日本語 
Ubita has revealed plans to build a large scale AI data center together with Maizuru City. The project is being positioned as a concrete step toward building next generation AI infrastructure inside Japan. Ubita’s group company, UBTECH, said it will hold a land signing ceremony and press conference with Maizuru City on January 29, 2026. That event is expected to formally mark the start of physical construction. The facility is planned as a major AI GPU center within Japan.
This land agreement follows UBTECH being selected for a large scale investment subsidy from Japan’s Ministry of Economy, Trade and Industry. With that support in place, the company has committed around 17 billion yen to the project. The move highlights growing coordination between the company and both central and local governments as Japan works to strengthen its domestic AI infrastructure base.
The AI data center will be located in Maizuru City on a site of about 2.3 hectares, roughly 23,000 square meters. The location was chosen for practical reasons. It offers stable land conditions, existing infrastructure, and access to ports and logistics routes. These points matter for operating AI systems that require high density computing and constant reliability.
Also Read: iGrafx and NTT DATA INTRAMART Drive Autonomous Ops
Construction is planned in stages. The first phase will focus on building core facilities and establishing basic operations. The second phase will expand computing capacity as demand increases, especially from generative AI and large language models. This step by step approach is intended to support long term growth while keeping operations stable. Construction is expected to begin in 2026, with major building work finishing toward the end of 2027.
From a technology standpoint, the facility will adopt NVIDIA’s Blackwell GPU architecture. These GPUs are designed for large scale AI training and inference and support dense deployment over long periods of operation. This puts the Maizuru site among a limited number of locations in Japan capable of handling next generation AI workloads.
The system design will be based on Ubita’s NeoCloud architecture. NeoCloud uses distributed computing, layered scheduling, and flexible resource control. GPU resources can be allocated dynamically across regions, industries, and projects. This is intended to improve efficiency, reduce latency, and avoid the bottlenecks seen in more centralized computing setups.
Maizuru City was selected after considering power supply stability, disaster risk, logistics strength, and cooperation from local institutions. These factors together make it suitable for long term and scalable AI operations.
The project fits into Ubita’s broader AI roadmap in Japan. In 2024, the company was selected for METI’s GENIAC program and developed a large scale East Asian language model supporting Japanese, Traditional Chinese, and English. In 2025, it rolled out AI solutions focused on tourism and culture, including the Chokimaru virtual tour guide already used in Maizuru. By 2026, Ubita plans to expand NeoCloud nationwide, building a flexible GPU infrastructure that can scale across regions and industries.
English