AI COMPUTE•NVIDIA•GPU CLOUD•DATA CENTERS•SEMICONDUCTORS•ENERGY•FUNDING•M&A•AI COMPUTE•NVIDIA•GPU CLOUD•DATA CENTERS•SEMICONDUCTORS•ENERGY•FUNDING•M&A•

NVIDIA

NVIDIA GB200

Grace Blackwell Superchip pairing one Grace CPU with two Blackwell GPUs; deployed in rack-scale systems including GB200 NVL72.

Architecture
Blackwell
Memory
372 GB HBM3e per Grace Blackwell Superchip
Memory bandwidth
16 TB/s per Grace Blackwell Superchip
Release
Announced 2024

Main use cases

Rack-scale AI training and inference

Cloud pricing

Full tracker →

No tracked rates for this accelerator yet.

Sold as rack-scale systems rather than single cards.

NVIDIA GB200 in the news

GPU news →

No coverage yet.