LLM prices / GPU prices
GPU rental prices · 17 providers · checked Oct 24, 2026
GPU rental prices per hour: H100, H200, B200, A100, L40S
The cheapest no-commitment price per GPU-hour at 17 clouds, read from their public price pages and APIs every day. Price your job below, then open the provider's page to rent.
H100$1.99an hour on RunPod (PCIe 80GB). Median of 17 providers: $4.41.H200$3.59an hour on RunPod (SXM 141GB). Median of 15 providers: $5.40.B200$5.43an hour on Massed Compute (SXM 180GB). Median of 11 providers: $8.19.A100$0.51an hour on Vast.ai (PCIe 40GB). Median of 13 providers: $1.99.L40S$0.79an hour on RunPod (48GB). Median of 10 providers: $1.56.
Short answerAn H100 costs from $1.99 an hour (RunPod) to $11.06 (Google Cloud) without a commitment, a 5.6x spread for the same chip. Big clouds (AWS, Google, Azure, Oracle) sell mostly 8-GPU machines and charge the most per GPU; specialist GPU clouds and marketplaces are cheaper but vary in reliability, networking and capacity.
| Provider | Option | Per GPU-hour | Your job |
|---|
Cheapest price per provider
Cheapest no-commitment price per GPU-hour, per card. Hover a price for the exact variant and instance; click it for the provider's page. thin = fewer than 5 offers on the marketplace that day.
| Provider | H100 | H200 | B200 | A100 | L40S |
|---|---|---|---|---|---|
| AWS EC2 | $6.88SXM 80GB | $7.91SXM 141GB | $14.24SXM 180GB | $2.74SXM 40GB | $1.8648GB |
| CoreWeave | $6.16SXM 80GB | $6.30SXM 141GB | $8.60SXM 180GB | $2.70SXM 80GB | $2.2548GB |
| Crusoe | $3.90SXM 80GB | $4.29SXM 141GB | n/a | $2.00PCIe 80GB | $1.5048GB |
| DigitalOcean | $4.41SXM 80GB | $4.47SXM 141GB | n/a | n/a | $1.5748GB |
| Google Cloud | $11.06SXM 80GB | $10.60SXM 141GB | n/a | $3.48SXM 40GB | n/a |
| Hyperstack | $2.50PCIe 80GB | $3.99SXM 141GB | $6.00SXM 192GB | $1.35PCIe 80GB | n/a |
| Jarvislabs | $3.49SXM 80GB | $4.59SXM 141GB | n/a | $0.8940GB | n/a |
| Lambda | $3.29PCIe 80GB | n/a | $6.69SXM 180GB | $1.99PCIe 40GB | n/a |
| Massed Compute | $2.73PCIe 80GB | $3.62NVL 141GB | $5.43SXM 180GB | $1.35PCIe 80GB | $0.9748GB |
| Microsoft Azure | $6.98NVL 94GB | $10.60SXM 141GB | n/a | $3.40SXM 40GB | n/a |
| Nebius | $4.50SXM 80GB | $5.40SXM 141GB | $8.50SXM 180GB | n/a | $1.5548GB (Intel host) |
| Oracle Cloud (OCI) | $10.00SXM 80GB | $10.00SXM 141GB | $14.00SXM 180GB | $4.00SXM 80GB | $3.5048GB |
| Paperspace (DigitalOcean) | $5.95SXM 80GB | n/a | n/a | n/a | n/a |
| RunPod | $1.99PCIe 80GB · community | $3.59SXM 141GB · community | $5.98SXM 180GB · community | $1.00SXM 40GB · community | $0.7948GB · community |
| Together AI | $3.99SXM 80GB | $5.99SXM 141GB | $8.19SXM 180GB | n/a | n/a |
| Vast.ai | $2.14SXM 80GB · marketplace | $4.00NVL 141GB · marketplace | $7.50SXM 180GB · marketplace | $0.51PCIe 40GB · marketplace | $0.8048GB · marketplace |
| Verda (DataCrunch) | $4.51SXM 80GB | $5.89SXM 141GB | $8.36SXM 180GB | $1.56SXM 40GB | $1.8148GB |
Price history
Cheapest and median no-commitment price per GPU-hour, one point a day since we started on Oct 24, 2026 (1 saved snapshot so far). Updates daily.
