[AI Prism] NVIDIA “Golden Age of Physical AI”… Accelerating cooperation with Korean companies

NVIDIA declared that the golden age of ‘physical artificial intelligence (AI)’, which moves and performs tasks directly in the real world, has begun.

As the development of agentic AI, which makes decisions and acts on its own, accelerates the development and introduction of robots into the field, it is predicted that AI based on the physical world will lead the next transition.

Cases of cooperation with Korean companies were also introduced, from Samsung Electronics and SK Hynix's semiconductor factory digital twin construction to Hyundai Motor Group's robot learning and LG's robot development.

◇ “Robotics is a three-computer problem”… From learning to deployment

Deepu Tala, Vice President of NVIDIA Robotics and Edge AI, said at the 'NVIDIA AI Day 2026' held at COEX in Gangnam-gu, Seoul on the 9th, "A golden age has begun as agentic AI has become a facilitator in accelerating the development and deployment of physical AI."

Vice President Tala pointed out, "Two-thirds of the world's gross domestic product (GDP) occurs in the physical world, but over the past 50 years, manufacturing automation has been limited to mass production and small-item processes. Small and medium-sized enterprises account for 90% of manufacturing, but the technology has not developed sufficiently until now."

However, as agentic AI technology has recently matured and robots can be learned and verified quickly and safely through simulation, conditions have been prepared to apply physical AI to industrial sites.

Vice President Tala said, "NVIDIA is not a company that directly manufactures robots, but a platform company that provides the core technologies needed for robot development," and noted that it is building an open ecosystem that connects everything from model learning to simulation and field deployment.

He defined robotics as 'three computer problems' including learning, simulation, and field deployment, and emphasized that the company's three computing technologies are needed in this process, including DGX, which trains robots, RTX, which simulates in a virtual environment, and Jetson, which is mounted on actual robots and runs in real time.

We also introduced 'Omnibus', a digital twin construction environment that virtually reproduces real space in order to learn and verify robots before putting them into the actual field.

“Just a few years ago, it took months and dozens of people to create a digital twin, but recent technological advancements have significantly reduced the time it takes to build one,” Tala explained.

In addition, he introduced the World Foundation model platform 'Cosmos', the humanoid robot platform 'Groot', and the robot development framework 'Isaac', and said, "Through an open model strategy that discloses model weights and learning and test scripts, we are supporting companies to utilize the latest technology while protecting their own data and intellectual property rights."

◇ “From industrial alliance to AI alliance”… Applying physical AI to Korean manufacturing sites

NVIDIA evaluated that as Korea builds AI infrastructure in addition to semiconductors, cooperation with NVIDIA is moving from an industrial alliance to an AI alliance.

Vice President Tala said that we are learning robots based on digital twins with Hyundai Motor Group and said, “We are speeding up the experimentation of the automobile manufacturing process through repetitive simulations that are hundreds of thousands of times faster than actual time.”

Samsung Electronics and SK Hynix announced that they are using omnibus to build a digital twin of a semiconductor factory and applying it to facility and process monitoring and standard operating procedure (SOP) analysis. He added that he is also collaborating with LG on everything from robot development to distribution.

Soyoung Jeong, CEO of NVIDIA Korea, emphasized, "The topic of AI is rapidly moving from generative AI to agentic AI, where software makes decisions and acts on its own, and further to physical AI, which moves and performs tasks directly in the real world. NVIDIA AI Factory is creating this trend of change."

CEO Jeong expressed his intention to contribute to the leap forward of Korea's AI ecosystem, saying, "I hope that this event will not only be about experiencing technology and products, but will also be a place for exchange where Korean companies, research institutes, and startups share experiences and ideas."

Meanwhile, 'NVIDIA AI Day 2026' will be held for two days from this day to the 10th.

On the first day, a keynote speech and 15 sessions in 3 tracks will be held on the topics of physical AI, agentic AI, and AI infrastructure. On the second day, special events such as hands-on workshops for developers, 'AI for Science Forum', and 'Physical AI Startup Connect' will be held.