impractical, such as autonomous robots, defense platforms, and remote edge infrastructure. By offering a scalable architecture with a consistent software environment, it may lower barriers for organizations deploying physical AI across varied performance tiers. However, the emphasis on aerospace and defense applications raises questions about how these capabilities might be applied, and the power demands of high-performance edge computing could still present challenges in energy-constrained settings.