AIST, LSTC develop 2nm edge AI chip combining RISC-V CPU and accelerator
Japan's national semiconductor research consortium is building a 2nm-generation edge AI accelerator integrating RISC-V CPUs and AI hardware, targeting robots and vehicles. Project led by AIST and LSTC, with Rapidus manufacturing and Tenstorrent CPU IP.
TOKYO — Japan’s National Institute of Advanced Industrial Science and Technology (AIST) and the Leading-edge Semiconductor Technology Center (LSTC) are developing a 2nm-generation edge AI accelerator integrating RISC-V CPUs and AI hardware to enable generative AI inference on robots, vehicles, and industrial equipment without cloud reliance.
Dr. Shinichi Ouchi of AIST will present the project November 11 at RISC-V Day Tokyo 2026 Autumn, held at the University of Tokyo. The project, funded by NEDO and selected in February 2024, combines 2nm process technology with specialized AI inference architecture to deliver high performance under the strict power and thermal constraints of edge devices. Development involves AIST, University of Tokyo’s AI Chip Design Open Innovation Laboratory (AIDL), Rapidus (manufacturing), and Tenstorrent (RISC-V CPU IP).
The project targets industrial robots, automobiles, and assistive devices that process sensor and camera data locally. Public filings show Rapidus will provide 2nm process design kits (PDKs) for Tenstorrent’s CPU design and chip fabrication.
Japan’s 2nm ambitions hinge on proving domestic fabs can support advanced AI workloads. This is a national project with no Kansai participation; LSTC and Rapidus operate primarily in Hokkaido and Tokyo.