South Korea has secured a massive supply of Nvidia's latest Blackwell architecture GPUs under a deal announced Friday that will distribute 260,000 cutting-edge chips across the country's sovereign AI infrastructure, major conglomerates, and government computing centers. The agreement was confirmed by Nvidia following a meeting between CEO Jensen Huang and South Korean President Lee Jae Myung on the sidelines of the Asia-Pacific Economic Cooperation (APEC) summit, marking a significant expansion of the U.S. chipmaker's footprint in a nation that already anchors the global memory supply chain through Samsung Electronics and SK hynix.
How Many Blackwell Chips Is South Korea Getting and Where Will They Go?
The deal encompasses a total of 260,000 Nvidia GPUs based on the Blackwell architecture, which the U.S. government has placed under "national security" export restrictions. According to Times of India, these are the specific chips subject to U.S. export controls. The allocation is precisely segmented across five key recipients, as detailed by LiveMint:
- NAVER Cloud: 60,000 chips to expand South Korea's largest search engine's AI infrastructure.
- Samsung Electronics: 50,000 chips for a new "AI factory" aimed at embedding AI throughout the company's entire manufacturing flow.
- SK Group: 50,000 chips for deployment in the conglomerate's AI facilities.
- Hyundai Motor Group: 50,000 chips for the automaker's respective AI facilities.
- Government and Cloud Providers: 50,000 chips for Seoul's National AI Computing Centre and other cloud service and IT providers.
This distribution strategy signals a whole-of-nation approach, covering hyperscale cloud, semiconductor manufacturing, automotive, telecommunications, and sovereign compute capacity simultaneously.
Why Is the U.S. Allowing Restricted Chips to Flow to South Korea?
The Blackwell chips involved in this transaction are subject to stringent U.S. export controls designed to limit the proliferation of advanced AI compute capabilities to strategic competitors, primarily China. The fact that a license was granted for 260,000 units — a volume that dwarfs typical single-country allocations — underscores South Korea's unique geopolitical and industrial position. As a treaty ally, a major non-NATO partner, and the home of the world's two largest memory manufacturers (Samsung and SK hynix), South Korea is viewed by Washington as a critical node in the trusted "friend-shoring" supply chain. The deal effectively rewards Seoul's alignment with U.S. technology security frameworks while bolstering a democratic partner's AI sovereignty.
What Role Did President Lee Jae Myung and Jensen Huang Play in Finalizing the Deal?
The agreement was catalyzed by high-level diplomacy. President Lee Jae Myung, who has made transforming South Korea into a "leading nation in the AI domain" a centerpiece of his administration, met directly with Jensen Huang during the APEC summit. Nvidia characterized the collaboration as "working with South Korea to expand the nation's AI infrastructure with over a quarter-million Nvidia GPUs across its sovereign clouds and AI factories," according to AFP reporting cited by LiveMint. Huang emphasized the strategic fit, stating that "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."
The personal rapport was on public display Thursday evening when Huang dined with Samsung Electronics Chairman Jay Y Lee and Hyundai Motor Executive Chair Chung Euisun at Kkanbu Chicken, a local Seoul restaurant. The gathering attracted significant media attention, with national broadcasters airing live footage. Huang gifted both executives Nvidia-labelled DGX systems — the company's integrated AI appliances — and presented a signed note to Jay Y Lee reading: "To our partnership and future of the world!"
How Does This Fit Into South Korea's Existing Semiconductor Dominance?
South Korea is already the linchpin of the global memory market. Samsung Electronics and SK hynix together control the majority of the world's DRAM and NAND flash production, components that are indispensable for AI accelerators and the data centers that house them. By securing 260,000 Blackwell GPUs, South Korea is not just buying compute; it is vertically integrating its advantage. Samsung's plan to embed AI across its "entire manufacturing flow" using 50,000 GPUs suggests a future where the foundries producing the world's most advanced logic and memory chips are themselves optimized by the very AI models those chips enable. SK Group's allocation similarly reinforces the memory supply chain, while Hyundai's share points to ambitions in autonomous driving and robotics. NAVER Cloud's 60,000 chips ensure the domestic software and services layer has sovereign training and inference capacity.
What Happens Next for AI Infrastructure in the Region?
The deployment of a quarter-million Blackwell GPUs will likely accelerate the construction of "AI factories" — a term Nvidia uses to describe data centers purpose-built for large-scale model training and inference. With the National AI Computing Centre receiving a portion of the government allocation, Seoul gains a sovereign resource for public research, startups, and strategic projects independent of foreign cloud providers. For Nvidia, the deal locks in a massive, multi-year revenue stream from a single democratic ally and deepens the ecosystem lock-in around its CUDA software stack and DGX hardware platforms. For the broader industry, it sets a precedent for how the U.S. may manage the distribution of controlled AI compute to trusted partners: large, transparent, government-to-government facilitated deals that reinforce allied industrial bases rather than fragment them.