Nvidia locks in $500bn SK Hynix memory supply for AI buildout
Nvidia's massive multi-year deal with SK Hynix secures critical high-bandwidth memory for its next-generation systems, signaling that global AI infrastructure spending is expanding well beyond the traditional hyperscalers.
Nvidia has agreed to a multi-year deal worth up to $500 billion with South Korea's SK Hynix to secure high-bandwidth memory for its advanced artificial intelligence processors. The agreement, announced late Friday in San Francisco, encompasses the construction of large-scale data centers that are expected to come online in 2027.
Under the terms, SK Hynix affiliate SK Telecom will construct a dedicated cloud business utilizing Nvidia's upcoming Vera Rubin systems. The planned buildout is massive by industry standards, targeting a power capacity of 2 gigawatts. That level of energy consumption indicates the deployment of hundreds of thousands of graphics processing units, highlighting the immense capital required to sustain AI growth.
The partnership exposes the acute supply chain constraints currently limiting the AI sector. Surging demand has created a global shortage of high-bandwidth memory, making the component a critical bottleneck for data center construction. Nvidia is moving to neutralize this risk by tying its fortunes directly to SK Hynix, which analysts consider the undisputed leader in HBM production.
Securing volumes is only part of the strategy, as Nvidia is also positioning itself to influence future memory designs. "The expansion will include a co-develop opportunity for us on the next-generation SK Hynix AI memory, and this will help us secure a stable supply of HBM memory," said Raj Mirpuri, Nvidia enterprise vice president.
For market participants, the deal signals a decisive shift in the profile of AI infrastructure investors. Capital deployments of this magnitude are moving beyond the traditional handful of American hyperscalers. Instead, foreign governments and sprawling industrial conglomerates are increasingly bankrolling the physical layer of the AI economy. The presence of South Korean President Lee Jae Myung at the San Francisco summit underscored this sovereign-backed momentum.
Nvidia paired the SK Hynix announcement with a separate $1 billion investment in Naver, a Korean cloud company developing GPU-driven data centers. Once completed, that project will add 200 megawatts of computing capacity for clients in South Korea and internationally. Together, these agreements validate the strategic rationale behind SK Hynix's recent Nasdaq listing. South Korea's second-most valuable company explicitly cited infrastructure financing as the goal of that US share sale.
