NVIDIA secures memory supply from SK Hynix as part of a $500 billion AI deal

NVIDIA announced that it has secured AI memory supply from South Korea's SK Hynix, as the chip manufacturer seeks to lock in key components for its advanced processors and systems.

NVIDIA stated that the agreement announced on Friday night in San Francisco could be worth $500 billion over several years and includes the construction of a large data center expected to go online in 2027.

SK Hynix's affiliate SK Telecom will use NVIDIA's Vera Rubin system to build its cloud business. NVIDIA indicated that its target capacity requires 2 gigawatts of power, suggesting this is a large-scale construction project involving hundreds of thousands of graphics processors.

In a separate deal, South Korean giant Samsung Electronics announced that it has signed a memorandum of understanding with chip design company Broadcom to expand cooperation in memory and foundry technology. Samsung stated in a release that this approximately $200 billion deal will help support next-generation AI infrastructure.

As the AI boom causes a global memory shortage, NVIDIA is actively securing supplies of high-bandwidth memory, which is crucial for its GPUs and systems. Analysts say that SK Hynix is a leader in HBM production.

"This expansion will include joint development opportunities on our next-generation SK Hynix AI memory, which will help us ensure a stable supply of HBM memory," said Rajeev Merpuri, NVIDIA's corporate vice president, during a press conference.

The agreement also indicates that large-scale AI infrastructure construction is extending beyond a few hyperscale companies, with foreign governments and large corporate groups beginning to get involved.

The agreement was announced at an AI summit in San Francisco, attended by South Korean officials including President Yoon Suk-yeol.

NVIDIA also announced on Friday that it will invest $1 billion in Naver, a South Korean cloud company building data centers around its GPUs. NVIDIA stated that the project will provide opportunities for potential customers in Korea and around the world to gain AI computing capabilities before the completion of the 200-megawatt capacity