NextFin News - U.S. President Trump’s administration continues to oversee a period of intense industrial competition as Nvidia CEO Jensen Huang confirmed that the company’s next-generation AI platform, Vera Rubin, has entered full production. Speaking at the GTC Taipei 2026 keynote, Huang ended months of supply-chain speculation by naming SK Hynix, Samsung Electronics, and Micron Technology as the primary suppliers for the high-bandwidth memory (HBM4) essential to the new chip’s performance. The announcement marks a critical milestone for Nvidia as it attempts to maintain its dominance in the AI accelerator market while navigating a global semiconductor landscape defined by chronic supply shortages.
The Vera Rubin platform represents a significant architectural leap, and its reliance on sixth-generation HBM4 memory underscores the growing cost of AI infrastructure. According to Morgan Stanley, memory now accounts for a share of the parts cost in a Rubin rack that is 435% higher than in previous generations. This shift has turned memory manufacturers into the ultimate gatekeepers of AI scaling. During a visit to the SK Hynix booth at Computex 2026, Huang met with SK Group Chairman Chey Tae-won for a second consecutive day, even writing "Please Make More" on an HBM4E wafer—a blunt signal of the desperation for capacity that currently grips the industry.
SK Hynix has emerged as a particularly vital partner in this ecosystem. The South Korean firm, which recently joined the "trillion-won club" in market valuation, is moving aggressively to expand its footprint. Chey stated that the group aims to double its wafer production capacity within five years to meet the "insane" returns Huang has pitched to investors and suppliers alike. While Samsung and Micron have also been certified for HBM4 supply, SK Hynix’s early lead in HBM development has historically given it a preferred status in Nvidia’s high-end configurations, a position it is fighting to defend as Samsung’s stock surged 10.1% following the certification news.
However, the concentration of supply among three major players introduces systemic risks. While Huang’s confirmation of "full production" suggests that the technical hurdles of HBM4 integration have been cleared, the sheer volume of demand remains a bottleneck. Analysts at Morgan Stanley, who have maintained a cautious but attentive stance on the semiconductor cycle, suggest that the massive increase in memory costs could eventually squeeze the margins of data center operators, even if Nvidia itself remains insulated by its pricing power. The current market enthusiasm assumes that demand for AI training will continue to outpace the rapid expansion of hardware costs, a premise that has yet to be tested by a significant cooling in enterprise AI spending.
The geopolitical dimension of this supply chain cannot be ignored. With the "Big Three" memory makers all operating outside the immediate domestic control of the U.S., the Trump administration’s focus on securing high-tech manufacturing remains a background tension. For now, the partnership between Nvidia and its Asian suppliers is a marriage of necessity. As Huang prepares for an anticipated visit to Seoul later this week, the focus remains on whether SK Hynix and its peers can actually deliver the volume required to turn the Vera Rubin platform from a production success into a global standard. The "Please Make More" plea is no longer just a request; it is the fundamental requirement for the next phase of the AI boom.
Explore more exclusive insights at nextfin.ai.
