Nvidia's Vera Rubin Platform Enters Full Production: Who Will Supply the HBM4 Memory?

• SK hynix is the clear leader. Multiple analyses consistently estimate the company will hold a dominant 60–70% share of Nvidia's initial HBM4 orders for Vera Rubin . Its superior mass-production yields and a long-standing, deep technological partnership with Nvidia are cited as key reasons for this lead .
• Samsung Electronics, which led the charge in early HBM4 validation and began shipping HBM4E samples, is projected to secure roughly a 20–30% share . Some early reports even suggested Samsung would lead specifically in Vera Rubin-dedicated HBM4 volume before a broader supply base was confirmed .
• Micron Technology faced the most uncertainty. Reports in March 2026 initially suggested it would be excluded as a supplier for Vera Rubin in favor of the two Korean giants, causing its projected market share to drop to zero in some analyst models . Huang's June announcement marks a significant reversal, confirming Micron has passed qualification, although its allocation is expected to be the smallest of the three, possibly focused on mid-tier accelerators .

Navigating a HBM4 Supply Crunch

Even with all three major suppliers now on board, Nvidia is grappling with an urgent supply-demand gap. Huang's keynote highlighted the immense scale of the Vera Rubin ramp, noting that its supply chain is "twice as large" as the previous-generation Grace Blackwell . The next day, he made a surprise visit to the SK hynix booth at Computex and left a telling handwritten message on an HBM4E wafer: "Please Make More HBM" . This plea underscored the severity of the supply bottleneck and the pressure Huang is placing on his partners to accelerate production.

The Strategic Focus on South Korea

Huang's moves are not just about semiconductors; they are about cementing a deep, multifaceted alliance with South Korea, which is now a critical pillar of Nvidia's AI strategy. Samsung and SK hynix together already supply approximately 70% of all memory used in Nvidia's AI chips .

Immediately following the Computex keynote, Huang embarked on a four-day visit to South Korea, his second in seven months. The trip's agenda was telling: while meetings with Samsung Electronics Chairman Lee Jae-yong and SK Group Chairman Chey Tae-won were central to securing long-term HBM4 supplies, the trip’s narrative was broadened to include "physical AI" — robotics, autonomous driving, and smart factories .

Upon landing in Seoul, Huang declared that robotics would be South Korea's "next major sector" and hinted at plans to build a new R&D center in the country . This strategic expansion drove a surge in shares of robotics-related companies like LG Electronics and Doosan Robotics, which momentarily stole the spotlight from the chipmakers . The partnership goes both ways, as Huang also floated the idea of hosting Nvidia's GTC developer conference in Seoul .

The evolution from exclusive reports in March to a three-supplier certification in June reveals a competitive landscape where SK hynix holds the pole position, Samsung is a crucial multi-generational partner, and Micron has fought its way back into the most coveted AI memory contract in the world. As the Vera Rubin platform ramps up, Jensen Huang’s actions make it clear that the battle for AI infrastructure is a team sport, and right now, his most important teammates are in Korea.