This trip is less about who buys Nvidia's graphics cards than about who can help the company teach machines to act in the physical world. South Korea, a manufacturing power with chips, robots and factories in equal measure, is being recast from customer to co-developer.
The pivot has a clear starting line. At the APEC summit in the southeastern city of Gyeongju last October, Huang unveiled a plan to deploy more than 260,000 Nvidia GPUs across South Korea, with Samsung, SK Group, Hyundai Motor Group and Naver each building artificial intelligence (AI)-dedicated factories to anchor the rollout.
That announcement seeded the infrastructure. The current visit is about putting it to work, turning raw computing capacity into trained systems that can drive cars, run assembly lines and move robots through real space.
The distinction matters because physical AI is where the next phase of the industry is being fought.
Cloud-based chatbots run on data centers alone, but robots and self-driving cars must perceive, decide and act in a messy physical world, and the companies that can supply both the hardware and the real-world testing grounds stand to capture the value. Few countries pair those strengths as tightly as Korea.
"Hyundai Motor has already said it would put Atlas to work on its factory floors. Korean firms are increasingly stressing automation through physical AI, embedding it in cars and then in humanoids," said You Bum-jae, principal research scientist and former head of humanoid development at the Korea Institute of Science and Technology.
Huang is expected to meet LG Group Chairman Koo Kwang-mo today, in talks centered squarely on physical AI, and the market has taken notice: shares of LG Electronics have more than tripled this year, hitting their daily limit twice on hopes of a deeper Nvidia tie-up.
"LG isn't widely known for this outside the company, but internally it has been working on a mobile-controlled, two-armed home robot," said You. "Because it is a machine that could work in both factories and homes, it could stand out all the more as physical AI."
LG Electronics has spent the past year pushing aggressively into the field, anchoring its effort on the AI home robot it calls CLOiD while developing actuators, robot sensing and smart-factory systems. It already runs Nvidia's Omniverse platform to build digital simulations of its own production lines.
Those simulations are the heart of physical AI. Before a robot lifts a part or a car merges into traffic, it rehearses inside a lifelike digital twin, repeating the task through reinforcement learning across millions of virtual trials until the behavior holds up in reality.
The conversation reaches well beyond a single company. Analysts expect cooperation to broaden across LG affiliates, drawing in the EXAONE generative model from LG AI Research, the semiconductor substrates and sensing technology of LG Innotek, and the AI cloud business of LG Uplus.
"With Huang's visit, interest in AI-RAN is rising, as Nvidia focuses on expanding the business alongside data centers," said Kim Hong-sik, an analyst at Hana Securities.
"The likely shape is a contest between a GPU-based Nvidia-Nokia camp and an Ericsson-Samsung approach built on CPUs with network solutions. More than market share, investors will watch how AI-RAN reshapes the equipment ecosystem, since it is likely to expand alongside the move to 5G standalone and 6G, completing the picture of physical AI."
South Korea's other giants slot into the same logic. SK Group and Samsung Electronics supply the high-bandwidth memory that feeds AI accelerators and are building digital twins of their own chip lines, while Naver is converting a GPU supply deal into a deeper technology alliance.
That last shift is telling. Naver Cloud said this week it would broaden its Nvidia relationship into a strategic alliance spanning AI infrastructure, models, services and physical AI, moving beyond a simple procurement arrangement.
"Our partnership with Nvidia goes beyond a simple GPU supplier-and-customer relationship," Naver Cloud CEO Kim Yu-won said. "It is a strategic decision to develop AI technology together and expand the global AI ecosystem."
Yet experts say Naver may be looking further ahead than simply becoming one of the country's top AI service providers.
"Naver doesn't build physical robots, but its framework technology for orchestrating and operating fleets of robots is said to be quite advanced," said You. "Beyond AI data centers, there could well be talk of expanding into robot-framework services."
Krafton this year launched Ludo Robotics, a humanoid venture, and has woven Nvidia's ACE character technology into its PUBG and inZOI titles. NC, long a gaming partner of Nvidia, now develops a "world model" through its NC AI unit that teaches robots the laws of physics.
The esports stop fits too. Korea turned competitive gaming into a national spectacle from the StarCraft era onward, an appetite that drove demand for graphics hardware and helped seed the computing culture Nvidia rode to dominance, and whose data-rich virtual arenas echo the simulations AI now learns from.
In the evening, Huang is set to share grilled pork belly and soju with corporate heads including SK, LG and Naver at a restaurant near Hongik University, a sequel to last year's "chimaek" gathering over chicken and beer.
The menu invites easy comparison with that earlier dinner, where the agenda does not: the chicken-and-beer night sealed Korea's role as a supplier of chips and memory, today’s is meant to bind the country into the harder, more lucrative business of making machines that think and move.
Copyright ⓒ Aju Press All rights reserved.