Cloud GPU Rate Summary
Overall median, low, high, and continuous monthly cost per GPU model · Click a model to filter · Scroll down for full provider rates ↓
| GPU Model | VRAM | Low / hr | Median / hr | High / hr | Spread Ratio of highest to lowest on-demand rate for the same GPU model (e.g. 2.5× means highest price is 2.5× the lowest). | Monthly (730h) |
|---|---|---|---|---|---|---|
SXM / PCIe 80GB | 80 GB | $1.49 | $2.59 | $3.98 | 2.67× | $1,890.70 |
SXM / NVL 141GB | 141 GB | $2.19 | $3.39 | $4.65 | 2.12× | $2,474.70 |
40GB / 80GB | 80 GB | $0.95 | $1.89 | $4.10 | 4.31× | $1,379.70 |
48GB GDDR6/GDDR7 | 48 GB | $0.52 | $0.76 | $0.80 | 1.54× | $554.80 |
48GB Ada Lovelace | 48 GB | $0.72 | $1.19 | $1.83 | 2.53× | $868.70 |
192GB Blackwell | 192 GB | $3.95 | $4.35 | $5.50 | 1.39× | $3,175.50 |
Complete Cloud GPU Rate Comparison
Showing 49 normalized rate entries across 12 cloud providers
| GPU & Variant | Provider | Normalised $/GPU/hr | Billing Model | Region |
|---|---|---|---|---|
NVIDIA H100(SXM / PCIe 80GB) | ||||
H100-SXM VRAM 80GB |  AWS | $3.88 (8x node @ $31.04/hr) | On-Demand | us-east-1 |
H100-SXM VRAM 80GB | AWS | $1.75 (8x node @ $13.96/hr) | Spot / Preemptible | us-east-1 |
H100-SXM VRAM 80GB |  Azure | $3.98 (8x node @ $31.84/hr) | On-Demand | eastus |
H100-SXM VRAM 80GB |  CoreWeave | $2.23 | On-Demand | us-east |
H100-SXM VRAM 80GB | CoreWeave | $1.85 | Reserved (12mo) | us-east |
H100-SXM VRAM 80GB |  Crusoe | $2.25 | On-Demand | us-north-1 |
H100-SXM VRAM 80GB | '%20d='M38.193,18.359c-0.771-2.753-2.319-5.177-4.397-7.03l-4.598,4.598%20c1.677,1.365,2.808,3.374,3.014,5.648v1.508c0.026,0,0.05-0.008,0.076-0.008c2.322,0,4.212,1.89,4.212,4.212S34.61,31.5,32.288,31.5%20c-0.026,0-0.05-0.007-0.076-0.008V31.5h-6.666H24V38h8.212v-0.004c0.026,0,0.05,0.004,0.076,0.004C38.195,38,43,33.194,43,27.288%20C43,23.563,41.086,20.279,38.193,18.359z'%20/%3e%3cpath%20fill='%23ffe082'%20d='M19.56,25.59l4.72-4.72c-0.004-0.005-0.008-0.009-0.011-0.013l-4.717,4.717%20C19.554,25.579,19.557,25.584,19.56,25.59z'%20opacity='.5'%20/%3e%3cpath%20fill='%2390caf9'%20d='M19.56,25.59l4.72-4.72c-0.004-0.005-0.008-0.009-0.011-0.013l-4.717,4.717%20C19.554,25.579,19.557,25.584,19.56,25.59z'%20opacity='.5'%20/%3e%3clinearGradient%20id='nBzteS51Qxjtut4Vh3V9rb'%20x1='24.835'%20x2='18.428'%20y1='9.608'%20y2='19.289'%20gradientUnits='userSpaceOnUse'%20%3e%3cstop%20offset='0'%20stop-color='%23f7572f'%20/%3e%3cstop%20offset='.523'%20stop-color='%23f7552d'%20/%3e%3cstop%20offset='.712'%20stop-color='%23f75026'%20/%3e%3cstop%20offset='.846'%20stop-color='%23f7461b'%20/%3e%3cstop%20offset='.954'%20stop-color='%23f7390a'%20/%3e%3cstop%20offset='1'%20stop-color='%23f73100'%20/%3e%3c/linearGradient%3e%3cpath%20fill='url(%23nBzteS51Qxjtut4Vh3V9rb)'%20d='M24,7.576c-8.133,0-14.75,6.617-14.75,14.75c0,0.233,0.024,0.46,0.035,0.69h6.5%20c-0.019-0.228-0.035-0.457-0.035-0.69c0-4.549,3.701-8.25,8.25-8.25c1.969,0,3.778,0.696,5.198