The Dawn of a Golden Age for Physical AI

At the recent "NVIDIA AI Day Seoul 2026," the chipmaker outlined a definitive vision for the future: Physical Artificial Intelligence has reached a pivotal inflection point. NVIDIA posits that as agent-based AI matures, the entire lifecycle of robots—from development and training to deployment—is being dramatically accelerated. Physical AI is set to become the core driver propelling the AI industry into its next phase.

A Full-Stack Platform for Robotics Development

Deepu Talla, Vice President of Robotics and Edge AI at NVIDIA, detailed the company's technological architecture. He explained that a complete robotics development pipeline typically involves three critical stages: model training, simulation, and real-world deployment. To bridge these phases, NVIDIA has built a full-stack support system.

  • DGX Systems: Provide the foundational compute power for large-scale AI model training.
  • RTX and Jetson Platforms: Serve high-performance edge computing and embedded devices respectively, acting as the "brains" for robot deployment.
  • Omniverse Digital Twin Technology: This acts as the bridge between virtual and physical worlds. Developers can test and train robots within highly realistic digital environments, significantly reducing the cost and risk of physical debugging and enabling a seamless transition from training to deployment.

Deepening Ecosystem Partnerships in Korea: Semiconductors and Automotive

NVIDIA is engaging in a series of deep industry collaborations within the Korean market, applying Physical AI to cutting-edge manufacturing sectors.

With Samsung & SK hynix: Building Smart Semiconductor Fabs

NVIDIA is collaborating with these memory chip giants to build high-fidelity digital twins of semiconductor fabrication plants using the Omniverse platform. These digital replicas enable real-time monitoring of complex equipment and full production workflows. By analyzing this data, companies can predict maintenance needs and optimize production cycles, thereby enhancing operational efficiency and yield.

With Hyundai Motor Group: Accelerating Robot Training in the Digital Realm

In its partnership with Hyundai, digital twin technology is being used to train industrial robots. The key advantage is that training within the simulation environment can run hundreds of thousands of times faster than real-time. This allows robots to accumulate vast "experience" in an extremely short period, quickly mastering complex tasks like assembly and handling, and rapidly adapting to changes on the actual production line.

With LG Electronics: Empowering Service Robots

NVIDIA's GR00T foundational model for robots is helping LG train its service robot, CLOiD. Through this platform, CLOiD can more accurately recognize, grasp, and transport a variety of everyday objects, making it more reliable and practical for home or commercial service scenarios.

Pushing the Frontiers of Autonomous Driving

Beyond robotics, NVIDIA announced a collaboration with Korean mobility service provider Kakao Mobility to advance the development of Level 4 autonomous driving systems. The plan is to integrate NVIDIA's Hyperion autonomous vehicle sensor and compute platform, along with its Halos safety validation technology, into Kakao Mobility's autonomous driving software stack. Initial development and testing will be based on Kia's Purpose-Built Vehicle (PBV) platform.

This flurry of collaborations demonstrates that NVIDIA, through its powerful AI computing platforms and digital twin technology, is deeply embedding itself into multiple critical industry chains—from semiconductor manufacturing and automotive to service robotics—aiming to become an indispensable infrastructure provider for the era of Physical AI.