Sakura Internet and TEPCO Discuss AI Power Demand

🔑 Enhanced Key Takeaways

•The 'Watt-Bit Collaboration' is a government-convened advisory body in Japan that aims to strategically redistribute compute capacity from urban centers to regions with abundant renewable energy and available grid capacity, such as Hokkaido, Tohoku, Chubu, and Kyushu.
•Sakura Internet has significantly invested in high-performance computing, launching its 'Koukaryoku' generative AI cloud services in January 2024, which utilize NVIDIA H100 and H200 GPUs to address the escalating demand for AI workloads.
•TEPCO Power Grid (TEPCO PG) is undertaking substantial grid reinforcement, including a ¥200 billion investment in northwestern Chiba prefecture and an additional ¥220 billion for distribution network enhancements, alongside plans to develop data centers adjacent to the Kashiwazaki-Kariwa nuclear power plant to secure stable power for AI.
•Japan's government initiated a program in May 2026 to fund the development and testing of low-carbon technologies for data centers, offering up to JPY500 million (US$3.2 million) in subsidies for projects spanning fiscal years 2026 through 2029.
•The global electricity demand from data centers is projected to more than double by 2030, reaching approximately 945 terawatt-hours, a figure exceeding Japan's current total annual electricity consumption, underscoring the critical need for energy optimization.

🛠️ Technical Deep Dive

'Watt-Bit' integration focuses on optimizing electricity supply and demand by dynamically coordinating data processing (bits) with energy availability (watts), potentially through workload shifting between geographically dispersed data centers.
TEPCO and Hitachi have demonstrated successful grid-linked energy management, enabling the shifting of computing workloads across data centers in different regions (e.g., Tokyo and Kanagawa Prefecture) to efficiently utilize local renewable energy sources and provide grid stability.
Sakura Internet's Ishikari Data Center employs direct liquid cooling (DLC) systems, which are designed to reduce cooling power consumption by directly targeting high-heat server components with cold water, thereby also increasing GPU server density from two to up to five units per rack.
Modern AI-optimized servers, particularly those equipped with high-end GPUs like NVIDIA H100/H200, exhibit significantly higher power consumption (700-1,200 watts per chip) compared to traditional CPUs (150-200 watts), necessitating advanced cooling and power delivery solutions.
Japan's strategy for urban data centers involves developing smaller, high-density facilities that incorporate advanced power architectures such as ±400V DC distribution, parallel UPS systems, silicon carbide (SiC) power electronics, and solid-state transformers (SSTs) to enhance efficiency and optimize space.
The University of Tokyo and Fujitsu are conducting trials as part of the Watt-Bit Collaboration Project to verify inter-regional workload shifting between data centers, aiming to balance electricity supply and demand.

🔮 Future ImplicationsAI analysis grounded in cited sources

Japan will accelerate its renewable energy deployment and grid modernization.

The escalating power demands of AI data centers, coupled with the government-backed 'Watt-Bit Collaboration' framework, provide a strong impetus for significant investments in new clean energy capacity and advanced grid infrastructure to meet decarbonization targets and ensure stable power supply.

Data center architecture in Japan will increasingly favor distributed, energy-aware designs.

Constraints such as high urban land costs, limited grid capacity, and the imperative to integrate with renewable energy sources are driving a strategic shift towards smaller, high-density urban facilities and the regional distribution of compute capacity, supported by technologies for grid-aware load shifting.

Nuclear power will play a more significant role in Japan's AI data center energy strategy.

TEPCO's initiatives to develop data centers near the Kashiwazaki-Kariwa nuclear power plant and the broader industry discussion about nuclear energy as a stable, carbon-neutral power source suggest a growing reliance on nuclear power to satisfy the continuous and massive energy requirements of AI.

⏳ Timeline

2024-01

Sakura Internet launches 'Koukaryoku PHY' GPU cloud services for AI/ML workloads.

2025-06-05

Sakura Internet and JERA sign agreement to explore data centers at power stations, promoting 'watt-bit collaboration'.

2025-06-11

Sakura Internet begins operating a container-type data center at Ishikari, equipped with NVIDIA H200 GPUs.

2025-10

The University of Tokyo announces the Watt-Bit Collaboration Project for Green Transformation with TEPCO Power Grid.

2026-05-25

Sakura Internet CEO Kunihiro Tanaka indicates significant capital spending increase (¥20-30 billion) to meet AI demand.

2026-05-27

TEPCO Power Grid establishes TEPCO Digital Infrastructure Co., Ltd. to advance its digital infrastructure business.

Sakura Internet and TEPCO Discuss AI Power Demand

⚡ 30-Second TL;DR

🧠 Deep Insight

🔑 Enhanced Key Takeaways

🛠️ Technical Deep Dive

🔮 Future ImplicationsAI analysis grounded in cited sources

⏳ Timeline

📎 Sources (19)

👉Related Updates

GMO AI & Robotics Becomes Official Unitree Distributor

Amazon employees file complaint over data center testimony probe

OpenAI Codex introduces Record & Replay for AI task automation

Gartner: Privacy Law Enforcement Fines Set to Triple