Jensen Huang Heads to Seoul This Week for Fourth Meeting with SK Group Chairman Chey Tae-won in Six Months
Miles Bennett
Nvidia CEO Jensen Huang arrives in Seoul Thursday for his fourth sit-down with SK Group chairman Chey Tae-won in six months, with the agenda expanding from AI memory into robotics and physical AI — and heads of four Korean conglomerates set to join.
Who is in the room, and what are they discussing?
Huang lands Thursday night and meets Chey on Friday, with an notably expanded guest list compared to previous sessions.
Confirmed attendees include LG chairman Koo Kwang-mo and Naver founder Lee Hae-jin; Hyundai Motor executive chair Euisun Chung is actively considering joining.
This means → the conversation has shifted from a bilateral Nvidia-SK affair to a multi-conglomerate summit spanning chips, internet, and autos — Nvidia's Korea footprint is stretching well beyond the supply chain.
Why is the agenda moving past chips?
Industry observers say the Seoul talks will likely extend into robotics and physical AI — the emerging field of embedding AI into real-world machines and systems.
In plain terms = Nvidia used to visit Korea to discuss "which memory goes with the GPU." Now it is discussing "using AI to drive cars, factories, and robots."
Huang has already flagged robotics as a potential investment topic for his Korea trips, and Hyundai's possible attendance underscores that signal.
They just met in Taipei — why meet again in Seoul?
On June 1 at Taipei Computex, Huang, Chey, and SK Hynix CEO Kwak Noh-jung held their third meeting this year.
That sit-down coincided with SK Hynix's market cap crossing the $1 trillion mark; both sides reviewed AI-memory milestones and reaffirmed their commitment to expanding into new AI-infrastructure domains.
This reflects a relationship that has evolved from "one annual contract meeting" into continuous strategic coordination — the frequency itself is the signal.
Four meetings in six months — what deal is being pushed?
Since a Silicon Valley meeting last February, the two have met at GTC San Jose in March, Taipei in June, and now Seoul — four times in six months, an extraordinary cadence by industry standards.
The market's core speculation: Nvidia's next-gen AI compute platform Vera Rubin, due in H2, will use sixth-generation HBM (HBM4), and the new supply contracts could carry prices up to 100% higher than before.
This means → if that price increase holds, SK Hynix's HBM revenue could roughly double — which explains why Chey is willing to meet four times in half a year. A contract of that size is worth showing up for personally.
Why does HBM matter so much?
In Taipei, Huang specifically outlined four critical factors for tackling the complexity of HBM — high-bandwidth memory, a technology that stacks multiple memory-chip layers to give GPUs ultra-fast data throughput — and stressed the importance of close collaboration with SK Hynix.
Nvidia has reportedly signed new HBM supply contracts with memory-chip makers including SK Hynix.
Put simply = no matter how powerful the GPU, it cannot run large AI models without matching high-speed memory. HBM is the bottleneck link in the AI compute chain — whoever can supply it reliably locks in Nvidia's next-generation platform.
Content is for reference only, not financial advice.