Nvidia Forms $500 Billion AI Alliance with SK

Long-term HBM Supply Agreement with SK hynix

NVIDIA has announced that it will jointly develop next-generation artificial intelligence (AI) memory with SK hynix and will secure a long-term supply agreement. This move is interpreted as a proactive effort to secure key supply chains amid an ongoing shortage of high-bandwidth memory (HBM) caused by surging demand for AI semiconductors.

Chairman Chey Tae-won and CEO Jensen Huang. Photo by Yonhap News.

According to CNBC on the 26th (local time), NVIDIA and the SK Group announced a strategic partnership worth over USD 500 billion (approximately KRW 720 trillion) in San Francisco, United States. This partnership includes AI data center construction and next-generation memory collaboration.

The scope of this project includes not only the supply of memory chips by SK hynix, but also SK Telecom's plan to build a large-scale AI data center. SK Telecom intends to establish a total 2-gigawatt (GW) scale AI cloud business based on NVIDIA's next-generation "Vera Rubin" system, with the first facility slated to begin operations in 2027.

2 GW is a vast electricity capacity, sufficient to run hundreds of thousands of graphics processing units (GPUs). This case is seen as an example of AI infrastructure investment—which has so far been led by a handful of major U.S. tech companies—spreading to foreign governments and large corporate conglomerates.

SK hynix and NVIDIA have agreed to jointly develop next-generation AI memory, including HBM, and to establish a long-term supply framework. For NVIDIA, this will ensure a stable supply of HBM, which is critical for AI accelerator performance; for SK hynix, it means participation in memory design from the early stages of next-generation NVIDIA platform development.

As the AI boom drives a surge in demand for GPUs, HBM has emerged as the key bottleneck in the global semiconductor supply chain. SK hynix is leading the HBM market and has positioned itself as a major supplier to NVIDIA.

Samsung Electronics also signed a memorandum of understanding (MOU) with U.S. semiconductor company Broadcom on the same day, to expand cooperation in both memory and foundry sectors. The two companies plan to jointly develop and supply memory and contract manufacturing technology necessary for next-generation AI infrastructure.

Samsung Electronics expects the scale of this cooperation to exceed USD 200 billion over five years until 2030. Since the cooperation includes not only memory but also foundry, it is seen as Samsung's effort to expand orders across the entire AI semiconductor supply chain.

NVIDIA has also decided to invest USD 1 billion in Naver. This investment is aimed at participating in Naver's domestic AI data center project, which is being pursued together with global asset manager Brookfield. Naver plans to initially build AI infrastructure with a capacity of 55 megawatts (MW), and later expand it to the gigawatt level.

This content was produced with the assistance of AI translation services.