Jensen Huang to Meet Samsung, SK Chiefs in New York on AI Chips

The repeated meetings between Jensen Huang, Jay Y. Lee, and Chey Tae-won confirm a fundamental shift in AI chip design. HBM is now a co-design element with GPUs, not a post-GPU selection. This means early access to Nvidia's future GPU roadmap directly determines a memory maker's competitive edge. Samsung and SK Hynix are competing intensely for this early integration.

For South Korea, this tightens its grip on the global AI infrastructure supply chain. Samsung's ability to offer memory, foundry, and advanced packaging in one vertical gives it a distinct advantage over SK Hynix. SK Group's plan to build a 2-gigawatt AI factory based on Nvidia's DSX platform, starting in 2027, also secures its role beyond just memory supply. This elevates Korea as a core partner in AI.

The key thing to watch is how quickly Samsung can expand its custom HBM offerings. Its February mass production of HBM4 and May sampling of 12-layer HBM4E are strong starts. The test for Samsung is whether it can convert these samples into broad adoption across Nvidia's next-generation GPU platforms, moving beyond HBM supply to full AI chip production.