Next-generation technology. Enterprise-scale compute.
Technical resourcesTrack your quote
GPUs & accelerators / Hopper
NVIDIA / SERVER

NVIDIA H200

NVL PCIe

Memory-intensive AI inference and HPC on supported PCIe GPU servers.

Official NVIDIA product image: H200 NVL four-GPU bridge group
Four-GPU NVL set shown; memory specifications are per GPU · NVIDIA
141 GBHBM3e
4.8 TB/sMemory bandwidth
Up to 600 WGPU board power
HopperGPU architecture
TECHNICAL SPECIFICATIONS

Know what you’re ordering.

ManufacturerNVIDIA
ProductNVIDIA H200
EditionNVL PCIe
ArchitectureHopper
Memory141 GB HBM3e
Memory bandwidth4.8 TB/s
Board powerUp to 600 W
CoolingPassive; qualified server airflow
Host interfacePCIe 5.0 x16
Form factorFull-height, full-length · dual slot

Reviewed September 11, 2026. Peak bandwidth and board power are manufacturer specifications, not measured NexTec application results.

Platform & integration notes.

01

Supported 2- or 4-GPU NVLink bridge groups provide up to 900 GB/s NVLink bandwidth per GPU.

02

An eight-card NVL server does not imply the eight-way NVSwitch fabric of HGX H200.

03

Specify bridge kit, spacing, power configuration, and qualified host system in the bill of materials.

LET’S BUILD WHAT’S NEXT

Your workload. Our starting point.

Tell us what you’re building. We’ll help scope the infrastructure.

Request a quote