NVIDIA is partnering with South Korean industrial leaders to integrate physical AI into semiconductor manufacturing, robotics, and autonomous vehicle development.
Using accelerated computing and simulation, the collaboration aims to move high-fidelity digital twins and AI-driven workflows from laboratory research into large-scale, real-world industrial production environments.
Samsung Electronics is deepening its technical integration with NVIDIA to overhaul its semiconductor development pipeline. The initiative focuses on computational efficiency, specifically targeting the bottlenecks in process and optical simulations that define modern lithography.
Samsung is deploying NVIDIA cuISS and cuDSS to handle the heavy lifting of sparse linear systems. These tools are critical for technology computer-aided design (TCAD), where the complexity of calculating electron behavior at the nanometer scale often exceeds traditional CPU capabilities. In materials research, the company’s sMOL platform has achieved a 5x end-to-end speedup in quantum chemical calculations using NVIDIA cuEST.
The performance gains in thermal and mechanical analysis are even more pronounced. By utilizing NVIDIA PhysicsNeMo, Samsung reported a 1,600x speedup in AI homogenization workflows. A task that previously required 18 hours on a 32-core CPU was completed in 41 seconds on a GPU-accelerated coarse-grid solve.
Kakao Mobility is targeting level 4 autonomous driving by marrying its proprietary software stack with NVIDIA’s vehicle computing hardware. The development roadmap centers on the NVIDIA DRIVE AGX Thor platform and the NVIDIA Hyperion reference architecture.
The collaboration is specifically designed to handle the dense, complex urban traffic patterns found in Seoul’s Gangnam district. By utilizing NVIDIA Halos for safety validation, the companies intend to integrate their combined software and hardware solutions into Kia purpose-built vehicles. This move represents a strategic push to move beyond testing and toward operational mobility services.
LG Electronics has successfully demonstrated robotic manufacturing tasks using the NVIDIA Isaac GR00T foundation model. The demonstration, which utilized LG’s CLOiD robot, highlighted the efficacy of a physical AI data flywheel in a live factory environment.
This workflow connects real-world data collection with simulation-based evaluation. The process relies on NVIDIA Cosmos for data augmentation and Jetson AGX Thor for onboard compute. By establishing these reusable workflows, the partnership aims to create a repeatable framework for training and deploying robots that can adapt to changing manufacturing requirements.
| Application | NVIDIA Technology | Primary Benefit |
|---|---|---|
| Semiconductor TCAD | cuISS / cuDSS | Sparse linear system acceleration |
| Materials Research | cuEST / cuEquivariance | Quantum calculation speed |
| Thermal Analysis | PhysicsNeMo | 1,600x simulation speedup |
| Robotics | Isaac GR00T | Reusable manufacturing workflows |
Samsung is already evaluating the NVIDIA Agent Toolkit to link its simulation tools with AI models. This would allow future engineering agents to autonomously run complex analyses and compare design iterations without manual intervention.
