With CEO Jensen Huang of Nvidia visiting Japan, Nvidia unveiled a strategy to expand artificial intelligence (AI) cooperation with Japanese manufacturing and automobile corporations. As Huang recently visited Korea and Taiwan in succession, a so-called "Japan passing" controversy arose in Japan. Through this announcement, Nvidia laid out a plan to speed up its push into Japan's manufacturing and robotics markets by expanding physical AI cooperation with Toyota, FANUC, NEC, SoftBank and others.
Huang visited Tokyo on the 15th to attend an event marking the 30th anniversary of the partnership between Nvidia and SEGA. In a separate media briefing on its Japanese partner ecosystem, Nvidia announced plans to expand AI cooperation with Japanese corporations, focusing on manufacturing, automobiles and robotics.
Nvidia positioned "Cosmos," a world foundation model for physical AI, as the key to its Japan strategy. Physical AI is technology that enables AI to perceive and judge its surroundings so that real devices such as robots or self-driving cars can carry out tasks on their own.
Nvidia is focusing on Japan because of its competitiveness in manufacturing and the robotics industry. Japan has many world-class corporations in automobiles, industrial robots and factory automation. Nvidia believes such a manufacturing base is essential to apply AI to real production sites. Its strategy is to expand the manufacturing AI market by supplying open AI models and software platforms so these corporations can develop AI using their own data.
Another notable point in this announcement is that Japan's leading manufacturing corporations have joined Nvidia's "Cosmos" ecosystem in large numbers. FANUC, Fujitsu, Kawasaki Heavy Industries, Kubota, Mitsubishi Electric, NEC, SoftBank and Yoshikawa will use Cosmos to develop AI models specialized for manufacturing and robotics. Omron, Sony and Woven by Toyota will use Nvidia's Isaac robotics platform and Jetson computers for robots to develop next-generation robots.
Nvidia supports corporations in developing industry-specific AI models by further training on data accumulated in manufacturing sites and robot operations instead of using general-purpose AI as-is. To that end, it is pushing an open-model strategy that discloses model weights, training data and development tools. The company said corporations can build and operate their own AI systems without disclosing production data and technical know-how externally.
Nvidia also offers open AI models by industry, including Nemotron, a large language model (LLM); Cosmos and GR00T for physical AI; and BioNeMo for life sciences.
Nvidia unveiled a new model for physical AI, "Cosmos 3 Edge." It downsized the model and computation requirements so it can run on Jetson, a compact Edge AI computer. Nvidia said it cut required computing resources to about one-fourth of previous Cosmos models, allowing corporations to reduce the time and expense of fine-tuning the model with their own data.
Another feature is that devices can perform AI inference directly without sending data from factories or robot sites to the cloud. It runs quickly even in environments with unstable network connections and does not require sending sensitive production data outside.
Nvidia also unveiled next-generation Edge AI computers, "Jetson T3000" and "Jetson T2000," to run AI in industrial sites such as factories and robots. It will allow software for the next-generation products to be developed in advance on existing Jetson developer kits and plans to supply the new hardware starting in the first quarter of 2027. By using Agentic AI, it will also support robot software development and memory optimization, aiming to shorten development time and expense.
It is also broadening the scope of cooperation with Toyota from automobiles to factories and Smart City projects. Toyota is developing next-generation driver-assistance systems based on Nvidia's in-vehicle AI platform DRIVE AGX and its operating system DRIVE OS. Nvidia's strategy is to expand cooperation with automakers beyond Autonomous Driving to production and urban infrastructure to widen the reach of its AI platforms.
On production sites, it will use Omniverse and Isaac Sim to build a Digital Twin of the production line. Toyota is also developing a code-generation AI using Nemotron that supports the automotive software safety specification (MISRA), and it plans to apply Nvidia platforms to AI-based urban mobility systems and traffic management technologies.
Nvidia said Japan has an AI ecosystem with more than 300,000 developers and more than 16,000 students. The industry views this announcement as a strategy by Nvidia to draw Japanese manufacturing and robotics corporations into its AI platform ecosystem. Analysts say it is a move to expand the market beyond supplying AI Semiconductor chips to also provide AI models, robotics platforms and Digital Twin software together.