The physical AI concept is being implemented at an accelerated pace! Nvidia (NVDA.US) and Toyota (TM.US) 10-year partnership upgrade: full-stack AI technology penetrates Toyota's “Weaving City”

The Zhitong Finance App learned that on July 14, Nvidia (NVDA.US) and Toyota Motor (TM.US) jointly announced the expansion of their 10-year strategic cooperation to expand the application of physical artificial intelligence (Physical AI) from the field of autonomous driving to a wider range of physical scenarios such as robots, smart factories, and smart cities. Nvidia CEO Hwang In-hoon personally visited Tokyo this week to promote this cooperation. This is a critical period for Japanese Prime Minister Sanae Takaichi to promote the development of the domestic AI industry.

This cooperative upgrade marks the most significant expansion since Toyota selected the Nvidia DRIVE PX platform to test autonomous driving systems in 2017. Last year, the two sides announced that Toyota would use the NVIDIA DRIVE AGX platform and the NVIDIA DriveOS operating system that has passed safety certification to develop next-generation models.

From L2++ to “Weaving City”: Four major landing scenarios for physical AI

The core implementation scenario of this collaboration is “Woven City” (Woven City), an experimental smart city built by Toyota in Shizuoka Prefecture. Nvidia will provide it with AI technology for traffic management. The “Weaving City” was completed and occupancy began in the fall of 2025, and it is expected that about 2,000 people will live there. The city was designed as a real-world laboratory to test cutting-edge technologies such as autonomous driving, intelligent transportation, and robotics in a real environment. Nvidia's participation means the smart city will receive full-stack technical support from chips to AI software.

A ten-year partnership: From Drive PX to L2++ mass production

The partnership between Nvidia and Toyota began in 2017, when Toyota selected the Nvidia Drive PX platform to test its early autonomous driving systems. In 2025, the two sides further deepened their cooperation, and Toyota promised to use the Nvidia Drive AGX Orin platform in its commercial vehicle fleet.

Following this upgrade, Toyota is using the NVIDIA DRIVE AGX platform equipped with a safety-certified NVIDIA DriveOS operating system to develop next-generation models with L2++ driver assistance functions. The L2++ level system can more accurately understand the road environment and make driving decisions while complying with Toyota's strict safety standards.

Four major landing scenes

Car: L2++ assisted driving. Toyota is developing next-generation vehicles using the NVIDIA DRIVE AGX platform, equipped with a safety-certified NVIDIA DriveOS operating system to achieve advanced L2++ driver assistance functions. Under the premise of complying with Toyota's strict safety standards, new cars can more accurately understand the road environment and make driving decisions.

Software development: AI accelerates code generation. Automobiles are accelerating their transformation to software definition. Toyota has developed a Code Assistant AI model that complies with Misra standards, uses NVIDIA Megatron-LM for training and fine-tuning, and refers to various data sets such as NVIDIA Nemotron. This customized AI model helps engineers generate, review, and validate safety-critical code more efficiently, while meeting strict industry specifications while shortening development cycles.

Factories: Omniverse digital twins. Toyota uses the NVIDIA Omniverse library and the NVIDIA Isaac Sim Open Framework at manufacturing sites to simulate production processes, robot motion, and a complete digital twin environment. Testing and optimizing production plans in a virtual environment can help reduce downtime, increase efficiency, and reduce costs.

Smart City: Woven City visual language model. Woven by Toyota, a subsidiary of Toyota, developed the Woven City AI Vision Engine, a multi-modal visual language model using Nvidia H100 Tensor Core GPUs and Megatron-Core. The model targets urban traffic intelligence and can analyze the actual environment, predict developments, and help traffic and infrastructure systems take timely countermeasures. Woven City is an experimental prototype community built by Toyota on the site of a former factory in Shizuoka Prefecture to test cutting-edge technology as a real-world laboratory.

Hwang In-hoon's Visit to Japan: Double Bets on Sovereign AI and Physical AI

Huang Renxun is visiting Japan to attend an event commemorating the 30th anniversary of the cooperation between Nvidia and Sega held in Tokyo on July 15. On the sidelines of the Tokyo developer event, while attending the event in Tokyo, Hwang In-hoon also clarified rumors of Vera Rubin's postponement. Hwang In-hoon told reporters that Nvidia's next-generation Vera Rubin AI accelerator hardware is moving towards a “huge” scale of production.

