"The semiconductor industry, centered on memory technology, is world-class, and heavy industry is also outstanding. With added software competitiveness, Korea has the perfect environment to harness the artificial intelligence (AI) revolution."
Nvidia CEO Jensen Huang offered this assessment of the backdrop to his back-to-back visits on the 8th to the headquarters of major Korean conglomerates including SK, LG, Hyundai Motor, and Naver, where he held a series of meetings with their chiefs. With Samsung Electronics Chairman Lee Jae-yong abroad on a business trip, Huang met separately that evening with Jun Young-hyun, vice chairman and head of Samsung's semiconductor (DS) division, at the Shilla Hotel to discuss AI memory strategies in depth. The discussions covered, in the short term, sufficient supply of sixth-generation high-bandwidth memory (HBM) and SOCAMM starting this year, as well as foundry (contract chip manufacturing) cooperation on seventh- and eighth-generation HBM4E and HBM5, and the Grok fourth-generation language processing unit (LPU).
Huang subsequently met with AI startup CEOs at the Shilla Hotel to advance ecosystem expansion. Analysts say the move is a strategic bid to make Korea's AI value chain — "memory, advanced manufacturing, communications networks, and proprietary AI models" — a testbed for Nvidia's AI infrastructure.
Taken together, Huang's actions and announcements that day lend weight to the view that he is positioning Korea as a whole as a "frontier outpost of global AI innovation." The cooperation between the two sides is concentrated in four major areas: AI factory construction and operation, physical AI, and mobility.
The AI factory, at the core, is not a mere data storage facility. It refers to a next-generation intelligent data center that draws on vast amounts of power and data as raw materials to continuously produce tokens — the key information unit of AI — around the clock. Just as Industrial Revolution-era factories churned out goods powered by electricity, AI-era factories produce intelligence.
Korean firms' hardware capabilities are being fully mobilized to build the massive AI infrastructure. As Huang explained, "Our large-scale partnership with SK is our first cooperation model spanning multiple platforms and technologies," SK hynix will supply AI memory, including HBM to be installed in next-generation accelerators, supporting the brain of the AI factory. "Thermal management" and "power," which have emerged as the biggest challenges in operating AI factories, will be handled by LG and Doosan. LG Electronics is conducting certification for a liquid cooling solution (CDU) that cools server heat, while Doosan Enerbility can supply the large-scale power needed by factories with low carbon emissions through small modular reactors (SMR) and gas turbines.
On the operations side, running the built infrastructure via the cloud, SK Telecom and Naver have taken the lead. SK Chairman Chey Tae-won said, "By sharing a future roadmap with Nvidia, we will rapidly adapt to AI demand," formalizing the construction of a full-stack AI cloud spanning everything from chip setup to data center operation.
At Naver, board chairman Lee Hae-jin personally stepped forward to agree on a joint gigawatt (GW)-scale global AI factory venture. As Lee assessed it as "presenting a concrete alternative for building independent sovereign AI capabilities," the blueprint calls for operating a 1GW-class mega-infrastructure capable of ultimately housing hundreds of thousands of the latest graphics processing units (GPUs) — starting with operation of a 55-megawatt facility each in Sejong in the first half of 2027, expanding to 200 megawatts in 2028, and extending infrastructure beyond the Asia-Pacific to Europe and the Middle East.
Physical AI, which brings AI out of virtual space and into the reality we live in, will be handled by Hyundai Motor, LG, and Doosan. LG Chairman Koo Kwang-mo said, "The (physical AI) ecosystem blueprint Nvidia envisions aligns with LG's future vision," and LG Electronics will jointly develop humanoid robots based on "Isaac GR00T," Nvidia's robot AI brain. Doosan Robotics' industrial robots and others will also join Nvidia's platform, accelerating innovation on manufacturing floors.
In the mobility sector, Hyundai Motor Group made a bold play. Hyundai Motor Group Chairman Euisun Chung met Huang at the company's Yangjae headquarters and abruptly proposed that Nvidia join as a co-investor in the roughly 9 trillion won "Saemangeum Project." The core is building large-scale AI data centers and an advanced robot manufacturing plant with a capacity of 30,000 units annually in the Saemangeum area.
Huang responded, "Saemangeum will become a wonderful 'AI Valley,'" adding, "I am very pleased to be building Nvidia (data centers) in Saemangeum."
