Price spread · 38 GPUs listed by 3+ providers · November 4, 2026
Price spread: the same GPU, different providers
4.0×the median gap between the dearest and the cheapest listed on-demand rate for the same card
Listed on-demand rate cards, each provider once at its cheapest region; marketplace floors, community capacity and shared GPUs are out. About this page ↓
- Widest12.8×V100: $0.24 Verda to $3.06 Microsoft Azure
- H100 SXM middle half$2.99–$4.4125th to 75th percentile, 41 providers
- Hyperscaler premium+109%median, over cards both kinds list
| GPU | Cheapest | Median | Middle half (25th–75th percentile) | Dearest | Dearest ÷ cheapest | Hyperscaler premium | Providers |
|---|---|---|---|---|---|---|---|
| NVIDIA V100 32GB | $0.16 Geodd · | $1.15 | $0.33–$1.45 | $3.20 Yandex Cloud · GPU only · | 20.0× | — | 11 |
| NVIDIA V100 | $0.24 Verda · | $0.83 | $0.32–$2.95 | $3.06 Microsoft Azure · eastus · | 12.8× | +759% | 13 |
| NVIDIA A10 | $0.34 Intelion Cloud · GPU only · | $1.22 | $1.10–$1.42 | $3.20Microsoft Azure · westus3 · | 9.5× | +136% | 9 |
| NVIDIA RTX A4000 | $0.12 GPU.ai · from · | $0.24 | $0.18–$0.53 | $1.10 airgpu · North America · | 9.2× | — | 11 |
| NVIDIA RTX PRO 6000 Blackwell | $0.65 Hostinger · from · | $2.20 | $1.90–$2.90 | $5.50Microsoft Azure · eastus2 · | 8.5× | +105% | 42 |
| NVIDIA A100 80GB | $0.92 Yotta Labs · | $1.99 | $1.49–$2.61 | $6.82 Servercore · | 7.4× | +110% | 63 |
| NVIDIA RTX A6000 | $0.35 Thunder Compute · | $0.59 | $0.50–$0.75 | $2.50 RunComfy · | 7.1× | — | 21 |
| NVIDIA H100 SXM | $1.79 Cudo Compute · no-kristiansand-1, GPU only · | $3.29 | $2.99–$4.41 | $11.06Microsoft Azure · eastus · | 6.2× | +209% | 41 |
| NVIDIA A30 | $0.22 CloudGPU · | $0.41 | $0.33–$0.68 | $1.35 E2E Networks · | 6.1× | — | 9 |
| NVIDIA H100 80GB (form factor not stated) | $1.72 Gcore · Everywhere Inference · | $3.75 | $2.70–$4.55 | $10.46 Cloud.ru Evolution · | 6.1× | — | 28 |
| NVIDIA RTX A5000 | $0.27GPU.ai · from · | $0.58 | $0.44–$1.31 | $1.56 Sesterce · from · | 5.8× | — | 10 |
| NVIDIA L40S | $0.63 CloudRift · | $1.42 | $0.99–$1.95 | $3.54 Lightning AI · | 5.6× | +109% | 40 |
| NVIDIA A40 | $0.39CloudGPU · | $0.64 | $0.59–$1.45 | $2.05Sesterce · from · | 5.3× | — | 8 |
| NVIDIA A100 40GB | $0.69 Trooper.AI · from · | $1.99 | $1.10–$2.61 | $3.40Microsoft Azure · westus2 · | 4.9× | +65% | 22 |
| NVIDIA Quadro RTX 6000 | $0.31 centron · GPU only · | $0.51 | $0.32–$0.89 | $1.50 Akamai Cloud (Linode) · | 4.8× | — | 4 |
| NVIDIA H200 | $2.29 Dataoorts · Orion Series · | $4.54 | $3.54–$5.99 | $10.60Microsoft Azure · westus2 · | 4.6× | +115% | 56 |
| NVIDIA GeForce RTX 3090 | $0.18 OpenRelay · | $0.42 | $0.31–$0.50 | $0.77 immers.cloud · | 4.3× | — | 10 |
| NVIDIA L4 | $0.28 Hikube · GPU only · | $0.80 | $0.58–$0.85 | $1.17 Neysa · from · | 4.2× | +-15% | 23 |
| NVIDIA T4 | $0.24immers.cloud · | $0.55 | $0.47–$0.65 | $0.99 Vagon · | 4.2× | +-11% | 19 |
