Nvidia CEO Physical AI Trend: Jensen Huang’s Big Bet on the Next Industrial Revolution — and What It Means Now That He’s Gone
Attending conferences in a leather jacket and making everything seem inevitable was Jensen Huang’s habit. Standing in front of a group of engineers and investors at Nvidia’s GTC event in Berlin last month, he shared the stage with a row of tiny duck-shaped robots created by a French startup. He requested that they perform somersaults. Without delay, three of them toppled over. As he concluded his speech, one turned completely in the wrong direction. He told it, seemingly without irony, “Pay attention, stay focused.” It was a humorous moment. In a roundabout way, it was also the whole point.
The phrase “physical AI,” which Huang had been promoting more and more over the past year, is precisely what it sounds like. It’s the notion that artificial intelligence must advance from text production and image recognition to machines that genuinely exist in the real world. Friction-aware robots. autonomous cars that have the ability to anticipate when a pedestrian will step off a curb. systems for warehouses that avoid people by using unplanned routes. Huang spent years arguing that Nvidia was bridging the huge gap between what AI can do on a screen and what it needs to do in a factory.
The engineering behind it is undoubtedly complex, but the fundamental argument is not. According to Huang, the majority of AI in use today is ignorant of the laws of physics, including cause and effect, friction, and inertia. Children learn these concepts before they are able to read. It irritates me when a chatbot responds incorrectly. When a robot makes the same kind of error, it can harm people or damage equipment. The difference between digital error and physical consequences is what distinguishes Physical AI from its predecessors.
Nvidia developed a whole platform to address this issue. Before they ever touch a real object, robots can practice in virtual environments created by Cosmos and Omniverse. The purpose of Isaac and Newton, physics simulation engines, is to assist machines in acquiring a sense of motor instinct. Robots can process data in real time instead of pinging a remote server thanks to Jetson’s edge computing hardware, which is integrated directly into robots. The stack is full, and that’s on purpose. Huang realized that Nvidia needed to control the entire ecosystem, not just sell chips.
The ensuing international alliances were important. SoftBank is spearheading a group of over 40 businesses in Japan, including Honda and Sony, to create a domestic AI model called Noetra, which incorporates Nvidia technology. According to several accounts, Huang’s reception by Japanese officials during his July visit to Tokyo felt more like a state visit than a business trip. Japan, a nation with strong manufacturing ties and an aging labor force, has clear reasons to be concerned about robotic automation and sees physical AI as a true engine of economic growth. Huang had been advocating for years for this framing—industrial necessity rather than technological novelty.
Then, on November 12, Nvidia announced that Jensen Huang had abruptly resigned from his positions as president, CEO, and board member. Colette Kress, the CFO, was appointed temporary CEO. That day, Nvidia’s stock fell more than 4%, but after-hours trading saw a partial recovery. His departure was the result of personal accusations, and the company only provided the most basic information about the change.
It’s difficult to ignore the timing. According to most analyzes, physical AI was Nvidia’s most obvious strategic move outside of its core chip business. Huang had personally promoted it in Berlin, Tokyo, London, Paris, Stockholm, Dublin, and throughout a European tour. His explanation of why it was important was the loudest. It’s genuinely unclear at this point whether Huang’s personally cultivated partners will remain as dedicated and whether Nvidia’s next permanent CEO shares that conviction.
The existence of the technology, the existence of the market, and the ongoing investment are all clear. In partnership with Nvidia, Germany’s Agile Robots intends to start manufacturing up to 2,000 humanoid robots this year. The Cosmos simulation platform is up and running. Whether physical AI would occur was never really a question. Simply put, the person who seemed most certain of it would no longer be employed there.
