Jensen Huang’s Visit to South Korea (Part 1): The Inception of the AI Infrastructure War

Let's learn more about Nvidia CEO Jensen Huang's visit to South Korea

— The Realignment of the Semiconductor Supply Chain (Upstream) — In June 2026, NVIDIA CEO Jensen Huang’s visit to South Korea carries significance that goes far beyond a typical corporate itinerary. From meeting Faker and appearing on You Quiz on the Block to throwing the ceremonial first pitch at a baseball game and hosting casual "Somaek" (somaek-and-fried-chicken) gatherings, his celebrity-like movements have consistently made headlines. However, this is not mere fan service; it is a meticulously calculated component of his "pro-Korea corporate marketing." It serves as clear evidence that the status and importance of South Korean companies in the global semiconductor supply chain remain irreplaceable. Immediately prior to his arrival in Korea, Huang attended COMPUTEX 2026 in Taiwan, where he unveiled NVIDIA's next-generation AI strategy and audited core supply chains, including TSMC. Synthesizing these recent developments brings a singular, critical agenda to the forefront: "The Realignment of the Global Supply Chain in Preparation for the AI Infrastructure Era." This article focuses specifically on the upstream segment (raw materials and front-end manufacturing processes) of the semiconductor supply chain—the bedrock of AI infrastructure —to provide a fact-based analysis of the multifaceted opportunities involving NVIDIA, South Korean enterprises, and the broader global supply chain, including China and India. 1. The Message from Taiwan: “AI is Becoming an Infrastructure Industry” During his Computex keynote, Jensen Huang moved past standard product announcements to clearly define the structural shifts occurring within the AI industry. Agentic AI (Self-Executing AI): The evolution of AI from a tool that merely answers questions to an autonomous ecosystem that makes decisions and executes objectives independently without human intervention. Unveiling the Next-Gen "Rubin" Platform: By introducing the Rubin architecture—the successor to Blackwell—NVIDIA signaled that the generational replacement cycle for AI infrastructure is accelerating. The Concept of the “AI Factory”: A paradigm shift that views AI not as a mere software tool, but as a "manufacturing infrastructure" that treats data as a raw material to pump out value. ◆ Components of an AI Factory AI Factory Component Traditional Manufacturing Equivalent Practical Role GPU(Graphics Processing Unit) Production Equipment (Machinery) The core hardware that generates value (computation). Data Center Factory (Infrastructure) The physical space where equipment operates and receives power. AI Model Final Product The output generated through factory operations. This shift demonstrates that the core of AI competition has migrated from a software race of "designing better models" to a physical infrastructure supply chain race of "who can physically secure overwhelming computational capacity." 2. Where the True Bottlenecks Lie: “Packaging, Memory, and Talent” Currently, the AI semiconductor supply chain is highly centralized. Design is driven by NVIDIA (USA), foundry and advanced packaging are handled by TSMC (Taiwan), and core memory is supplied by South Korean HBM manufacturers. Within this loop, three primary bottlenecks currently restrict actual market supply: ① The Limits of CoWoS (Chip on Wafer on Substrate) Packaging Concept: TSMC’s proprietary, advanced 2.5D back-end packaging technology that precisely processes and integrates chips and High Bandwidth Memory onto a fine substrate. Reality: As physical scaling reaches its limits in front-end processes, the importance of back-end packaging has skyrocketed. However, TSMC’s CoWoS capacity is already fully booked by major tech clients through long-term contracts. Securing this capacity has become synonymous with market dominance. ② HBM(High Bandwidth Memory) An indispensable AI memory technology that stacks DRAM vertically to maximize data transfer speeds. To overcome the physical limits of hardware, close-knit collaboration with South Korea’s two memory giants is no longer optional—it is a baseline requirement. ③ The Invisible Variable: Human Skill While semiconductor manufacturing is shifting heavily toward advanced robotics and automation, the experience and skill level of on-site engineers have paradoxically become more critical than ever. Minimizing errors, boosting early-stage yields in sub-micro processes, and maintaining extreme precision in complex packaging ultimately depend on human institutional knowledge rather than the machinery itself. Leading global fabs are currently grappling with a shared obstacle: a severe shortage of highly skilled technicians to support equipment operations. 3. The Meaning Behind the Korea Visit: Structural Locking and 'Lock-in' Strategies Jensen Huang’s visit to South Korea immediately following his trip to Taiwan cannot be interpreted merely as a routine procurement business trip for HBM. Institutionalizing Alliances: NVIDIA must actively counter big tech companies att