[Journalism Internship] Why Nvidia chose Korea: Inside the 260,000 GPU Bet on ‘physical AI’

Home > Think English > Journalism Internship

print dictionary print

[Journalism Internship] Why Nvidia chose Korea: Inside the 260,000 GPU Bet on ‘physical AI’

Nvidia CEO Jensen Huang and Samsung Electronics Executive Chairman Lee Jae-yong embrace as Hyundai Motor Group Executive Chair Euisun Chung looks on during an event at COEX in southern Seoul marking the 25th anniversary of Nvidia’s GeForce GPU launch on Oct. 30. [JOINT PRESS CORPS]

Nvidia CEO Jensen Huang and Samsung Electronics Executive Chairman Lee Jae-yong embrace as Hyundai Motor Group Executive Chair Euisun Chung looks on during an event at COEX in southern Seoul marking the 25th anniversary of Nvidia’s GeForce GPU launch on Oct. 30. [JOINT PRESS CORPS]

 
Ham Min-soo, Kong Jun-hoo, Ha Yeon-jin, Park Jae-yun

Ham Min-soo, Kong Jun-hoo, Ha Yeon-jin, Park Jae-yun

 
At a time when countries are struggling to power the next wave of AI, Korea has emerged as one of the few nations capable of hosting large-scale GPU clusters — a key reason Nvidia is sending more than a quarter million of its top-tier processors to the country.
 
At the Asia-Pacific Economic Cooperation summit in Gyeongju, North Gyeongsang, on Oct. 31, Nvidia drew global attention by officially announcing that it will supply 260,000 units of its latest GPUs to Korea. The allocation included 50,000 each to the Korean government, Samsung, SK and Hyundai Motor and 60,000 to Naver Cloud.
 
The scale alone marks a historic moment in Korea's global AI race, as GPUs serve as the brains of AI robots, referred to as physical AI. However, the fierce competition among countries scrambling for high-performance GPUs raises the question: Why did Nvidia choose to distribute such unprecedented quantities to Korea, ahead of others?
 
Nvidia is a company that creates GPUs, which accelerate the training of AI models, and being able to train AI models faster and accurately allows for more trial and error, directly impacting the quality of AI models.
 
In Korea, GPUs became a focal issue when President Lee Jae Myung, while running for office, promised that he would work to secure more than 50,000 high-performance graphics processing units. Physical AI could be promising to companies since robots could reduce labor costs and risks of danger in factories.
 
 
Why Korea became Nvidia’s strategic choice
 
Isaac GR00T, an open humanoid robot developed by Nvidia, is pictured in this photo provided by the company. [NVIDIA]

Isaac GR00T, an open humanoid robot developed by Nvidia, is pictured in this photo provided by the company. [NVIDIA]

High bandwidth memory (HBM), the core of Nvidia’s products, helps understand their strategic decision. Modern high-performance GPUs, such as the H100 and GH200, are reliant on HBM to handle colossal amounts of data in advanced AI tasks.
 
Crucially, over 90 percent of global HBM production is controlled by Korea, mostly from SK hynix and Samsung Electronics. This makes Korea play a key role in producing GPUs.
 
From Nvidia's perspective, Korea is not just a customer but an indispensable strategic partner, as the U.S. company needs HBM produced in Korea. By granting 260,000 GPUs to Korea, Nvidia is making its GPU supply chain stable.
 
The second factor solidifying Korea’s strategic importance is its readiness in power and data center infrastructure. It is true that a large allocation of GPUs will allow more advanced research in AI, but high-capacity infrastructure is needed to run them safely and continuously.
 
In this regard, Korea stands out as having advanced levels of AI data centers. According to “Global Data Center Trends 2024” from CBRE, Korea’s AI data centers are not just fast but also possess a global-best standard for power grid stability and robust transmission infrastructure.
 
Additionally, the Korean government is in the process of developing long-term solutions for the immense power consumption required for AI data centers, which will lower the risk for large-scale research.
 
