Physical AI demands more than readiness

Nvidia CEO Jensen Huang's visit to South Korea triggered a frenzy. Young people followed his every move, gathering at restaurants, PC gaming cafés, baseball stadiums and corporate offices he visited. Shares of Nvidia-related companies rose, and interest became so intense that a website tracking his whereabouts, “Jensen Huang's Footsteps,” was launched. Huang's remarks only added to the excitement. “The era of robotics and physical AI has finally arrived. No country is better prepared than Korea,” he said. He also praised South Korea as a rare country with strengths spanning heavy industry, manufacturing, electronics, software and AI research. Many took those comments as a sign that Nvidia sees South Korea as a key partner in the age of physical AI. Students at Seoul National University reportedly even coined a nickname for him: “K-Jensen.”

The enthusiasm is understandable. Few countries can match South Korea's industrial breadth, from semiconductors, shipbuilding and automobiles to gaming, robotics and artificial intelligence. Physical AI may be powered by software, but it ultimately has to operate in the real world. That requires places where technologies can be tested, refined and deployed at scale. South Korea offers precisely that environment.

Still, it is worth listening carefully to what Huang actually said. He did not call South Korea a winner in the coming era of physical AI. He called it well prepared. Those are not the same thing. Success will likely belong to the countries that build the strongest physical AI ecosystems, from core technologies and software platforms to components, manufacturing capacity and supply chains. South Korea's strengths are undeniable, but important gaps remain. Boston Dynamics, owned by Hyundai Motor Group, may have developed advanced robots such as Atlas, yet the domestic content ratio in South Korea's robotics industry remains only in the 40% range. Critical components such as reducers and control systems still rely heavily on Japan, while rare-earth magnets used in motors largely come from China.

Questions also remain about who will control the software foundations of physical AI. World models, which help AI systems understand, predict and respond to changes in the physical environment, are emerging as a crucial technology. South Korean companies are investing in the field, but dependence on overseas platforms remains significant. Nvidia's Cosmos currently stands as the most influential world-model platform.

South Korea is positioned to become one of the most important markets and testing grounds for physical AI. For now, however, it is closer to being a country that excels at applying technologies developed elsewhere than one that defines the technologies and supply chains themselves.

Huang's visit underscored South Korea's potential in physical AI. But once the initial excitement fades, a more crucial task lies ahead. The country must take a clear-eyed look at both its strengths and its weaknesses. South Korea enters the AI era with advantages few nations have. Whether it will become a leader in physical AI or remain a top-tier user of others' platforms is still uncertain. That is the real question Huang's visit raised, and it deserves much deeper reflection than the initial buzz.