gpu catalog: h100 nvl
The inference-tuned H100 variant: 94 GB of HBM3 and higher bandwidth than the SXM part, purpose-built for serving LLMs in PCIe servers.
| GPU | VRAM | Regions | Availability | Price / GPU | |
|---|---|---|---|---|---|
| H100 NVLcommunity | 94G | 4 regions | running low | $2.67/hr | launch |
The inference-tuned H100 variant: 94 GB of HBM3 and higher bandwidth than the SXM part, purpose-built for serving LLMs in PCIe servers. VRAM94 GB HBM3 Memory bandwidth3.9 TB/s FP16 tensor835 TFLOPS ArchitectureHopper InterconnectPCIe 5.0 + NVLink bridge Board power400 W Released2023 Regionsasia-pacific, eu-east, eu-west, us-central | |||||
| H100 NVL | 94G | 1 region | running low | $3.19/hr | launch |
| RTX 5090community | 32G | 8 regions | high availability | $0.47/hr | launch |
| L40S | 48G | 3 regions | high availability | $0.80/hr | launch |
| A100 | 80G | 5 regions | high availability | $1.07/hr | launch |
| RTX 4090community | 24G | 7 regions | high availability | $0.34/hr | launch |
| RTX PRO 6000community | 96G | 3 regions | high availability | $1.69/hr | launch |
| RTX PRO 6000 | 96G | 3 regions | high availability | $2.09/hr | launch |
| H200 SXMcommunity | 141G | 4 regions | high availability | $3.59/hr | launch |
| RTX 3090community | 24G | 6 regions | available | $0.15/hr | launch |
| A100community | 80G | 6 regions | available | $0.95/hr | launch |
| H100 SXM | 80G | 4 regions | available | $3.49/hr | launch |
| H200 SXM | 141G | 4 regions | available | $4.59/hr | launch |
| A100 40GBcommunity | 40G | 5 regions | available | $0.47/hr | launch |
| RTX 6000 ADA | 48G | 2 regions | available | $0.65/hr | launch |
| L40Scommunity | 48G | 3 regions | available | $0.79/hr | launch |
| RTX 3080community | 10G | 4 regions | available | $0.11/hr | launch |
| RTX A4000 | 16G | 1 region | available | $0.12/hr | launch |
| RTX A6000 | 48G | 3 regions | available | $0.40/hr | launch |
| RTX 5080community | 16G | 6 regions | available | $0.23/hr | launch |
| RTX A4000community | 16G | 4 regions | available | $0.10/hr | launch |
| L4 | 24G | 3 regions | available | $0.49/hr | launch |
| RTX 4080community | 16G | 4 regions | running low | $0.21/hr | launch |
| H100 PCIE | 80G | 2 regions | running low | $1.98/hr | launch |
| RTX 2000 ADA | 16G | 2 regions | running low | $0.24/hr | launch |
| A40 | 48G | 2 regions | running low | $0.49/hr | launch |
| B300 | 288G | 2 regions | running low | $7.89/hr | launch |
| L40 | 48G | 1 region | running low | $0.78/hr | launch |
| RTX A6000community | 48G | 2 regions | running low | $0.40/hr | launch |
| RTX 4000 ADAcommunity | 20G | 1 region | running low | $0.20/hr | launch |
| RTX A5000 | 24G | 1 region | running low | $0.27/hr | launch |
| RTX 4000 ADA | 20G | 1 region | running low | $0.28/hr | launch |
| L4community | 24G | 3 regions | running low | $0.33/hr | launch |
| RTX 3090 | 24G | 1 region | running low | $0.50/hr | launch |
| L40community | 48G | 1 region | running low | $0.69/hr | launch |
| RTX PRO 4500 | 32G | 1 region | running low | $0.72/hr | launch |
| RTX 6000 ADAcommunity | 48G | 1 region | running low | $0.74/hr | launch |
| RTX 5090 | 32G | 1 region | running low | $0.99/hr | launch |
| H200 NVLcommunity | 141G | 2 regions | running low | $5.47/hr | launch |
| B200community | 192G | 2 regions | running low | $7.50/hr | launch |
| H100 SXMcommunity | 80G | 2 regions | running low | $3.56/hr | launch |
| V100community | 16G | 1 region | running low | $0.13/hr | launch |
| RTX A4500community | 20G | 1 region | running low | $0.13/hr | launch |
| RTX A5000community | 24G | 1 region | running low | $0.30/hr | launch |
| H200 NVL | 141G | 1 region | running low | $3.29/hr | launch |
lowest live on-demand rate per GPU, refreshed continuously · per-second billing
Our floor price for the H100 NVL, sampled continuously across every provider in the fleet and rolled up to one point per day. A dot on the zero line is a day we checked and found nothing launchable; a shaded break is a day we did not sample. Those are different facts, so this chart never draws one as the other.
Daily trend, updated hourly · Full 15-minute history available via API
| Provider | Price / GPU / hr | Notes |
|---|---|---|
| GPU.ai (Secure Cloud) | $3.19/hr | Vetted data center capacity, per-second billing |
| GPU.ai (Community Cloud) | $2.67/hr | Marketplace capacity from independent suppliers |
| CoreWeave | — | Public on-demand list rate |
| AWS | — | Public on-demand list rate |
| Azure | — | Public on-demand list rate |
competitor rates are public on-demand list prices, refreshed daily · gpu.ai prices are live floor rates with zero markup
faq
H100 NVL cloud instances on GPU.ai currently start at $2.67 per GPU per hour, billed per second. That is the live floor price across every provider in our fleet. We route your launch to the cheapest provider with stock and add zero markup.
The NVIDIA H100 NVL has 94 GB of HBM3 memory with up to 3,900 GB/s of bandwidth. GPU.ai reports the minimum VRAM actually delivered across our fleet, so the figure you see at launch is the figure you get.
Yes: real 2× H100 NVL nodes are launchable right now, from $2.73 per GPU per hour. We only advertise node sizes that exist in live inventory, not theoretical configurations.
GPU.ai aggregates a dozen-plus GPU clouds and passes through each provider's raw price with no markup, so the same silicon is typically far cheaper than hyperscaler list rates. The comparison table on this page shows the current public on-demand list prices side by side.
Billing is per second, starting when your instance is SSH-ready and stopping the moment you terminate. There are no minimum commitments, no reservation fees, and no egress surprises: you pay the listed hourly rate pro-rated to the second.