 
Inside Korea’s emerging physical AI powerhouses
 
SK Hynix’s HBM3E high-bandwidth memory is displayed at the SK AI Summit 2025 exhibition at COEX in Gangnam District, Seoul, on Nov. 3. [YONHAP]

SK Hynix’s HBM3E high-bandwidth memory is displayed at the SK AI Summit 2025 exhibition at COEX in Gangnam District, Seoul, on Nov. 3. [YONHAP]

 
On Oct. 30, Nvidia announced that it is working with Korea to expand the nation’s AI infrastructure with over a quarter million Nvidia GPUs across its sovereign clouds and AI factories.
 
Nvidia CEO Jensen Huang said, “We are in the process of creating a new industry called AI,” and emphasized, “Korean companies are necessary," showing Nvidia's view of Korea as a key testing ground for the next generation of physical AI.
 
Hyundai Robotics’ manufacturing systems, Naver Labs’ work in autonomous navigation and spatial intelligence, KAIST’s advanced robotics research and Samsung Electronics’ capabilities in both robotics and chip design together create an ecosystem where every layer of physical AI can be tested and integrated.
 
For Nvidia, this makes Korea one of the rare nations capable of evaluating the entire physical AI stack, positioning it as an ideal partner for validating and scaling new AI models in real-world industrial environments.
 
Korea’s major technology firms — including Samsung, SK, Hyundai, Kakao and Naver — plan to use the 260,000 GPUs to build large-scale “AI factories,” facilities designed to train and operate advanced machine learning and robotics systems.
 
Industry analysts note that these companies already operate semiconductor fabs, robotics divisions and cloud platforms, giving them the technical capacity to integrate high-performance GPUs into real industrial applications. Rather than signaling general readiness, their involvement illustrates how Korea’s existing industrial ecosystem can absorb and deploy accelerated computing at scale.
 
 
The opportunities and costs of a physical AI future
 
Yet, the rise of physical AI sparks new problems. According to the Chosun Ilbo, a set of GB220 graphic cards, which is equivalent to 36 GB220s, uses approximately 132 kilowatts. Thus, the 260,000 GPUs that Korea secured would use about 480 megawatts of electricity, which is a similar amount to what a medium-sized thermal power plant would produce. The overwhelming use of electricity could lead to the construction of additional energy plants, which in turn could exacerbate economic and environmental issues.
 
Still, this rise could help address one of Korea’s major problems: population decline, including the decrease in its work force. Physical AI presents a compelling solution to its long-term population challenges.
 
Physical AI offers a promising advancement in areas such as autonomous robots, smart factories and AI-driven logistics, which could help offset a severe labor shortage by replacing repetitive or physically demanding roles. Goldman Sachs predicts that up to 50 percent of jobs could be fully automated by 2045.
 
In this emerging landscape, as Korea is promised an unprecedented amount of GPUs, it is slowly approaching the global physical AI.
 
The United States faces power grid shortages and permitting delays for new AI data centers, while China remains blocked by U.S. export controls on advanced GPUs. Southeast Asian nations lack the semiconductor base and technical work force needed for high-end deployment.
 
Against this backdrop, analysts say Nvidia's selection of Korea carries geopolitical weight: Korea is one of the few U.S.-aligned countries with both a critical semiconductor supply chain and the industrial capacity to scale physical AI.
 
This combination of HBM production, reliable infrastructure, and advanced manufacturing makes Korea unusually well-positioned to absorb and operate Nvidia's most powerful systems.
 
“Korea’s leadership in technology and manufacturing positions it at the heart of the AI industrial revolution — where accelerated computing infrastructure becomes as vital as power grids and broadband,” said Huang while in Korea on Oct. 30. Ultimately, Nvidia increasingly views Korea not as a simple consumer of its technology but as a true “AI ally.”
 

BY HAM MIN-SOO, KONG JUN-HOO, HA YEON-JIN, PARK JAE-YUN [[email protected], [email protected], [email protected], [email protected]]
Log in to Twitter or Facebook account to connect
with the Korea JoongAng Daily
help-image Social comment?
s
lock icon

To write comments, please log in to one of the accounts.

Standards Board Policy (0/250자)