NVIDIA H20 8-GPU

Best-in-class inference GPU for large-model RAG and multi-tenant serving. 141 GB HBM3e per card.

Inference-first 141GB HBM3e 4.8 TB/s BW
Starting at
$5,500 /month
USD · Monthly reservation · Discounts available for annual terms
Request a Quote Email Sales

Specifications

GPU HGX H20 768 GB
CPU Intel Xeon 8480+ × 2 (56C)
Memory 2048 GB
Disk 2 × 960 GB + 8 × 3.84 TB
Network 4 × 400G + 1 × 200G + 1 × 25G
Power 4 × 2000 W (N+N)

Deployment

  • Bare-metal or fully managed
  • Provisioning in 72 hours
  • 8 US data center locations
  • 24/7 on-shore NOC