Full Breakdown
Nvidia Secures $500 Billion AI Memory Deal with SK Hynix
7/25/2026, 10:49:08 AM
Core Event: Multi-Billion-Dollar AI Memory Partnership
Nvidia Corp. announced a partnership with South Korea’s SK Hynix that will lock in more than $500 billion in high-bandwidth memory (HBM) supply for Nvidia’s AI GPUs. The agreement, disclosed at an AI summit in San Francisco, also includes plans to build AI-focused data centers powered by Nvidia’s Vera Rubin chips and SK Hynix’s forthcoming HBM4 modules. The first 2-gigawatt facility is slated to become operational in 2027.
Background & Context
The deal follows a broader wave of South Korean AI initiatives announced during President Lee Jae Myung’s visit to Silicon Valley. South Korea has pledged roughly $950 billion in AI projects, with SK Group committing about $750 billion to partnerships that span memory, data-center construction, and advanced packaging. The Nvidia-SK Hynix pact is the largest single component of these efforts, reflecting the global shortage of HBM needed for AI training and inference workloads.
Data & Statistics
- Deal value: > $500 billion (Nvidia statement)
- Power requirement: 2 gigawatts for the planned AI data center
- Capacity timeline: First facility online in 2027
- Memory share: SK Hynix controls roughly 50 % of global HBM production
- Projected demand: SK Group chairman Chey Tae-won warned Nvidia’s five-year memory needs could exceed current forecasts
Official Statements & Responses
- “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,” — Raj Mirpuri, Nvidia enterprise vice president
- “I wouldn’t be surprised if, the next time I see (Huang), he’ll tell me those figures were too low and say, ‘I want more’,” — Chey Tae-won, SK Group chairman
- Nvidia’s press releases note that forward-looking statements are subject to risks and uncertainties, including market acceptance and technological development.
Conflicting Reports & Gaps
- Deal valuation: CNBC and Nvidia press materials cite a $500 billion figure, while Reuters references a broader $950 billion AI initiative and a $750 billion SK Group commitment. Bloomberg reported “very large” sums but declined to disclose a specific amount.
- Scope of partnerships: Reuters details separate memoranda with Broadcom and Samsung, whereas CNBC focuses on the Nvidia-SK Hynix memory component and a $1 billion investment in Korean cloud firm Naver.
- Future capacity: Nvidia targets “200 megawatts of capacity” to be completed, yet the data-center plan mentions a 2-gigawatt power draw, leaving the relationship between these figures unclear.
What’s Next
- The 2-gigawatt AI data center, built by SK Telecom, is expected to be operational in 2027.
- President Lee will attend a Silicon Valley meeting with South Korea’s National Pension Service and U.S. venture-capital firms to promote investment in Korean AI startups.
- Ongoing co-development of HBM4 and future HBM variants is anticipated, though timelines remain subject to the forward-looking cautions noted by Nvidia.
