Nautilus Technologies, Internet Initiative Japan, TEPCO Power Grid, CHUBU Electric Power Grid, Fujitsu, and 1Finity have launched a collaborative proof-of-concept experiment in Japan to develop a next-generation distributed AI processing infrastructure. This initiative is part of the New Energy and Industrial Technology Development Organization's efforts to validate low-latency distributed processing technologies for next-generation infrastructure.
The project aims to achieve the Watt-Bit Convergence Vision, a strategic framework that integrates electricity grids and telecommunications to optimize energy usage and computing power. The experiment will employ an All-Photonics Network, utilizing optical fiber cables from TEPCO and CHUBU and communications equipment from Fujitsu and 1Finity, to enable low-latency, high-bandwidth data transmission.
The proof-of-concept will be tested during the final race of the KYOJO CUP formula racing series at Fuji Speedway. Real-time 4K video and telemetry data from race cars will be transmitted via the All-Photonics Network to GPU servers distributed across TEPCO and CHUBU service areas for AI processing. This setup allows data to be processed by any GPU server within the network, facilitating flexible adjustments in response to electricity demand.
This experiment marks Japan's first attempt to operate geographically distributed GPU servers as a unified AI processing environment across regions with different utility frequencies. The project aims to demonstrate the feasibility of using low-latency distributed processing technologies to balance electricity supply and demand by shifting AI workloads to areas with surplus power capacity.
The proof-of-concept period will run from August 2026 to March 2027, starting from Round 3 of the KYOJO CUP series. Results and system verification details are expected to be presented at industry events, including Inter BEE, from autumn 2026 onward.