On June 5, the South Korean equity market suffered a severe correction, with the KOSPI index closing down 5.54%. By June 8, intraday losses expanded beyond 8%, triggering exchange circuit breakers as major semiconductor firms Samsung and SK Hynix declined nearly 10%. Amidst this volatility, NVIDIA CEO Jensen Huang's visit to Seoul functioned as a critical market stabilizer. On the evening of June 7, local time, Huang convened a dinner with SK Group Chairman Chey Tae-won and SK Hynix CEO Lee Seok-hee. Post-meeting, Huang confirmed that NVIDIA's newly launched Vera CPU would utilize SK Hynix DRAM, with both entities preparing for large-scale cooperation extending through the second half of the current year and into the next. Data compiled by Woofun AI shows that Huang further projected the current memory chip shortage to persist for several years, countering immediate expectations of supply normalization.

NVIDIA and SK Hynix subsequently formalized a multi-year technical cooperation agreement encompassing AI supercomputers, robotics, digital twins, and semiconductor manufacturing. The Vera CPU represents NVIDIA's first dedicated data center processor, entering a competitive landscape against Intel's Xeon line, AMD's Epyc chips, and custom silicon from cloud providers like Amazon Graviton. From the outset, NVIDIA has anchored its memory supply chain with SK Hynix for this new product line. The partnership covers diverse product categories, including the Vera Rubin AI supercomputer, PCs equipped with RTX Spark, and the Jetson Thor robotic computing platform. The strategic objective is to secure advanced memory supplies against long development cycles, complex manufacturing processes, and high capital expenditures required for global AI factory construction.

Beyond memory supply, the collaboration integrates NVIDIA's AI technology directly into SK Hynix's chip design and manufacturing workflows. SK Hynix is deploying NVIDIA's CUDA-X libraries and AI capabilities to accelerate semiconductor simulations, specifically targeting technical computer-aided design (TCAD) and computational lithography. Woofun AI notes that this initiative extends beyond internal optimization, aiming to establish a trilateral cooperation model involving chip manufacturers, NVIDIA, and EDA software suppliers. In the manufacturing domain, SK Hynix is advancing digital twin capabilities for its wafer fabs using the NVIDIA Omniverse platform. By leveraging the Omniverse library and OpenUSD processes, the company can construct 3D factory scenes to visualize, simulate, and optimize complex semiconductor environments.

At the operational level, these digital twin capabilities integrate with NVIDIA's cuOpt decision optimization engine and Metropolis platform to schedule autonomous mobile robots and other assets within the wafer fab. The companies are also exploring methods to merge digital twins with traditional software and AI workflows, enabling AI systems to reason and automatically execute tasks to improve manufacturing decisions. This technological integration marks a shift from simple component supply to a deep symbiosis in production methodology, potentially redefining efficiency standards across the semiconductor industry.

The groundwork for this infrastructure was laid six months prior, in October 2025, when NVIDIA and SK Hynix announced a large-scale infrastructure collaboration. SK Group is constructing an AI factory equipped with over 50,000 NVIDIA GPUs, with the first phase targeted for completion by the end of 2027. Upon completion, this facility is expected to rank among the largest AI factories in South Korea. The project operates on a 'GPU as a service' model, accessible to SK Group subsidiaries and external organizations to accelerate digital transformation. SK Telecom, acting as NVIDIA's cloud partner, is deploying over 2,000 NVIDIA RTX PRO 6000 Blackwell server GPUs to build an industrial AI cloud in Asia, specifically running Omniverse workloads to support SK Hynix's manufacturing and internal AI agents.

Despite the deepened partnership with SK Hynix, NVIDIA maintains a diversified supply strategy for HBM4 memory. Upon arriving in Seoul, Huang clarified to reporters that all three qualified suppliers—Samsung Electronics, SK Hynix, and Micron Technology—are in production and competing to support the Vera Rubin system. In his keynote at the Taipei Computer Show, Huang confirmed that Vera Rubin is in full production with delivery scheduled for the third quarter of the current year. Each server rack system pairs the Vera CPU and Rubin graphics core cluster with TB-level HBM4 memory. Woofun AI analysis suggests that while SK Hynix currently leads in HBM4 progress, having completed internal certification and established a production system for customers, the competitive landscape remains fluid. Meritz Securities senior analyst Kim Sunwoo previously predicted that SK Hynix's HBM market share would remain just above 60% in 2026 due to early supply advantages.

The involvement of three suppliers does not alleviate supply pressure, as Huang provided a sobering assessment regarding the duration of the shortage. He stated that the entire industry supply chain, from wafers to packaging and silicon photonics, faces deficits due to unprecedented demand, a situation expected to last for years. This scarcity is driven by the near-unending consumption of advanced memory required for global AI factory construction. The launch of Vera Rubin, the promotion of AI factories, and the expansion into personal and physical AI sectors are collectively driving memory demand higher. Consequently, all three HBM4 suppliers are under intense pressure to support Vera Rubin, as no entity wishes to fall behind in a constrained market.

Huang's agenda in South Korea extended beyond SK Group, reflecting a systematic effort to deepen ties with the entire regional tech ecosystem. He arranged meetings with Hyundai Motor, LG, SK, Samsung, and Naver, while also disclosing active recruitment for a new NVIDIA R&D center in the country. These movements indicate a comprehensive strategy to embed NVIDIA's technology stack across South Korea's industrial base. As the market digests these developments, the focus shifts from immediate stock price fluctuations to the long-term structural implications of a multi-year memory shortage and the accelerating integration of AI into semiconductor manufacturing.