,1.851l4.598-4.598%20C31.188,9.003,27.761,7.576,24,7.576z'%20/%3e%3cpath%20fill='%2390caf9'%20d='M15.712,31.5L15.712,31.5c-0.001,0-0.001,0-0.002,0c-0.611,0-1.188-0.137-1.712-0.373%20l-4.71,4.71C11.081,37.188,13.301,38,15.71,38c0.001,0,0.001,0,0.002,0v0H24v-6.5H15.712z'%20opacity='.5'%20/%3e%3clinearGradient%20id='nBzteS51Qxjtut4Vh3V9rc'%20x1='31.305'%20x2='8.555'%20y1='40.347'%20y2='30.222'%20gradientUnits='userSpaceOnUse'%20%3e%3cstop%20offset='0'%20stop-color='%234caf50'%20/%3e%3cstop%20offset='.486'%20stop-color='%234aae50'%20/%3e%3cstop%20offset='.661'%20stop-color='%2343a94e'%20/%3e%3cstop%20offset='.786'%20stop-color='%2338a14c'%20/%3e%3cstop%20offset='.887'%20stop-color='%23279648'%20/%3e%3cstop%20offset='.972'%20stop-color='%23184'%20/%3e%3cstop%20offset='1'%20stop-color='%23088242'%20/%3e%3c/linearGradient%3e%3cpath%20fill='url(%23nBzteS51Qxjtut4Vh3V9rc)'%20d='M15.712,31.5L15.712,31.5c-0.001,0-0.001,0-0.002,0c-0.611,0-1.188-0.137-1.712-0.373%20l-4.71,4.71C11.081,37.188,13.301,38,15.71,38c0.001,0,0.001,0,0.002,0v0H24v-6.5H15.712z'%20/%3e%3clinearGradient%20id='nBzteS51Qxjtut4Vh3V9rd'%20x1='17.16'%20x2='7.285'%20y1='28.692'%20y2='19.067'%20gradientUnits='userSpaceOnUse'%20%3e%3cstop%20offset='0'%20stop-color='%23ffd747'%20/%3e%3cstop%20offset='.482'%20stop-color='%23ffd645'%20/%3e%3cstop%20offset='.655'%20stop-color='%23fed43e'%20/%3e%3cstop%20offset='.779'%20stop-color='%23fccf33'%20/%3e%3cstop%20offset='.879'%20stop-color='%23fac922'%20/%3e%3cstop%20offset='.964'%20stop-color='%23f7c10c'%20/%3e%3cstop%20offset='1'%20stop-color='%23f5bc00'%20/%3e%3c/linearGradient%3e%3cpath%20fill='url(%23nBzteS51Qxjtut4Vh3V9rd)'%20d='M11.5,27.29c0-2.32,1.89-4.21,4.21-4.21c1.703,0,3.178,1.023,3.841,2.494l4.717-4.717%20c-1.961-2.602-5.065-4.277-8.559-4.277C9.81,16.58,5,21.38,5,27.29c0,3.491,1.691,6.59,4.288,8.547l4.71-4.71%20C12.53,30.469,11.5,28.999,11.5,27.29z'%20/%3e%3c/svg%3e) Google Cloud | $3.73 (8x node @ $29.80/hr) | On-Demand | us-central1 |
H100-SXM VRAM 80GB | Lambda | $2.49 | On-Demand | us-tx-1 |
H100-SXM VRAM 80GB |  Nebius | $2.15 | On-Demand | eu-west-1 |
H100-SXM VRAM 80GB | Oracle | $2.95 (8x node @ $23.60/hr) | On-Demand | us-ashburn-1 |
H100-SXM VRAM 80GB |  RunPod | $2.69 | On-Demand | global |
H100-PCIE VRAM 80GB | RunPod Community | $1.79 | On-Demand | global |
H100-SXM VRAM 80GB |  Together | $2.85 | On-Demand | us-west |
H100-PCIE VRAM 80GB |  Vast.ai | $1.49 | On-Demand | global |
NVIDIA H200(SXM / NVL 141GB) | ||||
H200-SXM VRAM 141GB | AWS | $4.60 (8x node @ $36.80/hr) | On-Demand | us-east-1 |
H200-SXM VRAM 141GB | Azure | $4.65 (8x node @ $37.20/hr) | On-Demand | eastus |
H200-SXM VRAM 141GB | CoreWeave | $3.10 | On-Demand | us-east |
H200-SXM VRAM 141GB | Google Cloud | $4.45 (8x node @ $35.60/hr) | On-Demand | us-central1 |
H200-SXM VRAM 141GB | Lambda | $3.29 | On-Demand | us-tx-1 |
H200-SXM VRAM 141GB | Nebius | $2.95 | On-Demand | eu-west-1 |
