Nvidia bets big on Korea’s AI future with 260,000 Blackwell processors

GYEONGJU, North Gyeongsang — Four major Korean corporations — Samsung Electronics, SK Group, Hyundai Motor and Naver — as well as the Korean government have just sealed landmark deals with Nvidia to deploy 260,000 Blackwell processors in a sweeping push to infuse AI across the nation’s core manufacturing supply chains.

Of the total, most will be used in the private sector, as 50,000 GPUs each will go to Samsung Electronics, SK Group and Hyundai Motor, and an additional 60,000 units will be supplied to Naver Cloud. The remaining 50,000 GPUs will be allocated to the Korean government, according to the company. The timeline and investment size for the Blackwell deployment were not disclosed.

Nvidia CEO Jensen Huang pinpointed Korea, with its deep-rooted manufacturing base, as a test bed to transform entire industries including semiconductors, automobiles, shipbuilding, biotechnology and robotics with AI, turning the country’s industrial backbone into an intelligent, automated system.

“For Korea, this is an extraordinary opportunity because very few countries in the world have the essential, necessary qualities in order to succeed the next generation of AI [referring to physical AI],” Huang said in a keynote address at the Asia-Pacific Economic Cooperation (APEC) CEO Summit on October 31.

“You have the technical capabilities, you have the software capabilities, you have artificial intelligence capabilities and you also have such an incredible manufacturing base and industrial base.”

The Nvidia chief hailed Korea as poised to become “one of the largest GPU or AI infrastructure countries in the world” through the deal.

Huang arrived in Korea on October 30 to attend the APEC summit, sharing an intimate chimaek (chicken and beer) meal with Samsung Chairman Lee Jae-yong and Hyundai Executive Chairman and CEO Euisun Chung at a small fried chicken joint in southern Seoul. The power trio also appeared together shortly afterwards at the nearby Coex Mall to celebrate the 25th anniversary of Nvidia’s GeForce graphics card in Korea.

The following day, Huang traveled to Gyeongju, North Gyeongsang, to attend the summit where he joined not only Samsung’s Lee and Hyundai Motor’s Chung once again, but also Korean President Lee Jae Myung, SK Group Chairman Chey Tae-won and Naver founder Lee Hae-jin for a commemorative ceremony marking the GPU supply deal.

Nvidia’s processors remain in short supply worldwide amid an AI boom, making the 260,000-GPU commitment — one of its largest ever to a single market. The company has also pledged 120,000 Blackwell GPUs to the UK and 100,000 to Germany.

Virtuous AI chip cycle
Nvidia’s agreements with Samsung and SK, which includes memory giant SK hynix, mark the start of what Huang labeled “virtuous” AI chip cycle. Both companies are supplying Huang’s company with high bandwidth memory (HBM) chips, a key component in Nvidia’s GPUs. Samsung and SK will then in turn use the Blackwell GPUs to enhance their own manufacturing capabilities through AI-driven optimization.

Samsung Electronics will deploy its 50,000 processors to integrate its semiconductor manufacturing process — from design and production to equipment, operations and quality control — into a single, intelligent network. In this setting, AI will continuously analyze, predict and optimize workflows in real time.

Having recently earned Nvidia certification for its HBM3E chips and shipped samples of next-generation HBM4, Samsung has been one of Nvidia’s closest technology partners. Huang referred to the company on October 31 as Nvidia’s first collaborator in developing the memory technology that powers modern AI.

Samsung and Nvidia will also jointly implement digital twin manufacturing across memory, logic, foundry and packaging facilities using Nvidia’s Omniverse platform, enabling predictive maintenance and production optimization. The system will first be introduced at Samsung’s domestic fabs before expanding to overseas sites, including its Taylor, Texas facility.

The Korean tech giant is also in talks to supply high-performance graphics, seventh-generation graphic double data rate (GDDR7) and next-generation low-power memory module Socamm2, while exploring deeper collaboration in the foundry business, according to company officials.

Meanwhile, SK Group will serve as Korea’s sole AI cloud provider powered by Nvidia’s Blackwell and Omniverse platforms, tailored for local manufacturers and public-sector use. About 2,000 processors will be used to build the cloud system, deployed at SK hynix’s Icheon campus and Yongin semiconductor cluster in Gyeonggi.

Some of the remaining GPUs will also go toward SK’s Ulsan AI data center, a $5 billion joint project with Amazon Web Services, slated for completion by 2027. SK hynix announced earlier this week that HBM4 shipments to clients, including Nvidia, will begin in the fourth quarter of this year.

Building the physical AI cluster
Nvidia has also partnered with Hyundai Motor Group to advance what Huang calls “physical AI” — the integration of AI into the real world through robotics, mobility and automation. The companies plan to co-invest $3 billion, including the value of the 50,000 GPUs, to develop a physical AI ecosystem encompassing autonomous driving, smart manufacturing and next-generation vehicle platforms.

“We’re working with technology companies throughout the world in order to create a robotic factory,” Huang said. “The entire factory is a giant robot. And inside this robotic factory, it will orchestrate robots that are working together with humans, with all of us, and they’re going to produce products like self-driving cars that are robots. So robots orchestrating robots, building products that are robotic. This is the future of artificial intelligence.”

The investment will fund three physical AI-focused facilities — a technology center, application center and data center — though specific locations and stakes have not yet been revealed.

Hyundai plans to leverage Nvidia’s DGX, Omniverse and DRIVE Thor computing platforms to infuse AI into its vehicle design, production and operations. Omniverse Enterprise will enable digital twin modeling of production sites to streamline automation, predictive maintenance and manufacturing efficiency.

Expanding AI across industries
Naver Cloud will partner with Nvidia to build a physical AI platform using 60,000 GPUs, Omniverse and Isaac Sim to develop AI models for industries such as shipbuilding, energy and biotechnology. The initiative aligns with Naver’s goal of developing a sovereign AI ecosystem based on domestic data and infrastructure.

The remaining 50,000 GPUs will be used in Korea’s government-led AI projects, including national foundation models developed with LG AI Research, Naver Cloud, NC AI, SK Telecom and Upstage. Nvidia will also work with Samsung, SK Telecom, KT and LG U+ on 6G wireless network development, and partner with the Korea Institute of Science and Technology Information (KISTI) on the Hangang supercomputer project.

BY LEE JAE-LIM [lee.jaelim@joongang.co.kr]