| NVIDIA GB200 NVL72 | $6.94 Bitdeer AI · from · | $13.25 | $8.63–$16.70 | $27.04Microsoft Azure · westus3 · | 3.9× | +133% | 6 |
| NVIDIA B200 | $4.50Hostinger · from · | $6.94 | $6.00–$8.53 | $16.11 Google Cloud · us-west1, GPU only · | 3.6× | +109% | 40 |
| AMD Instinct MI300X | $1.71 TensorWave · from · | $2.65 | $2.49–$4.50 | $6.00 Oracle Cloud (OCI) · | 3.5× | +132% | 7 |
| NVIDIA RTX PRO 4500 Blackwell | $0.72GPU.ai · from · | $0.86 | $0.74–$1.28 | $2.52 Amazon Web Services · us-east-1 · | 3.5× | — | 6 |
| NVIDIA H100 PCIe | $1.98GPU.ai · from · | $2.82 | $2.52–$3.09 | $6.72Cloud.ru Evolution · | 3.4× | — | 10 |
| NVIDIA RTX 5090 | $0.50OpenRelay · | $0.79 | $0.66–$1.19 | $1.60 EmpirioLabs · | 3.2× | — | 23 |
| NVIDIA RTX 4000 Ada | $0.28GPU.ai · from · | $0.52 | $0.28–$0.76 | $0.87Sesterce · from · | 3.1× | — | 5 |
| NVIDIA B300 | $5.89 Deep Infra · | $8.99 | $7.42–$11.16 | $17.80Amazon Web Services · us-east-1 · | 3.0× | +82% | 26 |
| AMD Instinct MI355X | $2.95TensorWave · from · | $6.14 | $5.23–$6.88 | $8.60Oracle Cloud (OCI) · | 2.9× | — | 4 |
| NVIDIA GH200 | $2.29 Lambda · | $3.87 | — | $6.50 CoreWeave · | 2.8× | — | 3 |
| NVIDIA RTX 6000 Ada | $0.65GPU.ai · from · | $1.03 | $0.86–$1.51 | $1.84 SwissGPU · | 2.8× | — | 12 |
| NVIDIA RTX 4090 | $0.35OpenRelay · | $0.48 | $0.42–$0.89 | $0.99 Daytona · GPU only · | 2.8× | — | 17 |
| NVIDIA H100 NVL | $2.92 Massed Compute · | $4.20 | $3.19–$5.95 | $7.92Cloud.ru Evolution · | 2.7× | — | 8 |
| NVIDIA RTX PRO 5000 Blackwell | $0.80 Meshive · | $1.21 | $0.92–$1.50 | $1.60EmpirioLabs · | 2.0× | — | 4 |
| NVIDIA L40 | $0.78GPU.ai · from · | $0.97 | $0.84–$1.13 | $1.50EmpirioLabs · | 1.9× | — | 11 |
| AMD Instinct MI325X | $2.25 Bentaus · | $2.25 | — | $3.80 Paperspace · GPU Droplets · | 1.7× | — | 3 |
| NVIDIA H200 NVL | $3.29GPU.ai · from · | $3.62 | $3.39–$4.62 | $5.50EmpirioLabs · | 1.7× | — | 7 |
| NVIDIA A10G | $1.01Amazon Web Services · us-east-1 · | $1.13 | $1.04–$1.32 | $1.67Vagon · | 1.7× | — | 4 |
| NVIDIA GB300 NVL72 | $13.30Verda · | $18.00 | — | $20.00 Fireworks AI · GPU only · | 1.5× | — | 3 |
About this page
As of , across the 38 GPUs listed by 3 or more providers, the dearest listed on-demand rate is a median 4.0× the cheapest one for the same card. The widest is the NVIDIA V100 at 12.8×: $0.24 on Verda against $3.06 on Microsoft Azure. The cheapest and the dearest are one quote each; the middle half of rate cards is steadier: for the NVIDIA H100 SXM it sits between $2.99 and $4.41 per GPU-hour across 41 providers.
The Price Index says what an H100-hour costs. This page says how much that depends on where you rent it. For every card listed by at least 3 providers: the cheapest, median and dearest listed on-demand rate per GPU-hour, each provider counted once at its own cheapest region, and from 4 providers the middle half of those rate cards (25th to 75th percentile). Marketplace floors and community capacity are excluded — neither is a rate card you can hold.