The chart fills in as daily snapshots arrive.
Every price we read
Every row from today's check, including spot, reserved and the marketplace's lowest single offer. JSON · CSV.
| Provider | GPU | Per GPU-hour | Pricing | Instance | Region | Source |
|---|---|---|---|---|---|---|
| Vast.ai | H100 SXM 80GB | $1.79 | marketplace (low) | H100 SXM | api | |
| RunPod | H100 PCIe 80GB | $1.99 | community | H100 PCIe x1 | api | |
| Together AI | H100 SXM 80GB | $1.99 | spot | HGX H100 cluster | html | |
| Hyperstack | H100 PCIe 80GB | $2.00 | spot | NVIDIA H100 PCIe x1 | html | |
| Vast.ai | H100 PCIe 80GB | $2.00 | marketplace (low) | H100 PCIE | api | |
| Microsoft Azure | H100 SXM 80GB | $2.04 | spot | Standard_ND96is_H100_v5 x8 | eastus2 | api |
| Vast.ai | H100 SXM 80GB | $2.14 | marketplace (median) | H100 SXM | api | |
| Verda (DataCrunch) | H100 SXM 80GB | $2.26 | spot | 1H100.80S.30V x1 | api | |
| Vast.ai | H100 PCIe 80GB | $2.27 | marketplace (median) | H100 PCIE | api | |
| Microsoft Azure | H100 SXM 80GB | $2.27 | spot | Standard_ND96isr_H100_v5 x8 | eastus2 | api |
| CoreWeave | H100 SXM 80GB | $2.46 | spot | NVIDIA HGX H100 x8 | North America | html |
| Hyperstack | H100 PCIe 80GB | $2.50 | on-demand | NVIDIA H100 x1 | html | |
| RunPod | H100 NVL 94GB | $2.59 | community | H100 NVL x1 | api | |
| Hyperstack | H100 PCIe NVLink 80GB | $2.60 | on-demand | NVIDIA H100 NVLink x1 | html | |
| RunPod | H100 SXM 80GB | $2.69 | community | H100 SXM x1 | api | |
| Massed Compute | H100 PCIe 80GB | $2.73 | on-demand | 1x H100 (80GB) x1 | html | |
| RunPod | H100 PCIe 80GB | $2.89 | secure | H100 PCIe x1 | api | |
| Massed Compute | H100 NVL 94GB NVLink | $2.92 | on-demand | 2x H100 NVL (94GB) NVLink x2 | html | |
| Vast.ai | H100 NVL 94GB | $3.03 | marketplace (low) | H100 NVL | api | |
| Massed Compute | H100 NVL 94GB | $3.11 | on-demand | 1x H100 NVL x1 | html | |
| Massed Compute | H100 SXM 80GB | $3.14 | on-demand | 1x H100 SXM5 (80GB) x1 | html | |
| Vast.ai | H100 NVL 94GB | $3.18 | marketplace (median) | H100 NVL | api | |
| RunPod | H100 NVL 94GB | $3.19 | secure | H100 NVL x1 | api | |
| Hyperstack | H100 SXM 80GB | $3.20 | on-demand | NVIDIA H100 SXM x1 | html | |
| Microsoft Azure | H100 NVL 94GB | $3.22 | spot | Standard_NC40ads_H100_v5 x1 | eastus2 | api |
| Microsoft Azure | H100 NVL 94GB | $3.22 | spot | Standard_NC80adis_H100_v5 x2 | eastus2 | api |
| DigitalOcean | H100 SXM 80GB | $3.26 | reserved-1y | GPU Droplet HGX H100 | html | |
| Lambda | H100 PCIe 80GB | $3.29 | on-demand | 1x H100 PCIe x1 | html | |
| Jarvislabs | H100 SXM 80GB | $3.49 | on-demand | NVIDIA H100 SXM x1 | html | |
| Crusoe | H100 SXM 80GB | $3.90 | on-demand | NVIDIA H100 80GB HGX | html | |
