During a series of meetings in South Korea, NVIDIA CEO Jensen Huang strengthened relationships across memory, cloud infrastructure, robotics and AI data centers, emphasizing the country’s important role in the next phase of AI development.
Overall, the trip was a vivid illustration of Huang’s strategy of evolving NVIDIA from a top chip supplier to a vendor serving many aspects of the AI sector. For South Korea, the trip was an acknowledgment that it is already a leading supplier of AI infrastructure.
The most significant announcement was a multi-year agreement with SK hynix to jointly develop future memory technologies and secure long-term supply for NVIDIA’s upcoming AI platforms. The partnership covers future memory products intended for NVIDIA platforms, including Vera Rubin systems, RTX Spark PCs and Jetson Thor robotics platforms. The agreement offers NVIDIA greater visibility into future memory availability while providing SK hynix predictable demand as development cycles and manufacturing costs continue to rise.
“This partnership reinforces the necessity of co-design in the HBM [high bandwidth memory] supply chain,” Brendan Burke, Research Director at Futurum, told Techstrong.it. “As HBM capacity becomes the primary constraint on GPU scaling, Nvidia’s move to lock in SK Hynix for next-gen products signals a shift toward more integrated chip designs. This reduces supply risk while ensuring their roadmap for training and inference clusters keeps pace with the compute requirements of the next generation of LLMs.”
The agreement also expands beyond memory production. SK hynix plans to use NVIDIA software platforms including CUDA-X, PhysicsNeMo and Omniverse to accelerate chip development and factory operations. The company is creating digital twins of semiconductor manufacturing facilities and exploring AI-driven optimization of production processes and autonomous factory equipment.
“NVIDIA has relied on SK Hynix as its main memory provider and this agreement appears to extend this key relationship,” Gil Luria, Head of Technology Research at DA Davidson, told Techstrong.it. “While NVIDIA is also likely to incorporate memory from Micron and Samsung as well, it is likely to keep the biggest portion of its purchasing from SK Hynix.”
Agreements Across Sectors
While the SK hynix partnership addresses a key supply-chain challenge, NVIDIA’s other Seoul announcements focused on expanding AI infrastructure across South Korea.
The company unveiled a major collaboration with NAVER to expand sovereign AI capabilities. NAVER plans to increase AI infrastructure capacity beginning with a 55-megawatt deployment and ultimately scale toward gigawatt-class AI facilities. The project will use NVIDIA’s DSX AI factory platform and expand operations at NAVER’s GAK Sejong data center.
NAVER also plans to use NVIDIA technologies to advance future versions of its HyperCLOVA X models (which is NAVER’s flagship LLM that is similar to ChatGPT), support agent-based AI services and develop a Seoul World Model built from local spatial and street-view data.
The partnership demonstrates growing interest among governments and enterprises in sovereign AI infrastructure that allows organizations to operate AI services while maintaining strict control of data and regulatory compliance.
NVIDIA also expanded its relationship with LG Group, targeting robotics, mobility and AI data centers. LG Uplus plans to build a large-scale AI data center designed to support NVIDIA GPU infrastructure. The deployment is expected to provide computing resources for AI model training and deployment across LG businesses spanning electronics, automotive systems, cloud services and robotics.
The companies are also exploring future AI data center architectures, including cooling systems and facility design. As AI workloads continue to devour ever more energy, these areas are big differentiators for both hyperscale and enterprise operators.
Physical AI was a key theme of Huang’s visit. NVIDIA announced expanded cooperation with LG and Doosan Group around robotics and industrial automation. Doosan Robotics plans to integrate NVIDIA technologies including Isaac Sim, Isaac Lab, Cosmos and Jetson Thor into its robotic platforms. The companies are evaluating applications ranging from industrial automation to humanoid robotics.
Doosan Enerbility is also exploring how its power-generation technologies, including gas turbines, fuel cells and small modular reactors, could support future AI data centers.