Where both kinds of provider list a card, the hyperscaler median runs a median 109% above the specialist-cloud median.Card by card
- NVIDIA V100 32GB: $0.16 on Geodd to $3.20 on Yandex Cloud (GPU only) per GPU-hour, a factor of 20.0 across 11 providers; the median is $1.15, the middle half of rate cards sits between $0.33 and $1.45, and the cheapest rate sits 86% below the median.
- NVIDIA V100: $0.24 on Verda to $3.06 on Microsoft Azure (eastus) per GPU-hour, a factor of 12.8 across 13 providers; the median is $0.83, the middle half of rate cards sits between $0.32 and $2.95, and the cheapest rate sits 71% below the median, with hyperscalers a median 759% above specialist clouds.
- NVIDIA A10: $0.34 on Intelion Cloud (GPU only) to $3.20 on Microsoft Azure (westus3) per GPU-hour, a factor of 9.5 across 9 providers; the median is $1.22, the middle half of rate cards sits between $1.10 and $1.42, and the cheapest rate sits 73% below the median, with hyperscalers a median 136% above specialist clouds.
- NVIDIA RTX A4000: $0.12 on GPU.ai (from) to $1.10 on airgpu (North America) per GPU-hour, a factor of 9.2 across 11 providers; the median is $0.24, the middle half of rate cards sits between $0.18 and $0.53, and the cheapest rate sits 49% below the median.
- NVIDIA RTX PRO 6000 Blackwell: $0.65 on Hostinger (from) to $5.50 on Microsoft Azure (eastus2) per GPU-hour, a factor of 8.5 across 42 providers; the median is $2.20, the middle half of rate cards sits between $1.90 and $2.90, and the cheapest rate sits 70% below the median, with hyperscalers a median 105% above specialist clouds.
- NVIDIA A100 80GB: $0.92 on Yotta Labs to $6.82 on Servercore per GPU-hour, a factor of 7.4 across 63 providers; the median is $1.99, the middle half of rate cards sits between $1.49 and $2.61, and the cheapest rate sits 54% below the median, with hyperscalers a median 110% above specialist clouds.
- NVIDIA RTX A6000: $0.35 on Thunder Compute to $2.50 on RunComfy per GPU-hour, a factor of 7.1 across 21 providers; the median is $0.59, the middle half of rate cards sits between $0.50 and $0.75, and the cheapest rate sits 41% below the median.
- NVIDIA H100 SXM: $1.79 on Cudo Compute (no-kristiansand-1, GPU only) to $11.06 on Microsoft Azure (eastus) per GPU-hour, a factor of 6.2 across 41 providers; the median is $3.29, the middle half of rate cards sits between $2.99 and $4.41, and the cheapest rate sits 46% below the median, with hyperscalers a median 209% above specialist clouds.
- NVIDIA A30: $0.22 on CloudGPU to $1.35 on E2E Networks per GPU-hour, a factor of 6.1 across 9 providers; the median is $0.41, the middle half of rate cards sits between $0.33 and $0.68, and the cheapest rate sits 46% below the median.
- NVIDIA H100 80GB (form factor not stated): $1.72 on Gcore (Everywhere Inference) to $10.46 on Cloud.ru Evolution per GPU-hour, a factor of 6.1 across 28 providers; the median is $3.75, the middle half of rate cards sits between $2.70 and $4.55, and the cheapest rate sits 54% below the median.
- NVIDIA RTX A5000: $0.27 on GPU.ai (from) to $1.56 on Sesterce (from) per GPU-hour, a factor of 5.8 across 10 providers; the median is $0.58, the middle half of rate cards sits between $0.44 and $1.31, and the cheapest rate sits 53% below the median.
- NVIDIA L40S: $0.63 on CloudRift to $3.54 on Lightning AI per GPU-hour, a factor of 5.6 across 40 providers; the median is $1.42, the middle half of rate cards sits between $0.99 and $1.95, and the cheapest rate sits 56% below the median, with hyperscalers a median 109% above specialist clouds.
- NVIDIA A40: $0.39 on CloudGPU to $2.05 on Sesterce (from) per GPU-hour, a factor of 5.3 across 8 providers; the median is $0.64, the middle half of rate cards sits between $0.59 and $1.45, and the cheapest rate sits 39% below the median.
