Nvidia's Next AI Superchip Needs A New Kind of Memory — And One Company Controls It
Nvidia's next-gen Vera Rubin AI platform requires HBM4 memory, which SK Hynix is poised to supply with 70% of initial volume. SK Hynix is investing $12.85 billion in a new megafab to manufacture HBM4 memory.
Nvidia's next-gen Vera Rubin AI platform, unveiled at GTC 2026, requires HBM4 memory running at speeds over 10Gb/s. SK Hynix has locked in roughly 70% of initial HBM4 volume, making it a crucial supplier. SK Hynix is investing $12.85 billion in a new megafab in Cheongju to manufacture HBM4 memory. The company is co-developing HBM4 base die using TSMC's logic process. SK Hynix's operating margin reached 72% in the first quarter, indicating a significant structural shift in the memory market. Nvidia's Vera Rubin platform relies heavily on SK Hynix's ability to hit yield and scale at its new megafab.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.