The Definitive Guide toAI Data Centers
Ask the GuideAboutAccount
Guide › Numbers register › Claim

GB200 usable HBM per accelerator from the published two-GPU superchip capacity

186 GBderived

Value kindderived — Derived values are guide calculations; their result is only as fixed as the stated inputs, scope, and method.
Scope372/2=186 GB; raw 192 GB is a different capacity boundary.
As of2026-09
SourceNVIDIA — Specifications: 372 GB per two-GPU Grace Blackwell Superchip. · Specifications: 372 GB per two-GPU Grace Blackwell Superchip
DerivationPublished 372 GB superchip capacity / 2 GPUs = 186 GB usable HBM per GPU.
Reviewchecking…review by 2026-12-05 · standard cadence
Recorded changeslast 2026-09-16
Claim idpart7-gb200-usable-hbm-per-gpu-2026

Where the guide uses it

← Full numbers register — every date-stamped figure in the guide, with revision history.