SK Group Chairman Choi Tae-won and NVIDIA CEO Jensen Huang held a "family-style" meeting at a fried chicken restaurant in Silicon Valley, with both of their daughters in attendance. Analysts believe that the discussions went beyond the supply of HBM4 and included strategic cooperation in building next-generation AI data centers. Industry insiders consider this meeting a potential major restructuring opportunity for the global semiconductor industry chain, following the "AI Triangle Alliance" in 2021, with SK leveraging HBM as a key tool to formally enter the next-generation AI infrastructure market.
In a fried chicken shop in Silicon Valley, there may not be exquisite tableware, but it houses hundreds of billions of dollars' worth of business.
According to South Korean media reports, on February 5 local time, Chairman Tae-won Choi of the SK Group and Jensen Huang, CEO of NVIDIA, held an informal 'Chimaek (chicken and beer) talk' at a Korean-style fried chicken restaurant called '99 Chicken' in Silicon Valley during Tae-won Choi’s visit to the United States.
Observers noted that the discussion might not only involve negotiations over the supply volume of the sixth-generation high bandwidth memory (HBM4) but also cover strategic cooperation in building next-generation AI data centers.
Notably, this meeting was reported to have been far more private and intimate than ordinary business negotiations. In addition to key executives from SK Hynix, attendees included Tae-won Choi’s second daughter Min-jung Choi, representative of IntegralHealth, and Madison Huang, senior director of NVIDIA's robotics division and daughter of Jensen Huang.
Analysts believe that this 'family-oriented' social setting sends a strong signal: the relationship between SK and NVIDIA is upgrading from business partners to closer strategic allies.
Reviewing history, every meeting between Tae-won Choi and Jensen Huang has triggered significant waves in the semiconductor market. Their meeting in May 2021 directly led to the formation of the 'AI Triangular Alliance' among SK, NVIDIA, and Taiwan Semiconductor.
This time, industry insiders believe that Tae-won Choi's visit is not only about ensuring the timely delivery of HBM4 but also focuses on the strategic layout for SK Group’s transformation into a 'comprehensive AI solutions provider.' A semiconductor industry insider stated:
This is a signal that SK Group is officially entering the next-generation AI infrastructure market using HBM as leverage.
The focus of the meeting was on guaranteeing the supply of HBM4. The next-generation AI accelerator 'Vera Rubin,' which NVIDIA will launch in the second half of the year, will use HBM4 with a capacity of 288 gigabytes per chip.
SK Hynix committed to seamless delivery of HBM4 supplies.
Given that HBM production requires approximately four months, plus an additional two to three months for packaging by Taiwan Semiconductor, the entire cycle lasts six to seven months, deepening NVIDIA's reliance on SK Hynix, which holds the largest production capacity in the industry.
At the end of last year, SK Hynix reached an agreement with NVIDIA to supply more than 55% of the demand for HBM4, and is currently conducting performance optimization. Choi Tae-won assured Jensen Huang during their meeting of "unimpeded supply."
According to reports, although SK Hynix’s HBM4 utilizes relatively older-generation processes such as 12nm wafer fabrication and 1b DRAM, its performance matches that of Samsung Electronics’ products, which adopt 4nm wafer fabrication and 10nm 6th-generation DRAM.
The market landscape has shifted this year. In September last year, Samsung Electronics passed NVIDIA’s quality testing for its 12-layer HBM3E product and has now taken the lead in entering the mass production and shipment phase for HBM4.
Samsung’s HBM4 achieves a transmission speed of 11.7Gb per second, surpassing NVIDIA’s requirement standard of 10 to 11Gb per second, thereby breaking SK Hynix’s near-monopoly position in the 12-layer HBM3E product last year.
Industry insiders anticipate that the two parties also discussed cooperation plans for the seventh-generation HBM (HBM4E), which will officially enter the market in 2027, as well as customized HBM (cHBM).
From "selling memory" to "selling AI infrastructure"
If ensuring unimpeded HBM4 supply represents "defending the territory," then Choi Tae-won's true ambition during this visit lies in "expanding the frontier," targeting the AI data center infrastructure market.
Choi Tae-won is driving SK Group’s transformation into a "comprehensive AI solutions provider," and his discussions with Jensen Huang are believed to encompass collaboration across multiple levels, including AI semiconductors, servers, and data centers.
Further cooperation between SK Hynix and NVIDIA in the enterprise-grade solid-state drive (eSSD) sector has drawn significant attention.
NVIDIA announced in January this year that it would apply a new memory solution, 'ICMS,' in Vera Rubin. The entire system will be equipped with 9,600 terabytes of eSSD, representing a 16-fold increase in demand compared to existing products. This provides a significant opportunity for SK Hynix to expand in the storage field.
Since February 3, Chairman Choi has been in the United States holding a series of meetings with several major technology companies, including NVIDIA and Meta.
It was reported that the two parties may also have discussed the next-generation server memory module following high-bandwidth memory - SoCamm (low-power DRAM module for servers) - as well as flash memory supply. Industry insiders pointed out:
SoCamm is the next battleground for transforming the power architecture of AI servers, and the SK Group is officially entering the next-generation AI infrastructure market with HBM as a strong lever.
SK Hynix has renamed its U.S. flash memory subsidiary Solidigm to 'AI Company,' making it a platform dedicated to SK Group's AI investment and solutions business.
The scope of business will not only cover SK Hynix's AI semiconductor operations but will also integrate SK Telecom’s AI technologies and solution capabilities, creating an end-to-end service chain from AI data center design to semiconductor and server delivery. There are even plans to advance a pilot project for a data center in North America within the year.
This meeting revealed signs that Choi Taewon may have introduced this strategic concept to NVIDIA. Industry observers believe that the two parties may have discussed specific plans for SK Group to provide comprehensive AI solutions to NVIDIA.
This informal meeting at 99 Chicken restaurant could mark the starting point of another major restructuring of the global semiconductor industry chain, following the 'AI Triangle Alliance' in 2021.
Editor/KOKO
