Japan Data Center Update 39: Japan Explores Workload Shifting as a Source of Energy System Flexibility

Workload shifting (WLS) could allow data centers to adjust when and where they consume electricity, reducing demand in regions experiencing grid constraints or high electricity prices and shifting workloads to locations or periods with abundant renewable generation. This could help ease pressure on the grid, reduce electricity costs and improve renewable energy utilization.

Three projects selected under the MIC’s program (WLS category) will test different approaches to allocating computing workloads across geographically distributed data centers based on electricity supply and demand, with some also considering electricity market prices and graphics processing unit (GPU) utilization. All three trials are scheduled to run through February 2027:

Five companies, including J-Power and Hitachi, launched a WLS trial in August connecting AI data centers in Yokohama, in the Tokyo area, and Kitakyushu, in the Kyushu area. The trial shifts computing tasks across locations or time periods, taking into account forecasts of renewable generation and electricity trading prices, regional power supply and demand conditions, and GPU utilization. The companies will also explore the feasibility of an All-Photonics Network (APN), a next-generation high-speed communications network.

Meanwhile, a 12-member consortium, including TEPCO Power Grid and TEPCO Holdings, will test a new AI infrastructure model called a “virtual hyperscaler.” The trial will connect facilities primarily through TEPCO Power Grid’s fiber-optic network and SINET, the academic network linking universities and research institutions nationwide to optimize where computing workloads are processed.

Container-based AI data center developed by Morgenrot and MIRAIT ONE in Kyushu, Japan. Source: Morgenrot, August 2025.

Similarly, a five-company consortium including NTT DOCOMO BUSINESS, Chugoku Electric Power, and others has also launched a trial. The project uses the IOWN (Innovative Optical and Wireless Network)-based docomo business APN Plus network and distributed control technologies to connect multiple data centers and test distributed AI processing and WLS. It will also evaluate the trade-offs between the economic and environmental benefits of WLS and AI service performance.

–