For NVIDIA and CEO Jensen Huang’s recent trip to Japan, not only is he paying respect to his long-time partner, SEGA, in the gaming world by promoting RTX Spark, he’s also there to close some AI deals on a national scale, too, as the latest partnership is with Noetra Corp.
The Japanese state-funded initiative, FRONTia, focuses on physical AI, and they will be using NVIDIA hardware to build the country’s largest AI factories that host 27,500 Rubin GPUs and 13,750 Vera CPUs spread across 382 Vera Rubin NVL72 racks. The entire deployment will run on the DSX reference platform and use Spectrum-X Ethernet networking, creating a 140-MegaWatt AI infrastructure designed to train trillion-parameter multimodal AI models as the system expands over time.
The project received an initial ¥387.3 billion (around US$2.4 billion) in first-year funding, with total government support potentially reaching ¥1 trillion (around US$6.1 billion) over five years, although future funding will depend on annual performance reviews.
According to Noetra’s roadmap, the consortium plans to deliver a reasoning-focused AI foundation model in fiscal 2026, followed by an omni-modal model capable of understanding text, images, audio, and video by fiscal 2028. By fiscal 2030, the goal is to develop real-world native AI with advanced spatial awareness for robotics and industrial applications.
Noetra itself is a newly formed consortium backed by SoftBank Corp., Sony, NEC, Honda, and 44 participating companies and organizations. Together with Japan’s National Institute of Advanced Industrial Science and Technology (AIST), the consortium was selected by the Japanese government to lead the FRONTia program from fiscal 2026 through 2030.
