News

Nvidia Expands Cosmos Physical AI Platform with Edge Model, Japanese Manufacturing Partners

Key Takeaways

  • Nvidia introduced Cosmos 3 Edge, a 4-billion-parameter world model designed for on-device robot perception and action generation, extending the Cosmos 3 family it launched in May.
  • A group of Japanese manufacturing and technology companies, including Fanuc, Yaskawa Electric, Fujitsu, Hitachi, and Sony, intend to join the Nvidia Cosmos Coalition to help build open physical AI models.
  • The announcement came during a two-day visit to Japan by Nvidia Chief Executive Jensen Huang, who also unveiled new Jetson edge computing modules and met with Japanese government officials.

Nvidia has expanded its open physical AI platform with the introduction of Cosmos 3 Edge, a new world model designed to run directly on robots and other edge devices, while announcing that a group of Japanese manufacturing and technology companies intends to help develop and adopt the platform through the Nvidia Cosmos Coalition, the company said.

The announcement came during a two-day visit to Japan by Nvidia CEO Jensen Huang, part of the company's broader effort to deepen its presence in Japan's AI, robotics, and advanced manufacturing sectors.

Cosmos 3 Edge Targets On-Device Physical AI
Cosmos 3 Edge is a 4-billion-parameter model built on Nvidia's Nemotron architecture, designed to help robots and vision AI agents perceive their surroundings, reason in real time, and generate actions locally rather than relying on cloud-based inference. According to Nvidia's developer documentation and accompanying technical report, the model differs from the Cosmos 3 Super and Cosmos 3 Nano models introduced in May in that it is trained from scratch rather than initialized with pretrained Qwen3-VL weights.

Nvidia said developers can adapt the base model to specific robots, vehicles, sensors, and operating environments in about a day using the open Cosmos framework. The company said Cosmos 3 Edge is designed to run across Nvidia RTX GPUs, DGX systems, and Jetson edge computing modules, including the newly introduced Jetson T2000 and T3000.

Cosmos 3, the model family that Edge extends, debuted in May as what Nvidia described as the first fully open omnimodal foundation model for physical AI, combining world generation, physical reasoning, and action generation. Nvidia has published Cosmos 3 Super and Nano as open weights on Hugging Face and GitHub, together with training scripts and datasets, and has made the models available through cloud providers including Microsoft Azure and CoreWeave.

Japanese Manufacturers Join the Cosmos Coalition
Nvidia also announced that a group of Japanese companies intends to join the Nvidia Cosmos Coalition, an industry collaboration focused on advancing open world models and software for physical AI.

Companies named in the announcement include AIRoA, Fanuc, Fujitsu, Hitachi, Kawasaki Heavy Industries, Kubota, NEC, SoftBank, Sony, and Yaskawa Electric.

According to Nvidia, Fujitsu is exploring a collaborative control platform for physical AI, while Fanuc, Yaskawa Electric, and Kawasaki Heavy Industries are integrating Nvidia technologies into their robotics systems. Another group of organizations, including Honda R&D, Omron, Mitsui, Telexistence, and Enactic, is building applications on the Cosmos platform for retail automation, elder care robotics, and industrial inspection.

During the visit, Huang framed the initiative in the context of Japan's long history in manufacturing. "Japan invented modern manufacturing. Now, it has the opportunity to reinvent it for the age of intelligent industries," according to a report from Evertiq.

Broader Context
's visit also included a meeting with Japan's Ministry of Economy, Trade and Industry and Minister Ryohei Akazawa, according to BigGo Finance, which cited Japanese media reports. The visit followed criticism in parts of the Japanese press that Huang's previous Asia itinerary included South Korea and Taiwan but not Japan.

Alongside Cosmos 3 Edge, Nvidia introduced the Jetson T2000 and T3000 edge computing modules. Huang also addressed reports concerning the company's upcoming Vera Rubin computing platform, saying it remains in production and on schedule, according to BigGo Finance. That statement had not been independently confirmed by English-language wire services at the time of publication.

About the Author

John K. Waters is the editor in chief of a number of Converge360.com sites, with a focus on high-end development, AI and future tech. He's been writing about cutting-edge technologies and culture of Silicon Valley for more than two decades, and he's written more than a dozen books. He also co-scripted the documentary film Silicon Valley: A 100 Year Renaissance, which aired on PBS.  He can be reached at [email protected].