H200 Specifications
Nearly double the memory capacity of H100 with 4,800 GB/s HBM3e bandwidth, enabling larger models, longer context windows, and faster inference without compromising precision.
141 GB
HBM3e Memory
Run 70B+ parameter models on a single GPU.
900 GB/s
NVLink Bandwidth
Fourth-gen NVLink for multi-GPU scaling.
1,979
TFLOPS FP16
3,958 TFLOPS at FP8 with Transformer Engine.
9X
Faster Pre-Training
vs A100 on large language models.
NVIDIA B200 on Denvr AI Cloud
1T Parameter Training
Scale across 8-GPU NVLink nodes for 1,440 GB of total VRAM and 1.8 TB/s per-GPU interconnect.
LLM Inference
Serve >70B parameter models on a single GPU, or scale to 8 GPUs with 1,536 GB VRAM for >1T+ parameter models at FP8 precision.
Production AI
H200 delivers 2x faster inference vs H100 with 1.8x the memory. For FP4 and next-gen training, consider the Blackwell GPUs with 180 GB HBM3e.
Managed Storage
High-performance Weka filesystem and local NVMe available. No external storage to provision for datasets, checkpoints, or model artifacts.
Configurations
Per-minute billing with on-demand and reserved options. All configurations available as bare metal, VM, or model endpoints.
Platform
GPUs
On-Demand
VRAM
vCPUs
Memory
Local Storage
Interconnect
NVIDIA H200 SXM
8
141 GB
208
2048 GB
6x 3.8TB NVMe
RoCE 3200G
Reserved only