• Nvidia is expanding Japan’s AI infrastructure with a 140MW facility.
• Japanese groups are using Nvidia for robotics and quantum research.

Nvidia has expanded its work in Japan through new projects covering national AI infrastructure, industrial robotics, edge computing, and quantum research.

The announcements include a 140-megawatt AI factory, new physical AI systems for manufacturers, and Blackwell-based supercomputers for scientific and quantum workloads.

Noetra Corp. plans to build the AI factory using 13,750 Nvidia Vera CPUs and 27,500 Rubin GPUs. The facility will support the government-backed FRONTia Project, which is developing multimodal foundation models for robotics and physical AI.

The AI factory will use Vera Rubin NVL72 racks based on Nvidia’s DSX architecture. It will include Spectrum-X Ethernet networking and BlueField data processing units.

Japan’s Ministry of Economy, Trade and Industry, or METI, launched FRONTia under its wider industrial AI policy. The project brings together industrial data, manufacturing expertise, and computing infrastructure to develop open multimodal foundation models for AI agents, digital twins, robotics, and other physical AI applications.

“Japan has launched the FRONTia Project, which will serve as the core of the country’s physical AI ecosystem,” Ryosei Akazawa, Japan’s Minister of Economy, Trade and Industry, said.

Akazawa said the programme would combine Japan’s industrial expertise and manufacturing infrastructure with technology from international partners.

Noetra plans to make the pretrained weights of its multimodal models available to Japanese developers and enterprises. The infrastructure will also support Nvidia software and models, including Nemotron, Cosmos, Isaac GR00T, and NeMo.

“Bringing physical AI into the real world requires enormous computing, data and foundational technologies — challenges no single company can solve alone,” Hironobu Tamba, CEO of Noetra, said.

Tamba said Noetra would share the results of its research while working with partners in Japan and overseas on the foundation models.

Nvidia said the facility will support the training of trillion-parameter-scale AI models as its capacity expands. The company identified manufacturing, logistics, healthcare, and telecommunications among the sectors covered by the initiative.

Japan’s AI Robotics Strategy, released in March, sets a target of securing more than 30% of the global AI robotics market by 2040. The government estimates that share would represent an opportunity worth about $133 billion.

Manufacturers expand robotics projects

Nvidia also said several Japanese robotics, manufacturing, and technology companies plan to join its Cosmos Coalition. The coalition brings together developers and industrial groups working on world models and physical AI systems.

Prospective members include FANUC, Fujitsu, Hitachi, Honda R&D, Kawasaki Heavy Industries, Kubota, NEC, Preferred Networks, SoftBank, Sony, TIER IV, and Yaskawa Electric.

Nvidia introduced Cosmos 3 Edge alongside the coalition expansion. The four-billion-parameter model is designed to run on edge systems, allowing robots and vision-based applications to interpret surroundings, process information, and generate actions locally.

Cosmos 3 Edge is built on Nvidia Nemotron and can run across RTX GPUs, DGX systems, and Jetson modules, including the T2000 and T3000. Nvidia said developers can adapt the model for specific machines, vehicles, sensors, and operating environments.

“The next frontier of AI is in the physical world, and this is a once-in-a-generation opportunity for Japan,” Jensen Huang, Nvidia’s founder and CEO, said. “Japan invented modern manufacturing. Now, it has the opportunity to reinvent it for the age of intelligent industries.”

Nvidia also introduced Metropolis libraries for developing and operating video intelligence systems. The company said the tools can reduce development time for Cosmos-based vision applications by at least six times.

Fujitsu is exploring a shared physical AI control platform with FANUC, Yaskawa Electric, and Kawasaki Heavy Industries. The planned platform will combine Nvidia Cosmos, Isaac, Omniverse NuRec libraries, and the Newton physics engine.

The companies intend to use the platform for AI model development, digital twins, robot learning, simulation, and pre-deployment validation.

NEC, Hitachi, OMRON, and Preferred Networks are using Cosmos and related Nvidia systems for industrial AI research. SoftBank is developing a physical AI platform based on Cosmos, Omniverse, and Isaac Sim.

SoftBank is also working on AI-enabled radio access networks through Nvidia AI Aerial. Nvidia said the work is intended to support connectivity for physical AI devices.

Mujin is assessing Cosmos for autonomous industrial robots operated through its MujinOS platform. TRON K.K. is developing manufacturing data workflows for assembly, picking, inspection, material handling, and factory digitisation.

Kawasaki Heavy Industries is applying Nvidia technology in healthcare, shipbuilding, transport, aerospace, and energy. Kubota is examining its use in autonomous agricultural equipment and farm management systems.

Enactic is fine-tuning Nvidia’s Isaac GR00T model for semi-humanoid elder-care robots. GROOVE X is using Jetson systems in its LOVOT companion robots, while Telexistence is using Isaac and evaluating Cosmos for retail automation.

Hitachi, OMRON, and Shimizu Corp. are also using Nvidia Metropolis for building management, automated inspection, and construction safety applications.

RIKEN deploys Blackwell quantum systems

Nvidia is also working with Japanese research institutions and companies on AI-for-science and hybrid quantum computing projects.

Two Nvidia-based supercomputers are beginning operations at RIKEN. RIKYU, an AI-for-science system, uses 1,600 Blackwell GPUs through the GB200 NVL4 platform.

RIKEN plans to use RIKYU for open foundation model development and research in life sciences, materials science, and laboratory automation.

A second system, ROQUO, combines 540 Blackwell GPUs with quantum computers at RIKEN facilities in Wako and Kobe. These include Quantinuum’s trapped-ion Reimei quantum processor.

Researchers are using the system to examine hybrid quantum and high-performance computing workloads. Initial work includes an evolutionary AI framework integrated with Nvidia CUDA-Q to generate quantum circuits for the Reimei system.

Nvidia is also working with AIST’s Global Research and Development Center for Business by Quantum-AI Technology. The collaboration covers Nvidia NVQLink, which provides a low-latency connection between GPUs and quantum processors, and Nvidia Ising open models for quantum processor calibration and error-correction research.

Mitsubishi Chemical, Mizuho Bank, Keio University, AIST, the University of Toronto, and Nvidia have tested a hybrid workflow for molecular spectral analysis. Nvidia said the GPU-based system completed the workload 13.4 times faster than CPU-only nodes.

The work is being applied to chemical simulations, including research into extreme ultraviolet photoresist materials used in semiconductor manufacturing.

Fujitsu and Nvidia are separately studying the use of CUDA-Q for large-scale quantum chemistry simulations. Fujitsu has also started testing NVQLink in its quantum-classical computing environment.

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.

Tech Wire Asia is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.