NVIDIA helps Korea map a nationwide AI buildout with new labs and plans for data centers and “physical AI”

South Korea used an AI Summit in San Francisco to put a national flag in the ground for artificial intelligence, positioning the country not just as a participant in the coming compute boom but as a “key engine” behind it. President Jae Myung Lee met with NVIDIA co-founder and CEO Jensen Huang ahead of and during the event, while South Korean business leaders and researchers announced a series of initiatives that connect AI research to data-center expansion and efforts to bring AI into the physical world through robotics and vehicles.

The centerpiece announced by NVIDIA and its Korean partners is a push to expand “full-stack” AI capacity, a term used to describe the layered system needed for modern AI, from chips and memory to cloud infrastructure, software models, and deployment in real-world products. NVIDIA said it is building on its long relationship with Korea, including work in semiconductors and AI infrastructure, and that the summit discussions advance Korea’s path toward becoming a global center for AI innovation.

NVIDIA and the Korea Advanced Institute of Science and Technology, known as KAIST, announced a joint AI research lab at KAIST’s Kim Jaechul Graduate School of AI in Seoul. NVIDIA called it the first joint AI lab between a Korean university and any global technology company. The lab will focus on “agentic AI,” or AI systems that can take actions toward goals rather than only generating text or predictions, and it is intended to bring together NVIDIA’s research and engineering expertise with KAIST’s scientific talent. NVIDIA said it will connect that academic program to NVIDIA Nemotron open models and NVIDIA AI Cloud partner computing.

Beyond research, the summit agenda also emphasized large-scale data-center buildouts framed as “national AI factories.” NVIDIA said NAVER, Brookfield, and NVIDIA will expand NAVER’s NVIDIA DSX AI factory at its GAK Sejong data center to 200 megawatts, describing that level as roughly 100,000 GPUs, built on the NVIDIA Vera Rubin platform. The expanded effort is more than triple an initial buildout announced in June. NVIDIA also said that “Vera Rubin” is the data-center platform underpinning its next generation of AI systems, while DSX refers to NVIDIA’s AI infrastructure software and reference deployments for running AI workloads at scale.

In parallel, NVIDIA and SK Group described a broader, multi-party partnership to build AI infrastructure at a scale measured in global compute demand. The announcement said the expanded collaboration is sized at “$500-billion-plus” and plans to deploy NVIDIA Vera Rubin infrastructure powered by SK hynix HBM4 memory. HBM stands for high-bandwidth memory, a specialized chip-stacked memory technology used to keep pace with the bandwidth demands of accelerated AI computing, and HBM4 is SK hynix’s next iteration referenced in the partnership description.

Hyundai Motor Group added a “physical AI” element tied to robotics and autonomous development. NVIDIA said Hyundai’s physical AI strategy is anchored by a robot reference platform developed with NVIDIA, and that the companies plan to extend a buildout with 50,000 NVIDIA Blackwell GPUs. Blackwell refers to NVIDIA’s GPU architecture marketed for accelerating AI workloads. NVIDIA also said Hyundai plans to integrate NVIDIA DRIVE Hyperion, an autonomous vehicle development platform, with Hyundai vehicle platforms. NVIDIA described DRIVE Hyperion as combining NVIDIA DRIVE AGX in-vehicle computing, which runs on the safety-certified NVIDIA DriveOS operating system.

NVIDIA said it is also supporting physical AI research through planned centers at universities. It described planning an NVIDIA AI Technology Center, or NVAITC, with KAIST’s Department of Mechanical Engineering for collaborative research, talent development, and knowledge exchange in physical AI. NVIDIA said the work may tap into NVIDIA Nemotron open models and complement the earlier agentic AI research lab. It also said NVIDIA and Seoul National University are planning an NVAITC spanning foundation models, accelerated computing, physical AI, and AI for science, with potential cooperation on AI infrastructure, including work harnessing NVIDIA Nemotron and NVIDIA Cosmos open world foundation models. “Open world foundation models” are designed to learn across broad environments and scenarios rather than only a single closed task.

The event’s political and industry framing was a notable part of the message. NVIDIA said President Lee opened by referencing the enthusiasm he saw around Jensen Huang’s visit to Seoul and noted his own wish to join in the experience. NVIDIA also reported Huang’s comments at the summit about the level of excitement across Korea. In a stage lineup that included OpenAI’s Sam Altman, Broadcom’s Hock Tan, and Anthropic’s Dario Amodei, the common theme was that Korea’s role extends “throughout the stack,” from semiconductors and infrastructure to talent concentration. NVIDIA further said Samsung Electronics Chairman Jay Y. Lee, SK Group Chairman Chey Tae-won, Hyundai Motor Group Executive Chair Euisun Chung, and NAVER founder Hae-jin Lee joined Huang for tours and photos, emphasizing government-industry coordination.

NVIDIA did not lay out detailed timelines, contractual terms, or unit economics for the proposed expansions in the announcement. It also did not specify whether all planned capacity will be delivered in a single phase or how much of the compute is targeted for internal use versus customer services. For investors and observers, the most concrete items are the capacity targets and technology labels, such as the 200-megawatt DSX expansion and the 50,000 Blackwell-GPU Hyundai buildout, but the full implementation plan remains to be clarified.