Nvidia has signed a multi-year agreement with SK Hynix for the supply of high-bandwidth memory (HBM), a critical component for AI accelerators. Nvidia says the cooperation could underpin infrastructure investments with a total value of up to $500 billion and includes the construction of large-scale data centers.
Timing and capacity
The planned data centers are expected to begin operations in 2027. Nvidia estimates the infrastructure will require roughly 2 gigawatts of electrical power, which implies the operation of several hundred thousand graphics processing units (GPUs).
SK Telecom to offer cloud services on Vera Rubin
As part of the arrangement, SK Telecom, which is part of the SK Hynix group, will create a cloud service built on Nvidia's Vera Rubin systems. Thus the deal covers not only memory supply but also the creation of a service and infrastructure ecosystem tailored to AI workloads.
Why HBM matters
HBM is one of the most important components of modern AI accelerators. The rapid expansion of artificial intelligence has produced global shortages of these memories, prompting leading chip designers to lock in manufacturing capacity years in advance. Industry analysts consider SK Hynix a technology leader in this market.
Additional moves in South Korea and the chip industry
The announcement also highlights that building AI infrastructure increasingly involves more than large U.S. tech companies: national governments and industrial conglomerates are treating compute capacity as a strategic investment.
Separately, Samsung Electronics and Broadcom signed a memorandum of intent to expand cooperation in memory technology and semiconductor manufacturing; the two companies estimate that this partnership could support roughly $200 billion in AI infrastructure development.
Nvidia also announced a $1 billion investment in the South Korean cloud provider Naver. That project aims to build a new data center capable of delivering about 200 megawatts of AI compute capacity for South Korean and international customers.
Implications
These announcements indicate that global AI investments are reaching very large scales: locking in memory capacity and building data centers are strategic steps to remain competitive. The SK Hynix–Nvidia cooperation and the Samsung–Broadcom initiative show that major memory and semiconductor manufacturers are positioning themselves to support next-generation AI systems.
This article is not investment advice.
