We take cryptocurrency — so the coins you hold can buy real hardware.Pay in crypto, spend it on real hardware. Free insured shipping worldwide over $399. Plain, unbranded boxes, sent from Germany.Free insured shipping over $399. Your rate is locked for 30 minutes once checkout opens.Rate locked 30 minutes. Every item sealed, serial-checked and covered by a 24-month warranty.Sealed, 24-month warranty.

Saved items Sign in
Full-length PCIe 5.0 x16 AI accelerator card with passive heatsink

NVIDIA

NVIDIA H200 NVL AI Accelerator Card 141GB HBM3e PCIe 5.0 x16 900-21010-0040-000

SKU CH-BRVX-8Z0N1 MPN 900-21010-0040-000

An AI accelerator with 141GB HBM3e memory and 4.8TB/s bandwidth for enterprise generative AI and HPC workloads

  • Graphics processorNVIDIA H200
  • Memory capacity141 GB
  • Interface typePCI Express 5.0 x16
  • Type of memoryHBM3e
  • Chipset makerNVIDIA
  • 141 GB HBM3e memory enables larger AI models
  • 8 TB/s bandwidth accelerates data-intensive workloads
  • PCIe 5.0 x16 interface provides high-speed host connectivity
  • 700 W TDP fits standard air-cooled server racks
  • Passive cooling design reduces infrastructure complexity

H200 NVL AI Accelerator for Enterprise Servers

The NVIDIA H200 NVL is an AI accelerator card built for enterprise generative AI, LLM inference, fine-tuning and high-performance computing workloads. It uses the Hopper architecture GPU with 141GB of HBM3e memory connected through a PCI Express 5.0 x16 interface. The card targets mainstream enterprise servers that need flexible, air-cooled rack configurations.

Passive Cooling with 700W TDP and 4.8TB/s Bandwidth

This card employs a passive cooling design optimised for low-power air-cooled enterprise rack deployments. It operates at a 700W TDP while delivering 4.8TB/s memory bandwidth from the HBM3e memory. The thermal solution is built for dense server environments where active fan cooling on the card itself is not used.

141GB HBM3e Memory Decides Fit for Large Models

The single specification that determines suitability is the 141GB HBM3e memory capacity with 4.8TB/s bandwidth. This capacity enables larger model inference and fine-tuning without offloading to system memory. Buyers who need more memory per GPU or who require active cooling on the card should evaluate other form factors.

How it compares

Measured against everything we stock in GPUs for AI & Data Centers, at today’s prices. Recalculated every night.

Price $25,413.99
Pricier than about 99% of the 36 comparable items (median $5,369.99).
Memory capacity 141 GB
Higher than about 99% of the 12 comparable items (median 79 GB).

New to this? Buying server hardware with crypto — what to check, what it costs to pay, and what is never asked for.

Specifications

Product
Brand nameNVIDIA
Product lineTesla Hopper
Model nameGPU-NVH200NV
Part number900-21010-0040-000
Memory
Type of memoryHBM3e
Graphics
Graphics processorNVIDIA H200
Memory capacity141 GB
Chipset makerNVIDIA
Ports & slots
Interface typePCI Express 5.0 x16
Size & build
Shipping weight1.38 kg
Package size324 × 160 × 70 mm

Questions about this item

How do I install the PCIe 5.0 x16 card in an air-cooled rack, and what step is often missed?

Seat the card in a PCIe 5.0 x16 slot, secure the bracket, then connect NVLink bridges between GPUs before powering on; the passive 700W design needs confirmed rack airflow.

What maintenance or replacement will the 141GB HBM3e module need later?

The HBM3e memory is permanently packaged with the GPU; no user-serviceable parts exist, so replacement means swapping the entire accelerator card.

What limit appears first under sustained 700W load with passive cooling?

Thermal headroom in the rack is the first constraint; without sufficient air flow the GPU will throttle to stay within its 700W TDP envelope.

More in GPUs for AI & Data Centers

Similar items