Cluster · NVIDIA · platform

NVIDIA HGX H200 (8-GPU baseboard)

H100's memory-refresh sibling: same footprint, 141 GB HBM3e per GPU (+76%) and 4.8 TB/s bandwidth. Inference-oriented shops upgraded here.

Key facts

Chips per instance: 8
Power (typical): 10.2 kW
Interconnect: NVLink 4
Cooling: air
Announced: 2023-11-13
GA / release: 2024-03-01

Chips inside

×8NVIDIA H200 Tensor Core GPUH200 SXM5 GPU

Where you'll find it

Data centers publicly running the chips inside this design. Per-rack counts are rarely disclosed, so these are presence rows: this campus runs these chips.

xAI Colossus (Memphis) · Second-wave upgrade GPUs added late 2024
Khazna QAJ1 Ajman AI-Optimized Data Center · H200/GB200-generation capacity for G42 / UAE AI stack
Lake Mariner Data Campus · H200-generation HPC capacity for AI-training tenants
Crusoe Childress Campus · Recent Crusoe expansions include H200 capacity

Vendor page: https://www.nvidia.com/en-us/data-center/h200/

See also: all rack designs · chip catalog · NVIDIA.