Originally published by:Engineering.com
M4S Take

Architecture: Built on NVIDIA Vera Rubin NVL72 racks and the NVIDIA DSX platform, connected via NVIDIA Spectrum-X Ethernet networking.

  • Partnership: NVIDIA and Noetra Corp. are launching an NVIDIA Vera Rubin AI factory in Japan, billed as the world's first national AI infrastructure for physical AI.
  • Hardware scale: The facility will contain 13,750 NVIDIA Vera CPUs and 27,500 NVIDIA Rubin GPUs, with 140 megawatts of data center capacity.
  • National mission: The factory supports Japan's FRONTia Project for multimodal foundation models in robotics and physical AI, strengthening the AI ecosystem across manufacturing, logistics, healthcare, and telecommunications.
  • Open access: Pretrained model weights and software including NVIDIA Nemotron, NVIDIA Cosmos, NVIDIA Isaac GR00T open models, and NVIDIA NeMo libraries will be available to domestic developers and enterprises.

NVIDIA is working with Noetra Corp. to launch an NVIDIA Vera Rubin AI factory in Japan — a facility the companies describe as the world's first national AI infrastructure for physical AI. The factory will support Japan's FRONTia Project, an initiative focused on multimodal foundation models for robotics and physical AI.

Scale and Architecture

The hardware numbers are substantial. The facility will house 13,750 NVIDIA Vera CPUs and 27,500 NVIDIA Rubin GPUs, delivering 140 megawatts of data center capacity. That power envelope puts it in the upper tier of dedicated AI infrastructure builds.

The AI factory will be built using NVIDIA Vera Rubin NVL72 racks and the NVIDIA DSX platform, with the deployment connected and scaled through NVIDIA Spectrum-X Ethernet networking. According to the companies, this architecture is intended to provide breakthrough AI performance, lower token costs, and massive scale for frontier AI training.

Purpose: Foundation Models for Physical AI

The factory's core mission is enabling the development of open multimodal foundation models — the kind that power AI agents, digital twins, robotics, and other physical AI applications. For manufacturing professionals, the digital twin and robotics angles are the ones to watch: these are the workloads closest to the factory floor.

The broader goal is strengthening Japan's AI ecosystem across manufacturing, logistics, healthcare, and telecommunications.

Open Access to Model Weights

One notable element of the plan is accessibility. The AI factory will make the pretrained weights of Noetra's multimodal foundation models available to domestic model developers and enterprises, rather than keeping them behind a closed API. The software stack includes NVIDIA Nemotron, NVIDIA Cosmos, NVIDIA Isaac GR00T open models, and NVIDIA NeMo libraries.

That open-weights approach could lower the barrier for Japanese enterprises — including manufacturers — looking to build physical AI applications on top of nationally hosted infrastructure.

SM

Simon Morton

Editor, M4SNews

With a background in heavy engineering, process engineering, digital marketing & AI. My mission, to cut through the news and make it easy to digest.

M4SNews marks eighteen years of independent operation, connecting manufacturers and engineers with the intelligence that actually matters on the factory floor.

Is this your company?

This article features your business. Claim it to add your logo, contact details, and a link to your website — or upgrade to reach more buyers.

Did you know 80% of Press Releases trigger AI content warnings? Reach out and the M4S team can assist.