Compare
Compare GPU Pricing
Select a GPU to compare per-hour rental pricing across all cloud providers.
Providers
26
Price Points
237
GB300 279–288GB · 2 GB200 186–189GB · 3 B300 263–288GB · 11 B200 179–192GB · 15 H200 141GB · 17 H100 80–94GB · 25 GH200 96GB · 3 A100 40–80GB · 20 L40S 48GB · 12 L40 48GB · 6 L4 24GB · 8 A40 48GB · 4 A10 24GB · 7 RTX PRO 6000 96GB · 13 RTX PRO 5000 48GB · 1 RTX 6000 Ada 48GB · 4 RTX A6000 48GB · 9 RTX A5000 24GB · 3 RTX A4000 16GB · 2 RTX 5090 32GB · 4 RTX 4090 24GB · 6 RTX 3090 24GB · 3 V100 16GB · 4 T4 16GB · 5 MI355X 288GB · 2 MI325X 256GB · 2 MI300X 192GB · 5
| Provider | Variant | VRAM | On-Demand | Spot | Reserved | Min GPUs | Billing | Notes |
|---|---|---|---|---|---|---|---|---|
| Latitude.sh Cheapest | H100 80GB | 80GB | $1.68 /hr | — | — | 1x | hr | Bare metal server |
| Vast.ai | H100 SXM | 80GB | $1.73 /hr | — | — | 1x | sec | Marketplace minimum |
| Voltage Park | H100 SXM | 80GB | $1.99 /hr | — | — | 1x | hr | 'Starting at' price for on-demand Ethernet H100 without a contract; InfiniBand and long-term reserve via sales |
| TensorDock | H100 SXM | 80GB | $2.25 /hr | — | — | 1x | sec | Starting price; marketplace hosts set rates. GPU only, vCPU/RAM/storage extra |
| Vultr | HGX H100 | 80GB | $2.30 /hr | — | — | 8x | hr | 24-month contract available |
| Vast.ai | H100 PCIe | 80GB | $2.33 /hr | — | — | 1x | sec | Marketplace minimum; prices vary by host |
| Hyperstack | H100 PCIe | 80GB | $2.50 /hr | $2.00 /hr | $1.75 /hr | 1x | min | — |
| Hyperstack | H100 NVLink | 80GB | $2.60 /hr | — | $1.82 /hr | 1x | min | — |
| Vast.ai | H100 NVL | 94GB | $2.60 /hr | — | — | 1x | sec | Marketplace minimum |
| RunPod | H100 PCIe | 80GB | $2.89 /hr | — | — | 1x | ms | — |
| Prime Intellect | H100 80GB | 80GB | $3.00 /hr | — | — | 8x | min | Cheapest marketplace offer was an 8x node ($24/hr). Live marketplace prices move with supply |
| RunPod | H100 NVL | 94GB | $3.19 /hr | — | — | 1x | ms | — |
| Hyperstack | H100 SXM | 80GB | $3.20 /hr | — | $2.72 /hr | 1x | min | — |
| Thunder Compute | H100 SXM | 80GB | $3.20 /hr | — | — | 1x | min | Instance on-demand; per-second sandboxes $2.95/GPU-hr |
| Lambda Cloud | H100 PCIe | 80GB | $3.29 /hr | — | — | 1x | hr | 1x instance |
| RunPod | H100 SXM | 80GB | $3.49 /hr | — | — | 1x | ms | — |
| Jarvislabs | H100 SXM | 80GB | $3.49 /hr | — | — | 1x | min | Term discounts 2-9% (1 month to 1 year) |
| Oblivus | H100 PCIe | 80GB | $3.50 /hr | — | $2.56 /hr | 1x | hr | — |
| Oblivus | H100 NVLink | 80GB | $3.64 /hr | — | $2.71 /hr | 1x | hr | — |
| Verda | H100 SXM | 80GB | $3.70 /hr | $1.85 /hr | — | 1x | hr | — |
| Modal | H100 SXM | 80GB | $3.95 /hr | — | — | 1x | sec | $0.001097/sec; GPU only |
| Lambda Cloud | H100 SXM | 80GB | $3.99 /hr | — | — | 8x | hr | — |
| Together AI | H100 | 80GB | $3.99 /hr | $1.99 /hr | $3.19 /hr | 1x | hr | GPU cluster on-demand; dedicated-inference endpoint $6.49/GPU-hr |
| Oblivus | H100 SXM5 | 80GB | $4.31 /hr | — | $2.92 /hr | 1x | hr | — |
| DigitalOcean | H100 SXM | 80GB | $4.41 /hr | — | $3.26 /hr | 1x | sec | HGX H100, 1x or 8x; reserved = 12-month contract |
| Crusoe | H100 SXM | 80GB | $4.50 /hr | — | — | 1x | min | HGX H100, on-demand |
| Nebius AI Cloud | H100 SXM | 80GB | $4.50 /hr | $0.79 /hr | — | 1x | sec | HGX H100, 1 or 8 GPUs. Preemptible is a dynamic 'from' floor |
| Paperspace | H100 | 80GB | $5.95 /hr | — | $2.24 /hr | 1x | hr | Paperspace lists $5.95 as a promo price; reserved is a 3-year commitment |
| CoreWeave | HGX H100 | 80GB | $6.16 /hr | $2.46 /hr | — | 8x | hr | Per-GPU price from 8-GPU instance ($49.24/hr) |
| Baseten | H100 80GB | 80GB | $6.50 /hr | — | — | 1x | min | $0.10833/min dedicated deployment; MIG half-H100 (40GB) $0.0625/min |
| AWS | H100 SXM | 80GB | $6.88 /hr | $3.92 /hr | $2.59 /hr | 1x | sec | p5.4xlarge (1 GPU) us-east-1; p5.48xlarge same per GPU. Reserved = 3-yr All Upfront Savings Plan |
| Microsoft Azure | H100 NVL | 94GB | $6.98 /hr | $1.29 /hr | $3.84 /hr | 1x | sec | NC40ads H100 v5 (1 GPU) East US. Reserved = 3-yr reservation |
| Oracle Cloud | H100 SXM | 80GB | $10.00 /hr | — | — | 8x | sec | BM.GPU.H100.8 bare metal |
| Google Cloud | H100 SXM | 80GB | $11.06 /hr | $6.60 /hr | $4.86 /hr | 8x | sec | a3-highgpu-8g $88.49/hr / 8 (us-central1). Reserved = 3-yr CUD; 1-4 GPU sizes Spot/Flex-start only |
| Microsoft Azure | H100 SXM | 80GB | $12.29 /hr | $2.27 /hr | $4.92 /hr | 8x | sec | ND96isr H100 v5, East US, $98.32/hr / 8. Reserved = 5-yr reservation (3-yr $5.40) |
Cost Estimate (1x H100, cheapest offer)
Cheapest / Hour
$1.68
Cheapest / Day
$40.32
Cheapest / Month
$1210
Need inference APIs instead?
Compare per-token and per-request pricing across AI inference providers.