SemiAnalysis Explains Why Nvidia Cut Rubin Ultra HBM to 8-Hi

SemiAnalysis argues Nvidia cut Rubin Ultra's HBM from 12-Hi to 8-Hi because cost per bandwidth, not capacity, is the real bottleneck, with the change delivering roughly 50% more bandwidth per GB.

2026-09-06 ~ 2026-09-06 · 2 related posts

• Nvidia de-specced Rubin Ultra HBM from 12-Hi to 8-Hi: $/bandwidth is the bottleneck — AccBalanced · 2026-09-06
• Cutting 12-Hi to 8-Hi lifts bandwidth-per-GB 50%: the Rubin Ultra HBM math — AccBalanced · 2026-09-06