| RunPod | H100 SXM 80GB | $3.99 | secure | H100 SXM x1 | api | |
| Together AI | H100 SXM 80GB | $3.99 | on-demand | HGX H100 cluster | html | |
| Lambda | H100 SXM 80GB | $3.99 | on-demand | 8x H100 SXM x8 | html | |
| Lambda | H100 SXM 80GB | $4.09 | on-demand | 4x H100 SXM x4 | html | |
| Lambda | H100 SXM 80GB | $4.19 | on-demand | 2x H100 SXM x2 | html | |
| Lambda | H100 SXM 80GB | $4.29 | on-demand | 1x H100 SXM x1 | html | |
| DigitalOcean | H100 SXM 80GB | $4.41 | on-demand | GPU Droplet HGX H100 | html | |
| Nebius | H100 SXM 80GB | $4.50 | on-demand | NVIDIA HGX H100 | html | |
| Verda (DataCrunch) | H100 SXM 80GB | $4.51 | on-demand | 1H100.80S.30V x1 | api | |
| Paperspace (DigitalOcean) | H100 SXM 80GB | $5.95 | on-demand | H100 x1 | manual | |
| Google Cloud | H100 SXM 80GB | $5.99 | spot | a3-megagpu-8g x8 | us-central1 | html |
| CoreWeave | H100 SXM 80GB | $6.16 | on-demand | NVIDIA HGX H100 x8 | North America | html |
| Google Cloud | H100 SXM 80GB | $6.27 | spot | a3-highgpu-8g x8 | us-central1 | html |
| AWS EC2 | H100 SXM 80GB | $6.88 | on-demand | p5.48xlarge x8 | us-east-1 | api |
| AWS EC2 | H100 SXM 80GB | $6.88 | on-demand | p5.4xlarge x1 | us-east-1 | api |
| Microsoft Azure | H100 NVL 94GB | $6.98 | on-demand | Standard_NC80adis_H100_v5 x2 | eastus2 | api |
| Microsoft Azure | H100 NVL 94GB | $6.98 | on-demand | Standard_NC40ads_H100_v5 x1 | eastus2 | api |
| Google Cloud | H100 SXM 80GB | $7.67 | reserved-1y | a3-highgpu-8g x8 | us-central1 | html |
| Google Cloud | H100 SXM 80GB | $8.03 | reserved-1y | a3-megagpu-8g x8 | us-central1 | html |
| Oracle Cloud (OCI) | H100 SXM 80GB | $10.00 | on-demand | BM.GPU.H100.8 x8 | api | |
| Microsoft Azure | H100 SXM 80GB | $11.06 | on-demand | Standard_ND96is_H100_v5 x8 | eastus2 | api |
| Google Cloud | H100 SXM 80GB | $11.06 | on-demand | a3-highgpu-8g x8 | us-central1 | html |
| Google Cloud | H100 SXM 80GB | $11.68 | on-demand | a3-megagpu-8g x8 | us-central1 | html |
| Microsoft Azure | H100 SXM 80GB | $12.29 | on-demand | Standard_ND96isr_H100_v5 x8 | eastus2 | api |
| CoreWeave | H200 SXM 141GB | $2.62 | spot | NVIDIA HGX H200 x8 | North America | html |
| Verda (DataCrunch) | H200 SXM 141GB | $2.94 | spot | 1H200.141S.44V x1 | api | |
| Together AI | H200 SXM 141GB | $2.99 | spot | HGX H200 cluster | html | |
| DigitalOcean | H200 SXM 141GB | $3.40 | reserved-1y | GPU Droplet HGX H200 | html | |
| Vast.ai | H200 NVL 141GB | $3.44 | marketplace (low) | H200 NVL | api | |
| RunPod | H200 SXM 141GB | $3.59 | community | H200 SXM x1 | api | |
| Massed Compute | H200 NVL 141GB | $3.62 | on-demand | 1x H200 NVL (141GB) x1 | html | |
| RunPod | H200 NVL 143GB | $3.79 | secure | H200 NVL x1 | api | |