- NVIDIA A100 40GB: $0.69 on Trooper.AI (from) to $3.40 on Microsoft Azure (westus2) per GPU-hour, a factor of 4.9 across 22 providers; the median is $1.99, the middle half of rate cards sits between $1.10 and $2.61, and the cheapest rate sits 65% below the median, with hyperscalers a median 65% above specialist clouds.
- NVIDIA Quadro RTX 6000: $0.31 on centron (GPU only) to $1.50 on Akamai Cloud (Linode) per GPU-hour, a factor of 4.8 across 4 providers; the median is $0.51, the middle half of rate cards sits between $0.32 and $0.89, and the cheapest rate sits 39% below the median.
- NVIDIA H200: $2.29 on Dataoorts (Orion Series) to $10.60 on Microsoft Azure (westus2) per GPU-hour, a factor of 4.6 across 56 providers; the median is $4.54, the middle half of rate cards sits between $3.54 and $5.99, and the cheapest rate sits 50% below the median, with hyperscalers a median 115% above specialist clouds.
- NVIDIA GeForce RTX 3090: $0.18 on OpenRelay to $0.77 on immers.cloud per GPU-hour, a factor of 4.3 across 10 providers; the median is $0.42, the middle half of rate cards sits between $0.31 and $0.50, and the cheapest rate sits 57% below the median.
- NVIDIA L4: $0.28 on Hikube (GPU only) to $1.17 on Neysa (from) per GPU-hour, a factor of 4.2 across 23 providers; the median is $0.80, the middle half of rate cards sits between $0.58 and $0.85, and the cheapest rate sits 65% below the median, with hyperscalers a median -15% above specialist clouds.
- NVIDIA T4: $0.24 on immers.cloud to $0.99 on Vagon per GPU-hour, a factor of 4.2 across 19 providers; the median is $0.55, the middle half of rate cards sits between $0.47 and $0.65, and the cheapest rate sits 57% below the median, with hyperscalers a median -11% above specialist clouds.
- NVIDIA GB200 NVL72: $6.94 on Bitdeer AI (from) to $27.04 on Microsoft Azure (westus3) per GPU-hour, a factor of 3.9 across 6 providers; the median is $13.25, the middle half of rate cards sits between $8.63 and $16.70, and the cheapest rate sits 48% below the median, with hyperscalers a median 133% above specialist clouds.
- NVIDIA B200: $4.50 on Hostinger (from) to $16.11 on Google Cloud (us-west1, GPU only) per GPU-hour, a factor of 3.6 across 40 providers; the median is $6.94, the middle half of rate cards sits between $6.00 and $8.53, and the cheapest rate sits 35% below the median, with hyperscalers a median 109% above specialist clouds.
- AMD Instinct MI300X: $1.71 on TensorWave (from) to $6.00 on Oracle Cloud (OCI) per GPU-hour, a factor of 3.5 across 7 providers; the median is $2.65, the middle half of rate cards sits between $2.49 and $4.50, and the cheapest rate sits 35% below the median, with hyperscalers a median 132% above specialist clouds.
- NVIDIA RTX PRO 4500 Blackwell: $0.72 on GPU.ai (from) to $2.52 on Amazon Web Services (us-east-1) per GPU-hour, a factor of 3.5 across 6 providers; the median is $0.86, the middle half of rate cards sits between $0.74 and $1.28, and the cheapest rate sits 16% below the median.
- NVIDIA H100 PCIe: $1.98 on GPU.ai (from) to $6.72 on Cloud.ru Evolution per GPU-hour, a factor of 3.4 across 10 providers; the median is $2.82, the middle half of rate cards sits between $2.52 and $3.09, and the cheapest rate sits 30% below the median.
- NVIDIA RTX 5090: $0.50 on OpenRelay to $1.60 on EmpirioLabs per GPU-hour, a factor of 3.2 across 23 providers; the median is $0.79, the middle half of rate cards sits between $0.66 and $1.19, and the cheapest rate sits 37% below the median.
- NVIDIA RTX 4000 Ada: $0.28 on GPU.ai (from) to $0.87 on Sesterce (from) per GPU-hour, a factor of 3.1 across 5 providers; the median is $0.52, the middle half of rate cards sits between $0.28 and $0.76, and the cheapest rate sits 46% below the median.
