NextFin News - In a move that underscores the critical interdependence of the global artificial intelligence supply chain, SK Group Chairman Chey Tae-won met with Nvidia CEO Jensen Huang in Santa Clara, California, on February 5, 2026. The meeting, held at a local Korean restaurant, served as a high-stakes strategy session disguised as a casual "chimaek" (chicken and beer) dinner. According to BusinessKorea, the discussions focused on the supply of sixth-generation High Bandwidth Memory (HBM4) for Nvidia’s upcoming "Vera Rubin" AI architecture, scheduled for release in the second half of 2026.
The timing of this meeting is pivotal. SK Hynix, the semiconductor arm of SK Group, currently commands approximately 70% of the market for HBM3E chips used in Nvidia’s Blackwell GPUs. However, the landscape is shifting rapidly. Samsung Electronics recently announced it would begin mass production of HBM4 in the third week of February 2026, aiming to challenge SK’s dominance. By meeting Huang in person, Chey sought to solidify SK’s position as the primary partner for Nvidia’s next-generation accelerators, with reports suggesting SK has already secured over 60% of the initial HBM4 orders for the Vera Rubin line.
Beyond immediate hardware procurement, the dialogue delved into the broader AI ecosystem. The presence of the leaders' daughters—Choi Min-jung, CEO of Integral Health, and Madison Huang, a senior director at Nvidia’s Robotics Division—indicates a long-term vision for cross-generational collaboration in robotics and AI-driven healthcare. Furthermore, SK Group’s recent decision to rename its U.S. NAND subsidiary Solidigm to "AI Company" and commit $10 billion to AI solutions suggests a transition from a component supplier to a full-stack AI infrastructure provider. This aligns with Nvidia’s expansion into autonomous systems and digital twins, where high-speed memory and specialized NAND flash are increasingly vital.
From a geopolitical perspective, this alliance is navigating a complex regulatory environment. U.S. President Trump, inaugurated in January 2025, has maintained a rigorous "America First" stance on technology, emphasizing the reshoring of semiconductor manufacturing and the security of AI supply chains. For SK Group, maintaining a "special relationship" with Nvidia is not just a commercial necessity but a strategic buffer against potential trade volatility. By deepening ties with the most valuable U.S. chip designer, SK positions itself as an indispensable partner in the U.S.-led AI bloc, potentially mitigating risks associated with tariffs or export controls.
The technical shift to HBM4 represents a fundamental change in chip architecture. Unlike previous generations, HBM4 will utilize a logic base die manufactured on advanced foundry processes—specifically the 3nm node. SK Hynix has partnered with TSMC for this transition, creating a powerful "triangular alliance" between the world’s top memory maker, the leading foundry, and the dominant AI chip designer. This integration makes it increasingly difficult for competitors to break into the ecosystem, as the customization required for HBM4 necessitates deep, early-stage collaboration between the memory provider and the GPU architect.
Looking ahead, the success of the SK-Nvidia partnership will likely depend on their ability to scale "Custom HBM" (cHBM). As AI models grow in complexity, generic memory solutions are reaching their thermal and bandwidth limits. The industry is moving toward a model where memory is co-designed with the processor. If SK can successfully integrate its memory directly onto Nvidia’s logic dies, it will effectively lock in its market share for the remainder of the decade. However, with Samsung doubling down on its own HBM4 capacity and U.S.-based Micron receiving significant federal support under the current administration, the race for AI memory supremacy is far from over. The Santa Clara meeting confirms that in the 2026 AI economy, personal diplomacy at the CEO level remains as critical as engineering breakthroughs.
Explore more exclusive insights at nextfin.ai.