| Massed Compute | H200 NVL 141GB NVLink | $3.82 | on-demand | 4x H200 NVL (141GB) NVLink x4 | html | |
| Hyperstack | H200 SXM 141GB | $3.99 | on-demand | NVIDIA H200 SXM x1 | html | |
| Vast.ai | H200 NVL 141GB | $4.00 | marketplace (median) | H200 NVL | api | |
| Crusoe | H200 SXM 141GB | $4.29 | on-demand | NVIDIA H200 141GB HGX | html | |
| DigitalOcean | H200 SXM 141GB | $4.47 | on-demand | GPU Droplet HGX H200 | html | |
| Jarvislabs | H200 SXM 141GB | $4.59 | on-demand | NVIDIA H200 SXM x1 | html | |
| RunPod | H200 SXM 141GB | $5.29 | secure | H200 SXM x1 | api | |
| Nebius | H200 SXM 141GB | $5.40 | on-demand | NVIDIA HGX H200 | html | |
| Verda (DataCrunch) | H200 SXM 141GB | $5.89 | on-demand | 1H200.141S.44V x1 | api | |
| Together AI | H200 SXM 141GB | $5.99 | on-demand | HGX H200 cluster | html | |
| CoreWeave | H200 SXM 141GB | $6.30 | on-demand | NVIDIA HGX H200 x8 | North America | html |
| Google Cloud | H200 SXM 141GB | $6.35 | spot | a3-ultragpu-8g x8 | us-central1 | html |
| Vast.ai | H200 SXM 141GB | $6.58 | marketplace (median) | H200 | api | |
| Vast.ai | H200 SXM 141GB | $6.58 | marketplace (low) | H200 | api | |
| Google Cloud | H200 SXM 141GB | $7.31 | reserved-1y | a3-ultragpu-8g x8 | us-central1 | html |
| AWS EC2 | H200 SXM 141GB | $7.91 | on-demand | p5en.48xlarge x8 | us-east-1 | api |
| Oracle Cloud (OCI) | H200 SXM 141GB | $10.00 | on-demand | BM.GPU.H200.8 x8 | api | |
| Microsoft Azure | H200 SXM 141GB | $10.60 | on-demand | Standard_ND96isr_H200_v5 x8 | eastus2 | api |
| Microsoft Azure | H200 SXM 141GB | $10.60 | spot | Standard_ND96isr_H200_v5 x8 | eastus2 | api |
| Google Cloud | H200 SXM 141GB | $10.60 | on-demand | a3-ultragpu-8g x8 | us-central1 | html |
| Together AI | B200 SXM 180GB | $4.09 | spot | HGX B200 cluster | html | |
| Verda (DataCrunch) | B200 SXM 180GB | $4.18 | spot | 1B200.30V x1 | api | |
| CoreWeave | B200 SXM 180GB | $4.26 | spot | NVIDIA HGX B200 x8 | North America | html |
| Google Cloud | B200 SXM 180GB | $4.95 | spot | a4-highgpu-8g x8 | us-central1 | html |
| Massed Compute | B200 SXM 180GB | $5.43 | on-demand | 8x B200 SXM6 x8 | html | |
| RunPod | B200 SXM 180GB | $5.98 | community | B200 x1 | api | |
| Hyperstack | B200 SXM 192GB | $6.00 | on-demand | NVIDIA B200 x1 | html | |
| Lambda | B200 SXM 180GB | $6.69 | on-demand | 8x B200 SXM6 x8 | html | |
| Lambda | B200 SXM 180GB | $6.79 | on-demand | 4x B200 SXM6 x4 | html | |
| Lambda | B200 SXM 180GB | $6.89 | on-demand | 2x B200 SXM6 x2 | html | |
| Lambda | B200 SXM 180GB | $6.99 | on-demand | 1x B200 SXM6 x1 | html | |
| Vast.ai | B200 SXM 180GB | $7.50 | marketplace (low) | B200 | api | |
