Samsung Cuts Simulation Time from 18 Hours to 41 Seconds in NVIDIA Physical AI Collaboration

Among the domestic corporate collaborations NVIDIA announced on November 9, 2026, Samsung Electronics used GPU acceleration to complete a semiconductor simulation that previously took about 18 hours in just 41 seconds. NVIDIA stated that it is supporting the development of Korea's 'physical AI' ecosystem alongside domestic companies including Samsung Electronics, LG Electronics, and Kakao Mobility, leveraging accelerated computing, AI, and simulation technologies. Physical AI refers to technology that enables AI to understand the physical environment of the real world and control robots, vehicles, and similar systems.

What Are the Three Companies Doing with NVIDIA Technology?

• Samsung Electronicsis applying AI technology to semiconductor materials research, process development, and device design. According to NVIDIA, in thermal and mechanical analysis of semiconductors using 'Physics NeMo,' a simulation that previously took about 18 hours on a 32-core CPU was completed in 41 seconds using GPU acceleration technology—roughly a 1,600-fold increase in computational speed.
• LG Electronicsis advancing robot training and manufacturing environment simulation using NVIDIA's robot AI model 'Isaac GR00T.' The two companies have completed a demonstration of LG Electronics' 'CLOi' robot operating in an actual manufacturing environment and have also built a technology development framework that repeatedly cycles through AI training and performance verification.
• Kakao Mobilityis developing a Level 4 autonomous driving system—one that operates within designated zones without driver intervention—using NVIDIA's autonomous driving platform 'Hyperion' and safety technology 'Halos.' The company plans to integrate these technologies centered on Kia's purpose-built vehicles (PBVs) to lay the groundwork for commercializing autonomous driving services in Korea.

Implications for AX at Korean Companies

What stands out in Samsung Electronics' case is not the speed itself but the number of iterations it enables. When a single simulation run takes about 18 hours, verifying even one change in design conditions per day is a tight squeeze—but at 41 seconds, you can run the simulation repeatedly while varying conditions. For organizations where design and process analysis takes a long time and are considering GPU acceleration, the logical approach is to start with the simulation task where the iteration bottleneck is most severe.

LG Electronics advanced robot training through manufacturing environment simulation and went as far as demonstrating CLOi operating in an actual manufacturing environment. The sequence involves repeated training and verification in a virtual environment before moving to the actual field. If your organization is considering robot adoption, it is safer to first examine whether you have the means to train and verify the robot in a virtual environment, rather than focusing on the robot model itself.

The three companies are also at different stages of progress. A single announcement bundled together results already achieved (Samsung Electronics), a completed demonstration (LG Electronics), and ongoing development toward a planned commercialization foundation (Kakao Mobility). When using collaboration announcements as reference material for adoption reviews, these three stages must be read separately.

There is one thing worth checking this week: make a list of which design and process analysis tasks in your organization take so many hours per run that iteration is effectively blocked.

Source: NVIDIA Partners with Samsung, LG, and Kakao to Grow Korea's 'Physical AI' Ecosystem