Nvidia secures $500B memory deal with SK Hynix, targets 2027 data center launch

Nvidia announced late Friday in San Francisco that it has secured a multi-year AI memory supply agreement with South Korea's SK Hynix, valued at $500 billion（約80兆円）. The deal includes construction of large-scale data centers expected to come online in 2027. SK Hynix affiliate SK Telecom will build a cloud business using Nvidia's Vera Rubin systems. The planned infrastructure requires 2 gigawatts of power, indicating a buildout with hundreds of thousands of graphics processing units.

Raj Mirpuri, Nvidia's enterprise vice president, explained the strategic rationale on a call with reporters: "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." Nvidia is aggressively pursuing high-bandwidth memory supply because this component is essential for its GPUs and systems, and the AI boom has created a global memory shortage. SK Hynix holds the leading position in HBM production, according to analysts.

The announcement reflects a broader shift in AI infrastructure investment. The deal signals that massive data center buildouts are moving beyond a handful of hyperscalers, with foreign governments and major conglomerates now directly participating. The agreement was announced at an AI summit in San Francisco attended by South Korean officials, including President Lee Jae Myung. In a separate but related move, Samsung Electronics signed a memorandum of understanding with chip designer Broadcom, estimated at $200 billion（約32兆円）, to expand collaboration across memory and foundry technologies for next-generation AI infrastructure.

Nvidia also committed $1 billion（約1600億円） to invest in Naver, a Korean cloud company building data centers around Nvidia's GPUs. This investment will provide customers in South Korea and globally an opportunity to secure AI computing capacity before the 200 megawatts of capacity is completed. SK Hynix, South Korea's second most valuable company, listed on the Nasdaq earlier this month as part of an effort to finance infrastructure developments, positioning itself to support the capital requirements of these large-scale projects.

Nvidia's $500 billion（約80兆円） commitment to SK Hynix represents a strategic shift in how AI infrastructure is being financed and built globally. Rather than relying on a small number of hyperscalers to absorb demand for chips and computing power, foreign governments and major conglomerates are now stepping directly into the market. The announcement came at an AI summit in San Francisco with South Korean President Lee Jae Myung present, underscoring the geopolitical dimension of AI infrastructure investment.

The deal's scale—requiring 2 gigawatts of power and hundreds of thousands of GPUs—reflects the massive computational demands of current AI systems. By securing long-term high-bandwidth memory supply through a co-development agreement with SK Hynix, Nvidia is hedging against the global memory shortage that has emerged as a bottleneck in AI infrastructure expansion. SK Hynix's recent Nasdaq listing earlier in the month was explicitly framed as a financing mechanism for infrastructure developments, making the timing of this announcement strategically aligned with capital-raising efforts.

The parallel $200 billion（約32兆円） Samsung–Broadcom memorandum of understanding on memory and foundry technologies suggests that South Korea is positioning itself as a critical hub for AI infrastructure components. Nvidia's additional $1 billion（約1600億円） investment in Naver to develop cloud data centers further reinforces this positioning, providing customers in South Korea and globally access to AI computing capacity before the full 200 megawatts of planned capacity is completed.

No comments yet. Be the first to share your thoughts!