| Vast.ai | B200 SXM 180GB | $7.50 | marketplace (median) | B200 | api | |
| RunPod | B200 SXM 180GB | $7.99 | secure | B200 x1 | api | |
| Together AI | B200 SXM 180GB | $8.19 | on-demand | HGX B200 cluster | html | |
| Verda (DataCrunch) | B200 SXM 180GB | $8.36 | on-demand | 1B200.30V x1 | api | |
| Nebius | B200 SXM 180GB | $8.50 | on-demand | NVIDIA HGX B200 | html | |
| CoreWeave | B200 SXM 180GB | $8.60 | on-demand | NVIDIA HGX B200 x8 | North America | html |
| CoreWeave | B200 GB200 NVL72 | $10.50 | on-demand | NVIDIA GB200 NVL72 x4 | North America | html |
| Google Cloud | B200 SXM 180GB | $11.12 | reserved-1y | a4-highgpu-8g x8 | us-central1 | html |
| Oracle Cloud (OCI) | B200 SXM 180GB | $14.00 | on-demand | BM.GPU.B200.8 x8 | api | |
| AWS EC2 | B200 SXM 180GB | $14.24 | on-demand | p6-b200.48xlarge x8 | us-east-1 | api |
| Oracle Cloud (OCI) | B200 GB200 NVL72 | $16.00 | on-demand | BM.GPU.GB200.4 x4 | api | |
| Microsoft Azure | B200 GB200 NVL72 | $27.04 | spot | Standard_ND128isr_NDR_GB200_v6 x4 | eastus2 | api |
| Microsoft Azure | B200 GB200 NVL72 | $27.04 | on-demand | Standard_ND128isr_NDR_GB200_v6 x4 | eastus2 | api |
| Vast.ai | A100 PCIe 40GB | $0.40 | marketplace (low) | A100 PCIE | api | |
| Vast.ai | A100 SXM 40GB | $0.41 | marketplace (low) | A100 SXM4 | api | |
| Vast.ai | A100 PCIe 40GB | $0.51 | marketplace (median) | A100 PCIE | api | |
| Vast.ai | A100 SXM 40GB | $0.53 | marketplace (median) | A100 SXM4 | api | |
| Microsoft Azure | A100 PCIe 80GB | $0.68 | spot | Standard_NC48ads_A100_v4 x2 | eastus2 | api |
| Microsoft Azure | A100 PCIe 80GB | $0.68 | spot | Standard_NC24ads_A100_v4 x1 | eastus2 | api |
| Microsoft Azure | A100 PCIe 80GB | $0.68 | spot | Standard_NC96ads_A100_v4 x4 | eastus2 | api |
| Microsoft Azure | A100 SXM 40GB | $0.74 | spot | Standard_ND96asr_v4 x8 | eastus | api |
| Verda (DataCrunch) | A100 SXM 40GB | $0.78 | spot | 1A100.40S.22V x1 | api | |
| Vast.ai | A100 PCIe 80GB | $0.87 | marketplace (low) | A100 PCIE | api | |
| Jarvislabs | A100 40GB | $0.89 | on-demand | NVIDIA A100 40GB x1 | html | |
| Vast.ai | A100 SXM 80GB | $0.94 | marketplace (low) | A100 SXM4 | api | |
| RunPod | A100 SXM 40GB | $1.00 | community | A100 SXM 40GB x1 | api | |
| Vast.ai | A100 PCIe 80GB | $1.00 | marketplace (median) | A100 PCIE | api | |
| Microsoft Azure | A100 SXM 80GB | $1.06 | spot | Standard_ND96amsr_A100_v4 x8 | eastus2 | api |
| Verda (DataCrunch) | A100 SXM 80GB | $1.06 | spot | 1A100.22V x1 | api | |
| Hyperstack | A100 PCIe 80GB | $1.08 | spot | NVIDIA A100 x1 | html | |
| Vast.ai | A100 SXM 80GB | $1.12 | marketplace (median) | A100 SXM4 | api | |
| RunPod | A100 PCIe 80GB | $1.19 | community | A100 PCIe x1 | api | |
