Nvidia SK Hynix deal secures AI memory in $500 billion buildout

The Nvidia SK Hynix deal gives the AI chipmaker access to a critical memory component at a time when supply is tight across the industry. For investors watching the AI trade, the key point is straightforward: Nvidia is trying to reduce a bottleneck that can limit how many advanced AI systems it can ship.

Nvidia said late Friday in San Francisco that it had secured AI memory supply from South Korea’s SK Hynix as part of an agreement that could be worth $500 billion over several years. Nvidia said the arrangement includes large data center projects expected to start operating in 2027.

The company said SK Telecom, an affiliate of SK Hynix, will build a cloud business using Nvidia’s Vera Rubin systems. Nvidia said it is targeting enough capacity to require 2 gigawatts of power, a level the company said points to a buildout involving hundreds of thousands of graphics processing units, or GPUs.

GPUs are the chips used to run the heavy parallel calculations behind modern AI models. Nvidia’s ability to sell complete AI systems depends not only on its own processors, but also on whether partners can supply enough memory, power, networking and data center capacity around them.

What is high-bandwidth memory and why does Nvidia need it?

High-bandwidth memory, often shortened to HBM, is a type of advanced memory that moves data quickly between processors and memory stacks. Nvidia uses HBM in its GPUs and AI systems because AI workloads need to process large amounts of data at high speed.

Nvidia has been working to secure HBM supply as the AI buildout has contributed to a global memory shortage. Analysts cited by CNBC describe SK Hynix as the leader in HBM production, which makes the South Korean company a key supplier for Nvidia’s next wave of AI infrastructure.

Raj Mirpuri, Nvidia’s enterprise vice president, told reporters that the expansion includes a chance for Nvidia to co-develop next-generation SK Hynix AI memory and will help Nvidia secure a stable HBM supply.

How does the deal fit into South Korea’s AI push?

The announcement came during an AI summit in San Francisco attended by South Korean officials, including President Lee Jae Myung. The deal shows how AI infrastructure spending is spreading beyond the largest U.S. cloud companies to include foreign governments, telecom groups and major industrial companies.

Nvidia also said Friday it would invest $1 billion in Naver, a Korean cloud company building data centers around Nvidia GPUs. Nvidia said the project could give potential customers in South Korea and abroad a way to secure AI computing capacity before the planned 200 megawatts of capacity is finished.

SK Hynix, South Korea’s second most valuable company, listed on Nasdaq earlier in July as part of an effort to fund infrastructure development, CNBC reported.

Samsung and Broadcom also expand AI chip ties

In a separate announcement, Samsung Electronics said it signed a memorandum of understanding with Broadcom to expand their work together across memory and foundry technologies. Samsung said the agreement is estimated at $200 billion and is intended to support the next generation of AI infrastructure.

For retail investors, the broader signal is that AI infrastructure spending is increasingly tied to supply contracts, memory access and power availability, not only headline demand for chips. Nvidia’s announcement puts HBM supply near the center of the next phase of AI data center expansion.

This story draws on original reporting from CNBC.