Nvidia is working with several companies in Japan to develop artificial intelligence (AI) infrastructure across the country.
The companies taking part include SoftBank Corp., GMO Internet Group, Highreso, KDDI, Rutilea, and Sakura Internet, and will be using Nvidia's computing, networking, and software to bring AI solutions to Japan's robotics, automotive, healthcare, and telecom industries.
Announced at the Nvidia AI Summit in Japan, the efforts are being supported by Japan's Ministry of Economy, Trade, and Industry.
GMO Internet Group is launching a GPU cloud offering in Japan that will feature full-stack Nvidia H200 GPUs, the Nvidia Spectrum-X Ethernet platform for AI, Nvidia BlueField-3 DPUs, and the Nvidia AI Enterprise software suite.
The GMO GPU Cloud will be built on Dell PowerEdge servers and is expected to go live later this month.
Highreso is developing an AI data center - dubbed Highreso Kagawa - which will feature Nvidia H200 GPUs and will provide high-performance infrastructure for Highreso’s GPUSOROBAN AI Supercomputer Cloud service.
The company is also developing a second AI data center that will go online in the summer of 2025. In total, the two data centers will have 1,600 Nvidia GPUs.
KDDI is launching AI computing infrastructure built with the Nvidia HGX systems to support generative AI and specialized large language model (LLM) development. The company is further planning a liquid-cooled data center that will feature the Nvidia GB200 NVL72 platform with Nvidia GB200 Grace Blackwell Superchips.
Rutilia is doubling its Nvidia Hopper computing capacity in its AI cloud data center and will provide more than 1,000 Nvidia Hopper GPUs for LLM development.
Sakura Internet, meanwhile, is planning to expand its Koukaryoku cloud services for generative AI from 2,000 to nearly 4,000 Nvidia Hopper GPUs. It also plans to install Nvidia HGX B200 infrastructure with Blackwell GPUs at its Ishikari data center, which is expected to be fully powered by renewable energy by 2027. In total, Sakura intends to offer around 10,800 GPUs as part of its AI cloud offering.
As already reported by DCD, SoftBank is set to receive the world’s first Nvidia DGX B200 systems, which will be used to build the Nvidia DGX SuperPod supercomputer. The system will also feature Nvidia AI Enterprise software and the company’s Quantum-2 InfiniBand networking. The company is also developing a supercomputer featuring Nvidia's GB200 NVL72 system.
“Japan’s companies stand to benefit tremendously from the new industrial revolution powered by AI,” said Jensen Huang, founder and CEO of Nvidia.
“Company employees will supercharge their speed and productivity by automating work with AI agents. Industrial companies of tomorrow will operate dual factories — new AI factories to produce software intelligence for the products and machines their factories make today. Working with Nvidia, Japan’s cloud providers are building the AI factories essential to reinvent the country’s automotive, robotics, telecommunications, and healthcare industries for the age of AI.”
