Compute · Proven Hopper
NVIDIA HGX H100
The battle-tested workhorse: cost-efficient fine-tuning, inference and large-scale training with the most mature software ecosystem in AI.
Specifications
Proven at every scale
| GPUs per node | 8 × H100 (SXM) |
| GPU memory | 80 GB HBM3 per GPU |
| Memory bandwidth | 3.35 TB/s per GPU |
| Node interconnect | NVLink 4, NVSwitch |
| Cluster interconnect | 400 Gb/s InfiniBand per GPU (Quantum-2) |
| Availability | On-demand, spot & reserved |
Interconnect
Linear scaling, validated
Every H100 cluster is delivered with NCCL all-reduce results across the full node count — you see the scaling efficiency before your first job starts.
Suited workloads
Fine-tuning
Production inference
Large-scale training
Mature-stack deployments
Start your journey today
Tell us about your workload — an engineer, not a sales bot, will get back to you within one business day.