Hwang In-hoon also emphasized the importance of sovereign AI in Tokyo: “National-level intelligence capabilities must be cultivated, strengthened, and developed domestically, and must be established within the country.” Japan is facing pressure from a declining workforce, and the factory automation and robotics industries are speeding up the introduction of artificial intelligence to improve efficiency. At the same time, he refuted the AI bubble theory, saying that demand for AI infrastructure is still “extremely strong.”

Sun Zhengyi spoke out in support of AI investment during the same period. Earlier this week, SoftBank founder Sun Zhengyi said at the SoftBank World Conference that questions about excessive investment in AI were “foolish.” SoftBank has previously finalized a $30 billion investment agreement for OpenAI, which will be settled in three installments. Sun Zhengyi predicts that AI data centers will require 3 terawatts of electricity by 2040, which is equivalent to 1.8 times the current total global electricity consumption.

Japan's AI ecosystem fully rolled out: Nemotron model and Jetson Thor

Nvidia is simultaneously expanding its corporate AI cooperation ecosystem in Japan. Japanese companies using Nvidia's Nemotron big language model include ENEOS (ENEOS), NTT DATA, and Hitachi. Nemotron is an open model series released by Nvidia in February 2026. It covers a variety of scales, from 9 billion to 320 billion parameters, and allows commercial use using relaxed licenses.

Nvidia has also launched a new Jetson Thor computer, which aims to advance the development of mainstream robotics technology and edge AI. The chip is based on a next-generation GPU architecture. The computing power is about 8 times higher than the previous generation Jetson Orin, and can support real-time inference scenarios such as humanoid robots, industrial automation, and edge AI servers. According to Futures Express, Toyota plans to integrate the Jetson Thor platform into its next-generation unmanned delivery vehicles and storage robots by 2027.

In the broader Japanese manufacturing AI layout, leading Japanese manufacturing companies such as FANUC (FANUC), Fujitsu, Kawasaki Heavy Industries, Kubota, Mitsubishi Electric, NEC, and SoftBank have joined Nvidia's “Cosmos” physical AI ecosystem. Omron, Sony, and Woven by Toyota will use the Nvidia Isaac robotics platform and Jetson computers to develop the next generation of robots.

Market and strategic significance: a key step in moving physical AI from concept to implementation

This cooperative upgrade marks that physical AI is moving from concept to large-scale industrial implementation. Deepu Talla, vice president of robotics and edge AI at Nvidia, said: “We are expanding our collaboration to advance the application of physical artificial intelligence in automotive, robotics, and smart cities.”

For Nvidia, this partnership is a key step in its transformation from an AI chip vendor to a physical AI full-stack platform provider. Japan has many world-class companies in the fields of automobiles, industrial robots, and factory automation, and Nvidia believes that such a manufacturing foundation is essential for applying AI to actual production sites.

This collaboration is a key step in extending its AI landscape from the data center to the physical world — the three platforms OmniVerse (digital twin), Isaac (robot simulation), and DRIVE (autonomous driving) have been collaboratively deployed on the same customer. For Toyota, this is the technological cornerstone of its transformation from an “automobile manufacturer” to a “mobility company” — from L2++ assisted driving to digital twin factories to smart city AI, Nvidia's full-stack technology is penetrating every aspect of Toyota's business.

For Toyota, this is an important technical support for its transformation from a traditional car manufacturer to a mobility and smart city service provider. The two sides hope to open up AI technology in the fields of automobiles, infrastructure, and industrial operations, and jointly build a new mobility system.

It is worth noting that after Hwang In-hoon's previous visits to South Korea and Taiwan, there was a “transit through Japan” dispute within Japan. Through this announcement to expand cooperation with leading Japanese companies such as Toyota and Fanuc, Nvidia has clarified plans to accelerate its entry into the Japanese manufacturing and robotics market.

As the focus of the global AI competition shifts from big language models to the physical world, this partnership upgrade between Nvidia and Toyota provides a reference model for “how to implement physical AI” — from chips to operating systems, from digital twins to real-world laboratories, a complete path from virtual to reality, and from code to steel is being paved.