H200-SXM VRAM 141GB | RunPod | $3.49 | On-Demand | global |
H200-NVL VRAM 141GB | Vast.ai | $2.19 | On-Demand | global |
NVIDIA A100(40GB / 80GB) | ||||
A100-80GB VRAM 80GB | AWS | $4.10 (8x node @ $32.77/hr) | On-Demand | us-east-1 |
A100-80GB VRAM 80GB | Azure | $3.67 (8x node @ $29.36/hr) | On-Demand | eastus |
A100-80GB VRAM 80GB | CoreWeave | $1.62 | On-Demand | us-east |
A100-80GB VRAM 80GB | Google Cloud | $3.67 (8x node @ $29.33/hr) | On-Demand | us-central1 |
A100-80GB VRAM 80GB | Lambda | $1.89 | On-Demand | us-tx-1 |
A100-80GB VRAM 80GB | Nebius | $1.60 | On-Demand | eu-west-1 |
A100-80GB VRAM 80GB | Oracle | $2.95 (8x node @ $23.60/hr) | On-Demand | us-ashburn-1 |
A100-80GB VRAM 80GB | RunPod | $1.89 | On-Demand | global |
A100-80GB VRAM 80GB | Together | $1.75 | On-Demand | us-west |
A100-40GB VRAM 40GB | Vast.ai | $0.95 | On-Demand | global |
NVIDIA RTX PRO 6000(48GB GDDR6/GDDR7) | ||||
RTX-PRO-6000 VRAM 48GB | CoreWeave | $0.76 | On-Demand | us-east |
RTX-PRO-6000 VRAM 48GB | Lambda | $0.80 | On-Demand | us-tx-1 |
RTX-PRO-6000 VRAM 48GB | RunPod | $0.79 | On-Demand | global |
RTX-PRO-6000 VRAM 48GB | RunPod Community | $0.58 | On-Demand | global |
RTX-PRO-6000 VRAM 48GB | Vast.ai | $0.52 | On-Demand | global |
NVIDIA L40S(48GB Ada Lovelace) | ||||
L40S VRAM 48GB | AWS | $1.83 (4x node @ $7.30/hr) | On-Demand | us-east-1 |
L40S VRAM 48GB | CoreWeave | $1.15 | On-Demand | us-east |
L40S VRAM 48GB | Google Cloud | $1.80 (4x node @ $7.20/hr) | On-Demand | us-central1 |
L40S VRAM 48GB | Lambda | $1.20 | On-Demand | us-tx-1 |
L40S VRAM 48GB | Nebius | $1.10 | On-Demand | eu-west-1 |
L40S VRAM 48GB | RunPod | $1.19 | On-Demand | global |
L40S VRAM 48GB | Vast.ai | $0.72 | On-Demand | global |
NVIDIA B200(192GB Blackwell) | ||||
B200 VRAM 192GB | AWS | $5.45 (8x node @ $43.60/hr) | On-Demand | us-east-1 |
B200 VRAM 192GB | Azure | $5.50 (8x node @ $44.00/hr) | On-Demand | eastus |
B200 VRAM 192GB | CoreWeave | $4.15 | On-Demand | us-east |
B200 VRAM 192GB | Lambda | $4.35 | On-Demand | us-tx-1 |
B200 VRAM 192GB | Nebius | $3.95 | On-Demand | eu-west-1 |
Frequently asked questions
How much does an H100 GPU cost per hour to rent in the cloud?
As of August 2026, renting an NVIDIA H100 (80GB) GPU costs a median of $2.49/hour on-demand. Rates vary significantly by provider tier: specialist clouds (like Lambda and CoreWeave) charge $2.15 to $2.85/GPU-hour, hyperscalers (AWS, Azure, Google Cloud) charge $3.70 to $3.98/GPU-hour, and unreserved marketplaces drop as low as $1.49/GPU-hour.
Why do cloud GPU prices vary so much across different providers?
What is the cheapest way to rent an NVIDIA H100 or H200 GPU?
What is the difference between on-demand, spot, and reserved GPU pricing?
Are the prices listed on this page per individual GPU or per server instance node?
When does buying dedicated bare-metal GPU servers beat renting hourly cloud instances?
Start with AWS 
Start with GCP 
Start with Azure