Jensen Huang Expands AI Full Stack, Tests Samsung and SK Hynix's Memory Strategies

By SEONGJUN JO Posted : June 2, 2026, 16:57 Updated : June 2, 2026, 16:57
Choi Tae-won, Chairman of SK Group, meets with NVIDIA CEO Jensen Huang in Taipei, Taiwan, marking their third meeting this year. [Photo: SK Hynix]

Jensen Huang, CEO of NVIDIA, is demonstrating a keen interest in the South Korean semiconductor ecosystem while on a busy schedule in Taiwan. As the AI semiconductor market shifts towards NVIDIA's leadership, the strategic importance of South Korean memory companies like Samsung Electronics and SK Hynix is becoming increasingly prominent.

On June 2, industry sources reported that Huang hosted a "Korea Partner Night" event in Taipei the previous evening. When asked by reporters about the possibility of holding GTC Seoul, he indicated that it could happen. He also expressed a desire to invest in South Korea's robotics sector, highlighting Taiwan's role as a key hub in the AI supply chain while sending a clear message of collaboration with South Korean firms.

That same evening, Huang met separately with Choi Tae-won and SK Hynix executives, showcasing their camaraderie by putting their arms around each other.

Huang's ongoing engagement with South Korean companies is driven by the growing importance of memory in NVIDIA's AI ecosystem. The performance of AI accelerators relies not only on graphics processing units (GPUs) but also on the supply capabilities of high-bandwidth memory (HBM), low-power DRAM, and high-performance storage. As demand for AI computing expands from data centers to AI PCs, robotics, and autonomous vehicles, the role of South Korean memory companies is also broadening.
 
Song Jae-hyuk, CTO of Samsung Electronics' Semiconductor Division, explains the world's first 8th generation high-bandwidth memory (HBM5) prototype at Computex 2026 in Taipei. [Photo: Samsung Electronics]

On the same day, Samsung Electronics unveiled its next-generation HBM technology at Computex 2026 in Taipei. Song Jae-hyuk, CTO of Samsung's Semiconductor Division, introduced the HBM5 prototype for the first time and presented a new heat management technology called "HPB (Heat Path Block)."

He stated, "As AI systems evolve into ultra-high-performance and ultra-dense architectures, not only memory performance but also data processing efficiency and heat control are becoming key competitive factors." Samsung aims to enhance operational stability by adding a separate heat transfer path to HBM5, thereby reducing thermal resistance.

SK Hynix is recognized as having the closest HBM partnership with NVIDIA. Choi's unique leadership style has facilitated their third meeting this year, solidifying a strong trust relationship. Recently, SK Hynix also introduced a next-generation heat management technology called iHBM, aimed at enhancing the competitiveness of high-density AI memory.

Kim Yong-seok, a distinguished professor at Gachon University’s Semiconductor College, noted, "There are two main reasons for Jensen Huang's visits to Korea. First, he seeks product supplies from Samsung and SK Hynix for NVIDIA's chips and software solutions. Second, he aims to secure stable access to memory, including HBM and DRAM, necessary for building AI infrastructure like smart factories."



* This article has been translated by AI.

Copyright ⓒ Aju Press All rights reserved.