Nvidia de-specced Rubin Ultra HBM from 12-Hi to 8-Hi: $/bandwidth is the bottleneck

AccBalanced · x · 2026-09-06

SemiAnalysis argues Nvidia de-specced Rubin Ultra's HBM from 12-Hi to 8-Hi because the real bottleneck is $/bandwidth, not $/capacity — and walks through the math in a thread.

Related event: SemiAnalysis Explains Why Nvidia Cut Rubin Ultra HBM to 8-Hi(2 posts)→

Original post →

More from Infra

Infra channel →