| CoreWeave | A100 SXM 80GB | $1.21 | spot | NVIDIA A100 x8 | North America | html |
| Hyperstack | A100 PCIe 80GB | $1.35 | on-demand | NVIDIA A100 x1 | html | |
| Massed Compute | A100 PCIe 80GB | $1.35 | on-demand | 1x A100 (80GB) x1 | html | |
| Massed Compute | A100 SXM 80GB | $1.38 | on-demand | 1x A100 SXM4 (80GB) x1 | html | |
| Massed Compute | A100 SXM 80GB (DGX) | $1.38 | on-demand | 1x DGX A100 (80GB) x1 | html | |
| RunPod | A100 SXM 80GB | $1.39 | community | A100 SXM x1 | api | |
| Hyperstack | A100 PCIe NVLink 80GB | $1.40 | on-demand | NVIDIA A100 NVLink x1 | html | |
| Jarvislabs | A100 80GB | $1.49 | on-demand | NVIDIA A100 80GB x1 | html | |
| Verda (DataCrunch) | A100 SXM 40GB | $1.56 | on-demand | 1A100.40S.22V x1 | api | |
| Hyperstack | A100 SXM 80GB | $1.60 | on-demand | NVIDIA A100 SXM x1 | html | |
| RunPod | A100 PCIe 80GB | $1.79 | secure | A100 PCIe x1 | api | |
| RunPod | A100 SXM 80GB | $1.79 | secure | A100 SXM x1 | api | |
| Lambda | A100 PCIe 40GB | $1.99 | on-demand | 2x A100 PCIe x2 | html | |
| Lambda | A100 PCIe 40GB | $1.99 | on-demand | 4x A100 PCIe x4 | html | |
| Lambda | A100 SXM 40GB | $1.99 | on-demand | 8x A100 SXM x8 | html | |
| Lambda | A100 PCIe 40GB | $1.99 | on-demand | 1x A100 PCIe x1 | html | |
| Lambda | A100 SXM 40GB | $1.99 | on-demand | 1x A100 SXM x1 | html | |
| Crusoe | A100 PCIe 80GB | $2.00 | on-demand | NVIDIA A100 80GB PCIe | html | |
| Google Cloud | A100 SXM 40GB | $2.09 | spot | a2-megagpu-16g x16 | us-central1 | html |
| Verda (DataCrunch) | A100 SXM 80GB | $2.13 | on-demand | 1A100.22V x1 | api | |
| Google Cloud | A100 SXM 40GB | $2.19 | reserved-1y | a2-megagpu-16g x16 | us-central1 | html |
| Google Cloud | A100 SXM 40GB | $2.20 | spot | a2-highgpu-8g x8 | us-central1 | html |
| Google Cloud | A100 SXM 40GB | $2.20 | spot | a2-highgpu-2g x2 | us-central1 | html |
| Google Cloud | A100 SXM 40GB | $2.20 | spot | a2-highgpu-1g x1 | us-central1 | html |
| Google Cloud | A100 SXM 40GB | $2.20 | spot | a2-highgpu-4g x4 | us-central1 | html |
| Crusoe | A100 SXM 80GB | $2.30 | on-demand | NVIDIA A100 80GB SXM | html | |
| Google Cloud | A100 SXM 40GB | $2.31 | reserved-1y | a2-highgpu-8g x8 | us-central1 | html |
| Google Cloud | A100 SXM 40GB | $2.31 | reserved-1y | a2-highgpu-4g x4 | us-central1 | html |
| Google Cloud | A100 SXM 40GB | $2.31 | reserved-1y | a2-highgpu-2g x2 | us-central1 | html |
| Google Cloud | A100 SXM 40GB | $2.31 | reserved-1y | a2-highgpu-1g x1 | us-central1 | html |
| CoreWeave | A100 SXM 80GB | $2.70 | on-demand | NVIDIA A100 x8 | North America | html |
| AWS EC2 | A100 SXM 40GB | $2.74 | on-demand | p4d.24xlarge x8 | us-east-1 | api |
| Lambda | A100 SXM 80GB | $2.79 | on-demand | 8x A100 SXM x8 | html | |
