Huang's Korea tour showcases Korea's rise from Nvidia supplier to physical AI partner

Huang's extended visit to Korea underscores the country's evolving status within Nvidia's ecosystem — no longer just a supplier of memory chips, but an increasingly important customer for Nvidia's advanced AI processors.

His latest visit, running from Friday through Monday, centered on key demand centers ranging from gaming outlets and internet firm Naver to companies that have adopted a focus on robotics such as Hyundai Motor, Doosan and LG.

An unexpected player on the roster was Doosan, an industrial conglomerate spanning machinery and construction, but the company has become not only a supplier of a key material used in AI accelerators but a client for Nvidia platforms for developing and training robots.

In a celebratory gesture, Huang threw the ceremonial first pitch at a Doosan Bears home game at Jamsil Baseball Stadium in southern Seoul alongside Chairman Park Jeong-won.

“Business is booming, and Korea is doing very well. My partners here in Korea are very important to me,” Huang told reporters on the sidelines of a Friday dinner with SK Chairman Chey Tae-won, LG's Koo Kwang-mo and Naver's Lee Hae-jin.

His final day itinerary on Monday is expected to include a meeting with Hyundai Motor Executive Chair Euisun Chung, followed by separate stops at the LG Electronics and Naver offices.

Although no specific deal had been announced as of press time on Sunday, nearly every stop on Huang's itinerary involved companies viewed as key customers or partners for Nvidia's newest AI platforms.

“Jetson Thor is widely expected to power Atlas, the humanoid robot developed by Hyundai Motor Group’s affiliate Boston Dynamics,” Kim Doo-un, an analyst at Hana Securities said. Jetson Thor is Nvidia's robot-focused AI processor designed to serve as the onboard computing platform for humanoids.

“RTX Spark could find applications in high-performance PCs and on-device AI systems. The result is that Korean companies are becoming increasingly intertwined with Nvidia's next wave of growth businesses, from next-generation AI accelerators and physical AI to AI PCs,” Kim said, speaking about Nvidia's Arm-based system on chip and computing platform.

PC bangs on the rise

PC bangs, Korea’s signature gaming cafes, emerged as recurring venues as well, serving as the first stop for the CEOs meeting with esports legend Faker on Friday and later hosting a separate gathering with Krafton Chairman Chang Byung-gyu and NC CEO Kim Taek-jin Sunday.

There, Huang pitched the upcoming RTX Spark chip that will be integrated directly into into laptops to deliver far more powerful on-device AI capabilities.

Developed through partnership with Krafton, an upcoming "co-playable character" powered by Nvidia ACE technology and optimized for the new RTX Spark laptops was demoed by Huang. The voice-activated AI acts as an interactive, real-time squad mate that can chat and strategize directly inside PUBG: Battlegrounds.

NC is also pushing its boundaries into physical AI through its subsidiary NC AI. The gaming company is developing AI systems capable of controlling robots and factory operations, while also working on models that enable robots to understand and interact with physical environments autonomously.

Memory still counts

As memory capacity remains vital to powering Nvidia's architecture, Huang pressed Korean suppliers to ramp up production of high bandwidth memory chips (HBM).

When he sat down for the star-studded dinner at a samgyeopsal (pork belly) restaurant in Mapo District, western Seoul, he noted that Nvidia's new products — such as Vera Rubin, Vera and RTX Spark — will require more memory, including high bandwidth memory and low-power double data rate, or LPDDR, memory.

Huang also acknowledged that three memory makers — Samsung, SK hynix and Micron — passed Nvidia’s quality test for their HBM4, which will be fitted into the Vera Rubin platform.

The Nvidia chief had another round of meetings with SK executives on Sunday, including Chairman Chey, SK hynix CEO Kwak Noh-jung and SK Telecom CEO Jung Jae-hun.

Additionally, Nvidia, Hyundai Motor and the Korean government are reportedly in the final stages of talks to establish an AI research and development center in Korea, which Huang said will be located in Seoul.

“The clearest bottleneck in the AI investment boom remains memory,” Kim said.

“If GPUs are the engines of AI, then HBM and high-capacity dynamic random-access memory are its fuel and circulatory system. As AI models grow larger, the amount of computation required rises exponentially, making memory bandwidth increasingly critical,” the analyst noted.

Physical AI and automobiles

With Hyundai Motor expanding its automotive footprint into robotics via its Boston Dynamics subsidiary, the synergy between the two industries has taken center stage.

Their deepened partnership with Nvidia leverages advanced AI processors to power both factory robotics and the automaker’s next-generation vehicle intelligence.

At the heart of this autonomous driving push is the Nvidia Drive AGX Thor—an in-vehicle compute platform designed to manage everything from advanced driver assistance systems to critical safety features.

Chung and Huang had a luncheon on Sunday, but did not disclose the contents of the meeting.

“One more thing: Nvidia introduced a brand-new line of robotics processors for self-driving cars. We have a big partnership with Hyundai,” Huang said. “All of my friends, [including those at] LG, SK hynix, Samsung, Hyundai and Naver, [our businesses] are all booming, so I came to congratulate them for their big partnerships.”

“Korea is going to be very, very busy,” he told reporters. “We are also building a very prominent research center here in Korea for AI research. [...] We are currently recruiting, so if you know anyone studying AI, tell them to apply for our center here.”

BY PARK EUN-JEE, LEE JAE-LIM [[email protected]]