Key Highlights

  • NVIDIA secured multi-party deals in Seoul spanning memory, sovereign cloud, and physical AI, reinforcing its Supply chain depth.
  • SK Hynix signed a multi-year technology Partnership to co-develop advanced memory for NVIDIA's full AI infrastructure roadmap, including robotics and personal AI platforms.
  • SK Telecom will build a gigawatt-scale AI Cloud using the NVIDIA DSX platform, with the first Facility targeted for 2027.
  • NAVER plans to expand sovereign AI infrastructure starting at 55 megawatts and scaling to gigawatt capacity, focused on enterprise and government workloads.

A Strategic Pivot, Not a Sales Tour

What Jensen Huang's Seoul visit produced was not a bundle of procurement contracts. It was a reconfiguration of how NVIDIA positions itself across the AI infrastructure value chain, from memory co-development at the chip level to sovereign cloud architecture at the national level.

The timing is deliberate. As AI workloads scale from experimentation to production-grade deployment, Demand is shifting toward integrated, full-stack solutions rather than Commodity GPU sales. Korea, home to global leaders in memory, semiconductors, Manufacturing, and robotics, is a natural anchor point for that transition.

Memory as Strategic Infrastructure

The most structurally significant agreement is between NVIDIA and SK Hynix. Under the multi-year technology partnership, SK Hynix will co-develop next-generation memory aligned directly to NVIDIA's product roadmap, spanning Vera Rubin AI supercomputers, Vera CPUs, RTX Spark-powered personal computers, and Jetson Thor robotic computing platforms.

This is a meaningful shift in how the memory Business works. SK Hynix has historically supplied advanced HBM at scale, but co-development at the silicon level with a defined roadmap is a deeper commitment. It converts a supplier relationship into a design partnership, with supply commitments structured to match NVIDIA's expanding infrastructure cycle rather than the open market.

Huang confirmed that NVIDIA already procures billions of dollars annually from SK Hynix, and called the deal extendable beyond its initial term. The strategic intent is to ensure memory supply does not become a constraint as AI factories scale globally.

Beyond supply security, SK hynix will apply NVIDIA's CUDA-X libraries and PhysicsNeMo framework to accelerate semiconductor simulation and computer-aided design workflows internally, while advancing digital twin capabilities for autonomous fab operations using Omniverse and cuOpt.

Korea's Sovereign Cloud Takes Shape

SK Telecom's announcement adds a different dimension: national AI infrastructure. The carrier will build a gigawatt-scale AI Cloud using the NVIDIA DSX full-stack architecture, with the first AI factory expected online in 2027. The initiative spans Training, inference, and agentic AI workloads, serving enterprises and industries across Korea with a stated ambition to expand into broader Asian markets.

NAVER's participation runs parallel. The internet and cloud group will expand its GAK Sejong data centre starting at 55 megawatts, with a trajectory toward gigawatt scale on the NVIDIA DSX platform. NAVER is also advancing HyperCLOVA X model development using NVIDIA Nemotron architecture and has become the first Korean company admitted to the NVIDIA Nemotron Coalition.

Both deals illustrate how sovereign AI is maturing from a policy concept into Capital-intensive infrastructure commitments. The NVIDIA DSX platform, which delivers what the company calls its lowest-cost token architecture through the MaxLPS software layer, is positioned as the blueprint rather than just the hardware.

The Macro Complication

The market reaction on June 8 complicated the narrative. South Korea's Kospi index fell 8.3%, its sharpest single-session drop in recent months, after U.S. non-farm payrolls data came in stronger than expected, reviving fears of a Federal Reserve rate hike. SK Hynix shares closed down 7.68%. Samsung Electronics declined 10.2%.

The selloff was not company-specific. It reflected a repricing of risk across the global technology sector as rate expectations shifted. Huang, when asked about the decline, framed it as a buying opportunity, pointing to NVIDIA's AI trajectory.

Whether that assessment proves correct depends on how quickly AI factory demand materialises at the scale these agreements envision. The deals signed in Seoul represent intent and architecture. Converting them into Revenue is a function of execution timelines, energy availability, and whether enterprise adoption of agentic AI accelerates at the pace NVIDIA's roadmap assumes.

The structural case remains intact. But the market was, at least for a session, more focused on the Cost of Capital than the promise of artificial intelligence.