| Google Cloud | A100 SXM 80GB | $3.04 | spot | a2-ultragpu-8g x8 | us-central1 | html |
| Google Cloud | A100 SXM 80GB | $3.04 | spot | a2-ultragpu-2g x2 | us-central1 | html |
| Google Cloud | A100 SXM 80GB | $3.04 | spot | a2-ultragpu-1g x1 | us-central1 | html |
| Google Cloud | A100 SXM 80GB | $3.04 | spot | a2-ultragpu-4g x4 | us-central1 | html |
| Microsoft Azure | A100 SXM 40GB | $3.40 | on-demand | Standard_ND96asr_v4 x8 | eastus | api |
| AWS EC2 | A100 SXM 80GB | $3.43 | on-demand | p4de.24xlarge x8 | us-east-1 | api |
| Google Cloud | A100 SXM 40GB | $3.48 | on-demand | a2-megagpu-16g x16 | us-central1 | html |
| Microsoft Azure | A100 PCIe 80GB | $3.67 | on-demand | Standard_NC48ads_A100_v4 x2 | eastus2 | api |
| Microsoft Azure | A100 PCIe 80GB | $3.67 | on-demand | Standard_NC24ads_A100_v4 x1 | eastus2 | api |
| Microsoft Azure | A100 PCIe 80GB | $3.67 | on-demand | Standard_NC96ads_A100_v4 x4 | eastus2 | api |
| Google Cloud | A100 SXM 40GB | $3.67 | on-demand | a2-highgpu-1g x1 | us-central1 | html |
| Google Cloud | A100 SXM 40GB | $3.67 | on-demand | a2-highgpu-8g x8 | us-central1 | html |
| Google Cloud | A100 SXM 40GB | $3.67 | on-demand | a2-highgpu-4g x4 | us-central1 | html |
| Google Cloud | A100 SXM 40GB | $3.67 | on-demand | a2-highgpu-2g x2 | us-central1 | html |
| Oracle Cloud (OCI) | A100 SXM 80GB | $4.00 | on-demand | BM.GPU.A100-v2.8 x8 | api | |
| Microsoft Azure | A100 SXM 80GB | $4.10 | on-demand | Standard_ND96amsr_A100_v4 x8 | eastus2 | api |
| Google Cloud | A100 SXM 80GB | $4.20 | reserved-1y | a2-ultragpu-1g x1 | us-central1 | html |
| Google Cloud | A100 SXM 80GB | $5.07 | on-demand | a2-ultragpu-4g x4 | us-central1 | html |
| Google Cloud | A100 SXM 80GB | $5.07 | on-demand | a2-ultragpu-8g x8 | us-central1 | html |
| Google Cloud | A100 SXM 80GB | $5.07 | on-demand | a2-ultragpu-1g x1 | us-central1 | html |
| Google Cloud | A100 SXM 80GB | $5.07 | on-demand | a2-ultragpu-2g x2 | us-central1 | html |
| Vast.ai | L40S 48GB | $0.53 | marketplace (low) | L40S | api | |
| RunPod | L40S 48GB | $0.79 | community | L40S x1 | api | |
| Vast.ai | L40S 48GB | $0.80 | marketplace (median) | L40S | api | |
| Verda (DataCrunch) | L40S 48GB | $0.90 | spot | 1L40S.20V x1 | api | |
| Massed Compute | L40S 48GB | $0.97 | on-demand | 1x L40S (48GB) x1 | html | |
| CoreWeave | L40S 48GB | $0.98 | spot | NVIDIA L40S x8 | North America | html |
| RunPod | L40S 48GB | $1.09 | secure | L40S x1 | api | |
| Crusoe | L40S 48GB | $1.50 | on-demand | NVIDIA L40S 48GB | html | |
| Nebius | L40S 48GB (Intel host) | $1.55 | on-demand | NVIDIA L40S with Intel CPU | html | |
| DigitalOcean | L40S 48GB | $1.57 | on-demand | GPU Droplet L40S | html | |
