NVIDIA and Japan Launch National AI Infrastructure
NVIDIA has partnered with Japanese political and economic stakeholders to launch the world's first national-level AI infrastructure. The project includes the development of a significant AI factory.
NVIDIA announced today its collaboration with Japanese political and economic sectors to launch the world's first national-level artificial intelligence infrastructure. This initiative marks a significant step in building domestic AI capabilities.
The core of the project involves the creation of a 140MW national physical AI factory. This facility will utilize NVIDIA's Vera Rubin NVL72 racks, housing 13,750 Vera CPUs and 27,500 Rubin GPUs, interconnected with Spectrum-X Ethernet. The infrastructure is designed to support training of T-level models.
This AI factory's data center will be dedicated to developing open-weight foundational models for physical AI. It aims to provide the necessary computing power for Japan's FRONTia project, which focuses on multimodal foundational model development for AI robotics and physical AI.
In conjunction with the launch, Noetra's multimodal foundational models will be made available to Japanese developers and businesses, alongside NVIDIA's open-source models such as Nemotron, Cosmos, and Isaac GR00T, and the NeMo library. This aims to accelerate the development of autonomous and physical AI applications within Japan.