Inside Nvidia’s Korea Pivot: Jensen Huang Locks in HBM4 Supply and Deepens Seoul Ties at Computex 2026

For context, this dinner is the latest in a series of personal gestures by Huang toward his Korean partners. In February 2026, he made a rare personal appearance at a Korean fried-chicken restaurant in Silicon Valley, spending roughly two hours drinking soju with over 30 SK Hynix engineers while urging them to deliver HBM4 on schedule .

Vera Rubin in Full Production: The Memory Order Gets Real

Hours before the dinner, Huang used his GTC Taipei keynote—held alongside the Computex trade show—to erase any doubt about the status of Nvidia’s next-generation AI accelerator platform. “Vera Rubin is fully in production now,” Huang declared, confirming that the platform had moved from roadmap to volume manufacturing .

He then explicitly named the three winners of the HBM4 memory supply contract. “HBM4 memory is supplied by SK Hynix, Samsung Electronics, and Micron,” Huang said, listing the companies onstage . The public acknowledgment settled what had been a closely watched competition among memory makers. Industry estimates reported after the keynote suggest that SK Hynix holds the largest share of the HBM4 allocation, with rough estimates pointing to 60–70% of the initial volume, while Samsung has recovered to secure around 25–30% .

Huang noted that the supply-chain scale for Vera Rubin had doubled compared with the previous Grace Blackwell generation . The Vera Rubin platform, designed to power the next wave of AI agents and large-scale inference workloads, is now the biggest single consumer of advanced HBM memory on the market.

A Pledged Visit to Seoul and a Stock-Price Rally

Huang’s Asia tour does not end in Taiwan. During the same engagements, he stated that his next stop would be Seoul, a visit intended to further solidify the partnerships he had just celebrated . The announcement was significant because it comes less than a year after his first official trip to South Korea in over a decade, when he met with the leaders of Samsung and Hyundai Motor and famously declared at the APEC CEO Summit in Gyeongju, “Korea is the world’s top memory technology nation. I am 100% confident that Samsung Electronics and SK Hynix will be Nvidia’s long-term partners” .

The combination of the confirmed Vera Rubin production status, the HBM4 supply lock-in, and the symbolic courtship of the Korean partner dinner had an immediate market impact. News of the deepened cooperation and the expanded AI-and-robotics collaboration scope drove a rally in South Korean technology stocks on expectations of sustained, multi-year AI-fueled demand .

From a Surprise Booth Visit to a Strategic Alliance

The relationship that appears so choreographed in 2026 was visibly more spontaneous just one year earlier. At Computex 2025, Huang made an unannounced visit to SK Hynix’s booth on the show floor. Guided personally by Kim Joo-seon, the head of the chipmaker’s AI infrastructure division, the Nvidia CEO inspected an HBM4 sample and remarked, “So beautiful” . That visit came just eight days before reports emerged that SK Hynix was close to finalizing its HBM4 supply deal, with mass production planned for Q4 2025 at its nearly completed M15X fab in Cheongju .

At the APEC summit in October 2025, Huang already foreshadowed the long-term commitment, telling reporters, “We will be close partners in manufacturing HBM4, HBM5, HBM97, and more going forward,” and confirming that HBM4 samples were “working great” . Computex 2026 was the moment that long-term pledge turned into hard production contracts and a very public, very personal celebration of the partnership.