Nvidia is making increasingly bold moves into physical AI for robots, cars and drones, moving beyond chips for AI training.

In a recent report, the Wall Street Journal said Nvidia has entered a stage of generating $10 billion a year in revenue from its physical AI business. The share is not large compared with its total revenue of $303 billion over the past four quarters, but the company has high expectations for physical AI.

CEO Jensen Huang said the physical AI business will grow tenfold within the next 10 years. At an event in San Francisco in July, he also said, "If you can generate video of fingers moving or a hand picking up a cup, there is no reason you can't make a robot move the same way."

Nvidia is pursuing a strategy of providing dedicated chips along with developer software to keep robot and robotaxi developers tied to its ecosystem. It also distributed open-weight models such as GR00T and Cosmos for free, allowing robot developers to download and modify them freely. Michael Frank (마이클 프랭크), CEO of intelligence startup Radiant Intel, said, "Physical AI is a channel for Nvidia to move beyond data centers into a much larger real economy of vehicles, factories, warehouses and robots."

Huang's children are also directly involved in the related business. Jensen Huang's daughter Madison Huang is in charge of marketing for physical AI, and his son Spencer Huang serves as a director in charge of product management. After visiting LG Electronics' robotics data factory in Seoul in August, Madison Huang toured a robot exhibition in Beijing the next day. She watched robots demonstrate kicking and also visited booths of Chinese companies using Nvidia technology.

The United States is blocking Nvidia from selling high-performance AI chips to China, but deals for robot and automotive chips remain open. China is the world's largest industrial robot market, and about 90 percent of the world's humanoid robots were shipped from China in the first half of this year. Brady Wang (브래디 왕), an analyst at Counterpoint Research, said, "China is key to Nvidia's physical AI roadmap," adding, "Chinese companies have strong manufacturing and commercialisation capabilities and provide Nvidia with many real-world use cases."

China has moved to develop its own AI chips at the national level, but local robot manufacturers still use Nvidia semiconductors and software to train and operate machines, the WSJ reported, citing industry sources. Chinese robot companies such as Unitree are also reported to be using Nvidia's latest Thor chip. Nvidia said, "Nvidia complies with U.S. laws and works with companies around the world."