At the APEC CEO Summit in Gyeongju, South Korea, NVIDIA CEO Jensen Huang announced a new and deeper partnership with the South Korean government and several industry giants to expand South Korea's AI infrastructure. This massive project is expected to deploy up to 26 NVIDIA GPUs in South Korea's Sovereign Cloud and AI factories, comprehensively boosting AI innovation, industrial development, and employment opportunities in the country.
In his keynote speech, Huang Renxun emphasized that countries around the world are using sovereign AI to drive growth and innovation. He stated, "South Korea's leading position in the technology and manufacturing industries places it at the heart of the AI industrial revolution. Just as South Korean factories inspired the world with their precision ships, automobiles, and chips, the country is now able to export intelligent technologies, driving global transformation."
South Korean Deputy Prime Minister and Minister of Science, ICT and Information, Bae Kyung-hoon, also stated, "This investment in expanding the national AI infrastructure and collaborating with NVIDIA will further solidify South Korea's strengths, such as its manufacturing capabilities."
Key partners in this collaboration include:
• Samsung: Building AI-powered super factories to drive the global transformation towards smart manufacturing.
• SK Group: Building AI factories to drive South Korea's manufacturing and digital transformation.
• Hyundai Motor Group: Building AI factories and driving AI-driven mobile solutions.
Samsung is building an AI super factory, deploying over 5 GPUs and introducing Omniverse.
Among the many collaborations, the alliance between NVIDIA and Samsung is particularly deep. Samsung Semiconductor also announced today that it will collaborate with NVIDIA to build an "AI Megafactory," marking a new milestone in its journey towards AI-driven manufacturing.
Samsung plans to deploy more than 5 NVIDIA GPUs, fully integrating AI into the entire manufacturing process from design, process, equipment, operations to quality control, accelerating the research and development and production of next-generation semiconductors, mobile devices and robots.
Collaborative development of HBM4 deepens 25-year chip alliance
Samsung and NVIDIA also celebrated their more than 25-year partnership (from early NVIDIA graphics cards using Samsung DRAM to recent wafer foundry collaborations) and announced that they are jointly developing HBM4 memory.
Samsung revealed that its HBM4 uses sixth-generation 10nm-class DRAM and a 4nm logic base die, targeting a processing speed of 11Gbps, significantly exceeding the JEDEC standard of 8Gbps. Samsung will continue to provide memory solutions and foundry services such as HBM, GDDR, and SOCAMM.
Samsung adopts NVIDIA software stacking to accelerate manufacturing processes and digital twins.
Samsung plans to expand its AI Factory by incorporating NVIDIA's accelerated computing technology and applying it across its global manufacturing infrastructure, spanning memory, logic chips, wafer foundry, and advanced packaging.
• Computational lithography: By leveraging NVIDIA cuLitho and CUDA-X libraries to enhance its Optical Proximity Correction (OPC) process, it claims to have increased computational lithography speed by 20 times.
• EDA: Both parties are working with EDA partners to develop a new generation of GPU-accelerated EDA tools.
• Digital twins: Use the NVIDIA Omniverse library to create a digital twin that virtualizes the entire wafer fab's operation for anomaly detection, predictive maintenance, and process optimization.
• AI Models and Robots: Samsung has developed its own AI model based on the NVIDIA Megatron architecture (which already supports 4 million devices) and is using NVIDIA RTX PRO 6000 Blackwell servers and NVIDIA Jetson Thor modules to drive manufacturing automation and the development of humanoid robots.
• AI-RAN: Both parties will build on last year's PoC results and collaborate with South Korean telecommunications operators and academia to jointly promote the development of AI-RAN (AI-Radio Access Network) to provide real-time inference capabilities for edge devices such as robots and drones.
This AI factory infrastructure will also be extended to Samsung's global manufacturing center in Taylor, Texas, in the future.
SK Group builds 5 GPU AI factory; SK hynix strengthens cooperation with HBM.
NVIDIA also announced a deepened collaboration with SK Group to jointly build an AI factory to advance semiconductor research and development, digital twins, and AI agent development.
• AI factory scale: SK Group will build an AI factory equipped with more than 50,000 NVIDIA GPUs. The first phase of the project is scheduled to be completed by the end of 2027, and it is expected to become one of the largest AI factories in South Korea. This factory will serve subsidiaries such as SK hynix and SK Telecom (SKT), and will provide services to external clients using a GPU-as-a-Service model.
• Deepen cooperation with HBM: The two parties will strengthen their cooperation on SK hynix HBM and future memory solutions.
SK hynix incorporates AI in physical and digital twin technologies:
• Employs NVIDIA CUDA-X and NVIDIA PhysicsNeMo frameworks, and simulates using AI-accelerated chip design (TCAD).
• Building a digital twin of the wafer fab using NVIDIA Omniverse and RTX PRO servers (equipped with RTX PRO 6000 Blackwell GPUs) is a step towards self-optimized fabs.
• AI agent import: SKT will develop a base model called AX and deploy AI agents for more than 4 SK hynix employees through NVIDIA NIM microservices and AI Enterprise software to improve productivity.
• SKT builds an industrial AI cloud: SKT plans to build an industrial AI cloud in Asia, using NVIDIA RTX PRO 6000 Blackwell GPUs (with an initial deployment of over 2,000 units), to support Omniverse workloads and empower manufacturers to develop physical AI and robotics.
Hyundai Motor Group is building an AI factory to deepen autonomous driving and smart manufacturing.
NVIDIA also announced a deepened partnership with Hyundai Motor Group to jointly build an AI factory powered by NVIDIA Blackwell, accelerating its innovation in autonomous driving (AV), smart factories, and robotics.
• AI supercomputer: Hyundai Motor Group will build an NVIDIA AI supercomputer (aiming to deploy 50,000 Blackwell GPUs) to accelerate the training, validation, and deployment of models for in-vehicle AI, autonomous driving, smart factories, and robotics.
• Entity AI Application Center: In response to the South Korean government's call, the two parties will jointly establish the "NVIDIA AI Technology Center," the "Hyundai Motor Group Physical AI Application Center," and a regional data center, with a total investment of US$30 billion, in order to cultivate AI talent in South Korea.
• Smart Factory: Hyundai Motor Group will use the NVIDIA Omniverse Enterprise platform to create digital twins of its factories and use NVIDIA Isaac Sim for virtual verification of robots (including humanoid robots).
• Autonomous driving: The test used the Omniverse and Cosmos platforms (running on RTX PRO servers) to build a digital twin of the regional driving environment, advancing AV development.
• In-vehicle AI: It utilizes the NVIDIA Nemotron open model and NVIDIA NeMo software to develop its own LLM for OTA upgrades, personalized digital assistants, and intelligent infotainment. The vehicle itself will be equipped with NVIDIA DRIVE AGX Thor as its AI brain, running DriveOS and supporting advanced driver assistance systems (ADAS) and in-vehicle AI experiences.
Huang Renxun concluded, "We are collaborating with Hyundai Motor Group, a South Korean industrial giant and one of the world's leading mobility solutions providers, to create smart cars and factories that will shape the trillion-dollar mobility industry of the future."