| Verda (DataCrunch) | L40S 48GB | $1.81 | on-demand | 1L40S.20V x1 | api | |
| Nebius | L40S 48GB (AMD host) | $1.82 | on-demand | NVIDIA L40S with AMD CPU | html | |
| AWS EC2 | L40S 48GB | $1.86 | on-demand | g6e.xlarge x1 | us-east-1 | api |
| AWS EC2 | L40S 48GB | $2.24 | on-demand | g6e.2xlarge x1 | us-east-1 | api |
| CoreWeave | L40S 48GB | $2.25 | on-demand | NVIDIA L40S x8 | North America | html |
| AWS EC2 | L40S 48GB | $2.62 | on-demand | g6e.12xlarge x4 | us-east-1 | api |
| AWS EC2 | L40S 48GB | $3.00 | on-demand | g6e.4xlarge x1 | us-east-1 | api |
| Oracle Cloud (OCI) | L40S 48GB | $3.50 | on-demand | BM.GPU.L40S.4 x4 | api | |
| AWS EC2 | L40S 48GB | $3.77 | on-demand | g6e.24xlarge x4 | us-east-1 | api |
| AWS EC2 | L40S 48GB | $3.77 | on-demand | g6e.48xlarge x8 | us-east-1 | api |
| AWS EC2 | L40S 48GB | $4.53 | on-demand | g6e.8xlarge x1 | us-east-1 | api |
| AWS EC2 | L40S 48GB | $7.58 | on-demand | g6e.16xlarge x1 | us-east-1 | api |
Method
Once a day a server reads each provider's public price page or the public API behind it, with no account and no key. Every price is per GPU per hour in US dollars: an 8-GPU machine's hourly price is divided by 8. "No commitment" means on-demand, plus Vast.ai's marketplace median and RunPod's community and secure clouds; spot and reserved prices are listed but kept out of the headline numbers. For Vast.ai we take the median (and separately the lowest) of verified, rentable on-demand offers. Big clouds are priced in one region (AWS us-east-1, Google us-central1, Azure East US 2). B200 excludes GB200 superchip systems. A price that fails a sanity range is dropped, never guessed.
What a price doesn't tell you: whether capacity is actually available, networking between GPUs (InfiniBand matters for multi-node training), storage and egress fees, and minimum rental size. Check those on the provider's page before you commit.
Sources
- RunPod: public API
- Lambda: price page
- Vast.ai: public marketplace API (median of verified on-demand offers)
- CoreWeave: price page
- AWS EC2: public price file, us-east-1
- Google Cloud: price page, us-central1
- Microsoft Azure: Retail Prices API, East US 2
- Oracle (OCI): public price list API
- Together AI: price page
- Nebius: price page
- Hyperstack: price page
- DigitalOcean: price page
- Jarvislabs: price page
- Crusoe: price page
- Massed Compute: price page
- Verda (DataCrunch): public API
- Paperspace: read by hand
Up next
Before you rent: Pick a model and quantization; it tells you which of these cards fits it, and how many.LLM VRAM calculatorWill the model fit your GPU? GB for weights + KV cache on 24 open models, and which cards fit.Read it next →