Nvidia CEO Jensen Huang met SK Group Chairman Chey Tae-won in Taipei on June 1, according to a post on SK Group's Facebook account reported by Yonhap. The meeting, held on the sidelines of Computex, covered cooperation on artificial intelligence memory and marked a celebration of SK hynix reaching a market capitalization of US$1 trillion, the Facebook post said, quoted by Yonhap. Yonhap also reports Huang visited SK hynix's booth, inspected next-generation high-bandwidth memory products, and left handwritten notes on sample wafers reading "Please make more" and "Love SOCAMM." Yonhap reports that SK hynix supplies Nvidia with HBM4 and that Nvidia has begun mass production of the Vera Rubin AI computing platform equipped with HBM4. Editorial analysis: Industry observers note that senior executive meetings like this reflect the tight supplier integraton around HBM and the premium placed on memory capacity for large AI models.
What happened
Jensen Huang, chief executive officer of Nvidia Corp., met with SK Group Chairman Chey Tae-won in Taipei on June 1, according to a post on SK Group's Facebook account cited by Yonhap. The Facebook post, as reported by Yonhap, said the executives met to mark SK hynix reaching a market capitalization of US$1 trillion and to discuss future cooperation in artificial intelligence memory. Yonhap reports Huang visited SK hynix's Computex booth to review next-generation high-bandwidth memory samples and left handwritten notes on wafers, writing "Please make more" on an HBM4E sample and "Love SOCAMM" on a SOCAMM module.
Technical details
Per Yonhap, SK hynix supplies HBM4 to Nvidia, and Yonhap reports Huang confirmed that Nvidia has begun mass production of the Vera Rubin AI computing platform, which is equipped with HBM4. Industry-pattern observations: High-bandwidth memory such as HBM4 is a concentrated critical input for large-scale GPU systems because it combines high throughput with lower power per bit, a factor that materially affects AI data center throughput and model scaling.
Industry context
Editorial analysis: Companies building large AI accelerators and data centers have increasingly limited vendor options for next-generation HBM, creating tight supplier relationships between GPU designers and memory manufacturers. Observers following the sector note that executive-level coordination at trade shows is a common mechanism to align product timing, validate samples, and highlight marquee partnerships to investors and customers.
Related events reported
Yonhap, citing industry sources, reports Huang plans a trip to Seoul later in the week with scheduled meetings with major South Korean conglomerate leaders, including LG Group Chairman Koo Kwang-mo and Naver founder Lee Hae-jin, and potential participation by Hyundai Motor Group Executive Chair Euisun Chung. Yonhap also reports a planned closed-door meeting with local robotics and AI startups and a meeting with Kim Taek-jin, CEO of NC Corp.
What to watch
For practitioners: monitor the following indicators to judge downstream implications:
- •Announcements from SK hynix orNvidia about HBM4 supply agreements, capacity expansions, or qualification milestones. - •Capacity and pricing signals in the memory market, especially for HBM4 and advanced module supply. - •Productization timelines for large-scale systems using Vera Rubin, since system ship schedules drive near-term demand for HBM.
- •Any formal statements or filings from the companies that detail supply commitments or joint development programs.
Editorial analysis: For engineers and infrastructure planners, these kinds of supplier-level alignments tend to affect procurement lead times and benchmarking expectations for new AI hardware generations, which in turn shape model deployment timelines and cost assumptions.
Scoring Rationale #
This is a notable infrastructure story because HBM supply is a constraining input for large AI systems and the meeting signals continued coordination between a major GPU vendor and a memory supplier. The story is company-level and relevant for capacity planning, but it does not by itself announce new contracts or capacity expansions.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.