Nvidia's South Korea Strategy: How Jensen Huang is Realigning the AI Supply Chain
At 7:10 p.m. Friday, global technology's center of gravity shifted to Hyeongnim Jeoyo, a Mapo-gu pork belly restaurant. Nvidia CEO Jensen Huang joined SK Group's Chey Tae-won, LG Group's Koo Kwang-mo, who arrived early at 6:52 p.m., and NAVER's Lee Hae-jin. They dined on samgyeopsal and soju, with Lee settling the bill via Naver Pay to the cheers of patrons.
For institutional investors, reading this as a photo opportunity is a catastrophic misread. Behind the casual BBQ lies the architecture of a $5 trillion enterprise securing its chokeholds. This was not a dinner; it was supply-chain diplomacy.
The Four Pillars of the Korean Pivot
Earlier at Gimpo Airport, Huang promised "a lot of surprises," confirming four explicit new businesses making Korea "very busy":
- Vera Rubin: Nvidia’s next-generation AI superchip platform (Vera CPU plus Rubin GPU), already in full production and ramping aggressively for H2 2026.
- Vera CPU: A purpose-built, standalone processor for agentic AI workloads.
- RTX Spark: Nvidia’s inaugural AI laptop superchip. Co-developed with Microsoft, it fuses a Blackwell GPU, Arm CPU, and 128GB unified memory for 1 petaflop of local inference, backed by Dell, HP, Lenovo, Asus, and MSI.
- Jetson Thor: An edge computing platform engineered for next-generation humanoid robots and physical AI.
Huang paired these deployments with a structural commitment: a new AI technology center in South Korea, actively recruiting AI and robotics engineers. Nvidia is fresh off a staggering Q1 FY2027—$81.6 billion revenue (up 85% YoY, including $75.2 billion in data center) and a $91 billion Q2 guide. The Vera Rubin ramp is the dominant global tech narrative, and South Korea is where it gets resourced.
From Supplier to Extension Node
To understand this shift, one must recognize a 15-year restructuring of compute economics. CPUs lost the growth node, GPUs became the accelerator layer, and today, High-Bandwidth Memory (HBM) is the ultimate gating input.
Historically, South Korea’s semiconductor supremacy rested on cyclical DRAM—a brutal, boom-and-bust commodity. The AI supercycle broke that physics. HBM is performance-critical, thermally complex, and co-designed alongside accelerator roadmaps. SK Hynix and Samsung no longer trade like memory companies; they are toll collectors on the global AI factory buildout.
The market is missing the final evolution. Korea is not becoming an AI hub by owning software. It is indispensable because Nvidia requires memory density, packaging, and robotics integration at a scale precluding arm's-length vendor relationships.
Huang is executing Apple's TSMC playbook, or oil majors' relationships with producing states. He is transforming indispensable suppliers into vested ecosystem stakeholders. By integrating Korean HBM, hardware, and engineering talent, Huang is repositioning South Korea from a memory supplier to a critical platform extension node. That is the structural epiphany.
Valuation Disconnects: Pricing the Perfect Trade
The equity market's reaction has created severe valuation disconnects. Investors are mistakenly applying identical multiple expansion to every company orbiting Huang's table.
The Control Plane: Nvidia trades at $213 (down 2.4%), commanding a $5.17 trillion cap at 22x forward earnings and 44.5x FCF. The market accurately prices it as infrastructure control. The fatal risk is customer digestion, not competition. The trade remains long on pullbacks, focusing strictly on hyperscaler capex.
The Memory Duality: SK Hynix (₩2,080,000, down 9.49% but up 856% YoY) is priced for perfection. At ₩1,627.7 trillion cap, 9.9x book, and 38.9x FCF, its 7.1x forward P/E relies on the heroic assumption that HBM permanently suspended cyclicality. The smarter asymmetrical play is Samsung (₩2,092.7 trillion cap, 6.1x forward P/E, 4.3x book). Despite execution stumbles, Samsung offers vastly superior catch-up convexity if HBM4 qualification deepens its Nvidia lock-in.
The Speculative Air Pocket: LG Electronics is the most dangerous long. Up 325% YoY at 61.2x trailing earnings (₩303,000), it sits 45.8% above its analyst target. The market is aggressively front-running its Jetson Thor robotics optionality without disclosed orders. If physical AI headlines do not rapidly translate to backlog, LG is an unwind trade. NAVER (₩255,500, down 4.49%, 21.3x P/E) holds a role in sovereign AI, but a dinner seat does not instantly rewrite economics.
The winning strategy is isolating true choke points and ignoring downstream noise.
not investment advice
Sources: https://blogs.nvidia.com/blog/korea-ecosystem-2026/