- NVIDIA B300: $5.89 on Deep Infra to $17.80 on Amazon Web Services (us-east-1) per GPU-hour, a factor of 3.0 across 26 providers; the median is $8.99, the middle half of rate cards sits between $7.42 and $11.16, and the cheapest rate sits 34% below the median, with hyperscalers a median 82% above specialist clouds.
- AMD Instinct MI355X: $2.95 on TensorWave (from) to $8.60 on Oracle Cloud (OCI) per GPU-hour, a factor of 2.9 across 4 providers; the median is $6.14, the middle half of rate cards sits between $5.23 and $6.88, and the cheapest rate sits 52% below the median.
- NVIDIA GH200: $2.29 on Lambda to $6.50 on CoreWeave per GPU-hour, a factor of 2.8 across 3 providers; the median is $3.87, and the cheapest rate sits 41% below the median.
- NVIDIA RTX 6000 Ada: $0.65 on GPU.ai (from) to $1.84 on SwissGPU per GPU-hour, a factor of 2.8 across 12 providers; the median is $1.03, the middle half of rate cards sits between $0.86 and $1.51, and the cheapest rate sits 37% below the median.
- NVIDIA RTX 4090: $0.35 on OpenRelay to $0.99 on Daytona (GPU only) per GPU-hour, a factor of 2.8 across 17 providers; the median is $0.48, the middle half of rate cards sits between $0.42 and $0.89, and the cheapest rate sits 27% below the median.
- NVIDIA H100 NVL: $2.92 on Massed Compute to $7.92 on Cloud.ru Evolution per GPU-hour, a factor of 2.7 across 8 providers; the median is $4.20, the middle half of rate cards sits between $3.19 and $5.95, and the cheapest rate sits 31% below the median.
- NVIDIA RTX PRO 5000 Blackwell: $0.80 on Meshive to $1.60 on EmpirioLabs per GPU-hour, a factor of 2.0 across 4 providers; the median is $1.21, the middle half of rate cards sits between $0.92 and $1.50, and the cheapest rate sits 34% below the median.
- NVIDIA L40: $0.78 on GPU.ai (from) to $1.50 on EmpirioLabs per GPU-hour, a factor of 1.9 across 11 providers; the median is $0.97, the middle half of rate cards sits between $0.84 and $1.13, and the cheapest rate sits 20% below the median.
- AMD Instinct MI325X: $2.25 on Bentaus to $3.80 on Paperspace (GPU Droplets) per GPU-hour, a factor of 1.7 across 3 providers; the median is $2.25, and the cheapest rate sits 0% below the median.
- NVIDIA H200 NVL: $3.29 on GPU.ai (from) to $5.50 on EmpirioLabs per GPU-hour, a factor of 1.7 across 7 providers; the median is $3.62, the middle half of rate cards sits between $3.39 and $4.62, and the cheapest rate sits 9% below the median.
- NVIDIA A10G: $1.01 on Amazon Web Services (us-east-1) to $1.67 on Vagon per GPU-hour, a factor of 1.7 across 4 providers; the median is $1.13, the middle half of rate cards sits between $1.04 and $1.32, and the cheapest rate sits 11% below the median.
- NVIDIA GB300 NVL72: $13.30 on Verda to $20.00 on Fireworks AI (GPU only) per GPU-hour, a factor of 1.5 across 3 providers; the median is $18.00, and the cheapest rate sits 26% below the median.
How to read it
A wide spread is not automatically an arbitrage: the cheapest row is often a community tier, a GPU-only line item where CPU and memory are billed separately, or a single region, and each of those is tagged in the table. What the spread does show is that “the price of an H100-hour” is not one number, and that a quoted market rate means little without the provider next to it. The median in this table counts each provider once, at its own cheapest region, so a provider with many regions cannot weigh it down.
Cheapest rate per card by provider type: /cheapest · the level over time: Price Index · every quote behind this page: prices.json.
Cite this page
Published under CC BY 4.0: use and redistribute with attribution and a link. Cite the linked source for a price itself, and cite us for the normalization, the index and the comparison.
| Plain | GPU Economy, "Cloud GPU price spread by model", November 4, 2026. https://gpueconomy.com/spread |
|---|---|
| Markdown | [GPU Economy, "Cloud GPU price spread by model" (November 4, 2026)](https://gpueconomy.com/spread) |
| APA | GPU Economy. (2026-11-04). Cloud GPU price spread by model. Retrieved from https://gpueconomy.com/spread |
Data updated: