NVIDIA H100 vs H200
Compare NVIDIA H100 and H200 side by side: compute, memory, and pricing. H100 runs on-demand on VESSL Cloud from $2.98/hr. H200 (141GB HBM3e) is reserved VM Cluster capacity.
from $2.98/hr
- GPU memory
- 80GB HBM3
- Memory bandwidth
- 3.35 TB/s
- GPU memory
- 141GB HBM3e
- Memory bandwidth
- 4.8 TB/s
Technical specifications
H100 NVIDIA H100 SXM | H200 NVIDIA H200 SXM | |
|---|---|---|
| Architecture | Hopper | Hopper |
| GPU memory | 80GB HBM3 | 141GB HBM3e |
| Memory bandwidth | 3.35 TB/s | 4.8 TB/s |
| NVLink | 900 GB/s | 900 GB/s |
| FP16/BF16 (Tensor) | 1,979 TFLOPS | 1,979 TFLOPS |
| FP8 (Tensor) | 3,958 TFLOPS | 3,958 TFLOPS |
| Max TDP | 700W | 700W |
| GPUs per node | 8 (HGX H100) | 8 (HGX H200) |
*Peak performance with sparsity, per NVIDIA official specs. Final specs may vary by node configuration.
Pricing & availability
What's Hopper best for?
Large-scale LLM training & fine-tuning
Run 70B–400B-class training on a proven Hopper stack: FP8 tensor cores, multi-node InfiniBand, and NeMo and Megatron support.
High-throughput LLM inference
H200's 141GB HBM3e fits 70B-class models on a single GPU, while H100 FP8 keeps token throughput high and tail latency low.
Research & HPC
Mature CUDA, PyTorch, JAX, and NeMo ecosystem: spin up Hopper nodes on demand for experiments, scientific compute, and deadline-driven runs.
Why industry-leading teams run GPUs on VESSL Cloud
Capacity across clouds
Access GPU capacity across clouds through one platform, without per-cloud quotas and contracts.
From one node to a cluster
Run up to 8 GPUs on one node, and move to a multi-node VM Cluster over InfiniBand when one node is not enough.
Transparent pricing
Published per-second rates for self-serve GPUs, and reserved terms quoted by GPU, term, and volume.
Enterprise-ready
SOC 2 Type II compliance, with dedicated support for production AI.
Frequently asked questions
How much does an NVIDIA H100 cost on VESSL Cloud?
H100 SXM (80GB) starts at $2.98/hr on-demand. Reserved rates are set by term and volume. Provision self-serve at cloud.vessl.ai, or reserve capacity to guarantee it.
What's the difference between the H100 and H200?
Both share the same Hopper compute, but the H200 carries 141GB of faster HBM3e memory (vs 80GB HBM3 on the H100) at 4.8 TB/s bandwidth, fitting larger models, bigger batches, and longer context windows.
Is the H200 available now?
H200 runs as reserved VM Cluster capacity. Talk to our team for current availability and pricing.
Can I run multi-node H100/H200 training?
Yes, on a VM Cluster: HGX H100 or H200 nodes (8 GPUs each) joined by high-speed InfiniBand for distributed training.
Do you offer reserved pricing?
Reserved commitments hold your capacity, at a rate set by term and volume. Contact us.
Where AI models
get their GPUs
- Start in minutes
- Scale to multi-node clusters
- Capacity reserved to your timeline
- Dedicated support