Samsung Electronics and SK Hynix, along with other South Korean companies strong in autonomous driving and physical AI, are increasingly joining NVIDIA's AI supply chain. This raises concerns about potential cracks in the existing TSMC-centered AI supply chain.
On June 1, NVIDIA CEO Jensen Huang began a series of meetings with major South Korean companies during his keynote speech at the 'GTC Taipei 2026' event. The inaugural 'Korea Partner Night' featured representatives from Samsung Electronics, SK Hynix, Hyundai Motor Group, LG Electronics, Doosan, and Naver.
NVIDIA's expansion of collaboration with South Korean firms is driven by their advanced manufacturing capabilities. The core competitiveness of AI accelerators is shifting from single GPU performance to include memory bandwidth, power efficiency, advanced packaging, data center operations, and physical AI implementation. To ensure a stable supply of next-generation AI accelerators, NVIDIA needs partners not only for memory supply but also in robotics and mobility sectors. Relying solely on a foundry-based value chain like TSMC's is no longer feasible.
The surge in AI chip production has exacerbated supply bottlenecks related to TSMC's proprietary packaging technology, CoWoS (Chip on Wafer on Substrate). While TSMC's foundry capabilities remain dominant, it has become increasingly difficult to manage the entire supply chain independently in the AI era. Analysts predict a multipolar system where NVIDIA handles GPU design, Taiwan focuses on advanced manufacturing, and South Korea takes charge of memory, packaging, and post-processing.
Samsung and SK Hynix, which have secured leadership in HBM supply, are expected to gain further prominence. Samsung recently began shipping samples of its seventh-generation HBM4E, accelerating its response to NVIDIA's next-generation roadmap. HBM4E is set to be integrated into NVIDIA's upcoming AI accelerator, Rubin Ultra, scheduled for release in 2027.
SK Group Chairman Chey Tae-won and CEO Jensen Huang have met three times over the past seven months in South Korea and the U.S., and they met again during the GTC event in Taiwan. There is growing interest in the potential for collaboration on next-generation HBM supply and AI memory roadmaps.
NVIDIA is also strengthening its collaboration with Hyundai Motor Group on autonomous driving. At the GTC conference held in the U.S. in March, NVIDIA announced plans to work with Hyundai on developing Level 4 autonomous driving. Hyundai is building an integrated system that can expand from Level 2 to Level 4 autonomy using NVIDIA's Drive platform, Hyperion.
With LG Group, NVIDIA is accelerating its collaboration in physical AI across various sectors. Collaborations are anticipated with LG AI Research, which is developing a large language model (LLM) called Exaone, LG Innotek, which specializes in robotic sensing technology, and LG Uplus, which has cloud infrastructure including data centers. Earlier this year, LG Electronics unveiled the intelligent home robot 'LG Cloid,' based on NVIDIA's high-performance chipset, Jetson Thor.
An industry insider noted, "The AI semiconductor supply chain is evolving from being centered around NVIDIA and TSMC to a 'physical AI manufacturing ecosystem' that includes memory, packaging, and robotics. South Korean companies are emerging as key players in the AI ecosystem due to their core capabilities in physical AI."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.