Samsung, SK Hynix to Mass-Produce 8-Layer HBM4 for Nvidia in Q2 Next Year

The staggered rollout is the tell: 12-layer HBM4 carries this year's volume, while 8-layer is positioned for Rubin Ultra rack-scale systems from mid-next year. That sequencing matters beyond memory suppliers — it signals when Nvidia's next rack architecture shifts from roadmap to bill-of-materials reality, with knock-on effects for the broader AI hardware supply chain. The open question is whether the two Korean suppliers can hold that Q2 window, since slippage would ripple into Nvidia's rack-scale timeline rather than stay a memory-sector story.