Inside Nvidia's $0 Billion Bet on South Korea: From Memory Chips to Robot Factories

In June 2026, Nvidia CEO Jensen Huang touched down in Seoul for a five-day charm offensive that reshaped the semiconductor giant’s relationship with an entire nation. It was a trip that mixed cold noodles with Hyundai’s executive chair, a ceremonial first pitch at a baseball game, and late-night barbecue with tech titans . But behind the spectacle, Huang walked away with a portfolio of agreements spanning memory chips, cloud infrastructure, and robotics—deals that collectively aim to transform South Korea from a critical component supplier into a full-stack, sovereign AI hub.

The most significant commitment was a multi-year memory partnership with SK hynix, which Huang himself called the “biggest achievement” of the visit . Under this deal, SK hynix will co-develop next-generation memory components for Nvidia hardware, including the upcoming “Vera Rubin” platform, moving beyond a commodity supply relationship into a strategic technology partnership . Huang warned that even SK hynix’s plan to double production capacity by 2030 would likely be insufficient to meet surging AI demand, signaling that the memory shortage could persist for several years .

The second pillar of the trip was the agreement with SK Telecom to build a gigawatt-scale AI Cloud in South Korea using Nvidia’s full-stack DSX platform . SK Group Chairman Chey Tae-won framed the project as an effort to become “a leading AI factory player shaping Asia’s AI ecosystem,” with the first AI factory expected to come online in 2027 . Telecom networks, Huang argued, are now “becoming national AI infrastructure” .

These two anchor deals were flanked by a series of collaborations that fill out the physical AI stack. Naver is expanding its GAK Sejong AI data center from 55 megawatts toward gigawatt scale and became the first Korean company to join the Nvidia Nemotron Coalition . LG Group agreed to build an AI factory focused on robotics, autonomous driving, and GPU cloud services . Hyundai Motor Group discussed expanding work with Nvidia across mobility, robotics, and physical AI, while Doosan Group separately announced a robotics partnership .

From memory supplier to sovereign AI powerhouse

The SK Telecom AI factory deal is the linchpin for South Korea’s ambitions in physical AI—the class of artificial intelligence that controls robots, autonomous vehicles, and industrial machinery in the real world . Because the infrastructure is designed as “sovereign” AI, Korean companies and government agencies can run sensitive industrial workloads domestically on Korea-owned compute, rather than depending on foreign cloud providers .

This matters because the AI factory is integrated with Korea’s industrial base. SK Telecom’s nationwide network means it can distribute AI inference and physical-AI control close to factories, ports, and logistics hubs . And the factory deal is paired directly with robotics collaborations at Hyundai and Doosan, giving Korea both the compute muscle and the end-user industries to lead in physical AI.

The binding problem: memorandums, not mandates

For all the fanfare, Korean industry observers have pointed out that most of the announced partnerships are “loosely binding and unclear” . The majority of the discussions amounted to memorandums of understanding (MOUs) or verbal agreements, which carry no legal force, performance obligations, or penalties for breach . This is a significant caveat: the SK Telecom AI factory’s 2027 target is a welcome exception with a clear milestone, but the broader transformation of Korea’s AI ecosystem hinges on converting these frameworks into funded, operational projects.

The trip, as some analysts see it, was largely about assigning Korea a defined role within Nvidia’s sprawling global AI infrastructure strategy . Whether the country can seize that role on its own terms—rather than as a captive supplier—depends on execution in the coming years. For now, the agreements have put South Korea on the map as one of the few places on earth with the combined memory manufacturing, telecom infrastructure, and industrial robot expertise to build AI not just for the cloud, but for the physical world.