Preferred Networks, Inc. (PFN) and Tenstorrent have been selected by Japan’s New Energy and Industrial Technology Development Organization (NEDO) for a five-year project to develop high-performance, energy-efficient AI processors for training and inference. The project theme is “xPU Design and Development Incorporating Photonics-Electronics Integration Technology,” under NEDO’s research and development project for post-5G information and communication systems and advanced semiconductor manufacturing technologies.
The companies said AI models and workloads are increasing in size and computational complexity and are being adopted across a wide range of industries. This growth has increased the compute needed for model training and inference, driving demand for more scalable and efficient AI infrastructure.
PFN and Tenstorrent will research system-level integration of AI processors using photonics-electronics integration and advanced packaging technologies. PFN will develop AI chiplets with ultra-high memory bandwidth enabled by 3D memory stacking technologies based on leading-edge semiconductor process technologies, while Tenstorrent will develop open-architecture CPU chiplets that support confidential computing.
The project will also involve subcontractors researching domain-specific applications for scientific AI and physical AI: The University of Tokyo, Kobe University, Toyo University, Japan Agency for Marine-Earth Science and Technology, and Japan Atomic Energy Agency. Tenstorrent Japan will design CPU chiplets and develop software to enable confidential computing.
The companies will conduct the research and development program for five years and then work to commercialize its outcomes for applications including AI agents.
