GPU rental prices, November 2026
An NVIDIA H100 rents for a median $4.14 per GPU-hour on demand across 4 GPU clouds (November 2026).
That is up 6.4% in 30 days and up 52.4% in 12 months. The tables list 15 GPU models at RunPod, Lambda, Nebius and CoreWeave, and on AWS, Azure and Google Cloud, as each provider lists them: per GPU, per hour, with no commitment.
GPU clouds
The median on-demand price across RunPod, Lambda, Nebius and CoreWeave, in US dollars per GPU-hour. A median needs three listings; with two, the range is shown.
costs more costs lessEach change compares the same providers in both months.
| Trend | Listings | |||||||
|---|---|---|---|---|---|---|---|---|
| NVIDIA H10080 GiB | $4.14 | $2.89 to $6.16 | up 6.4 percent over 30 days | up 16.1 percent over 90 days | up 32.9 percent over 180 days | up 52.4 percent over 12 months | 4 | |
| NVIDIA H200141 GiB | $5.40 | $5.29 to $6.30 | unchanged over 30 days | up 20.0 percent over 90 days | up 35.3 percent over 180 days | up 50.4 percent over 12 months | 3 | |
| NVIDIA B200179 GiB | $8.24 | $7.69 to $8.60 | up 7.8 percent over 30 days | up 19.1 percent over 90 days | up 35.3 percent over 180 days | up 47.4 percent over 12 months | 4 | |
| NVIDIA B300268 GiB | $8.99 to $9.50no median | $8.99 to $9.50 | up 6.3 percent over 30 days | up 21.3 percent over 90 days | no data | no data | 2 | |
| NVIDIA A100 80GB80 GiB | $2.70 | $2.09 to $3.19 | unchanged over 30 days | unchanged over 90 days | unchanged over 180 days | up 50.8 percent over 12 months | 3 | |
| NVIDIA L40S44.7 GiB | $1.55 | $1.09 to $2.25 | unchanged over 30 days | unchanged over 90 days | unchanged over 180 days | unchanged over 12 months | 3 | |
| NVIDIA RTX PRO 600096 GiB | $2.50 | $1.80 to $2.69 | up 19.6 percent over 30 days | up 25.6 percent over 90 days | up 18.2 percent over 180 days | no data | 3 |
- NVIDIA H10080 GiB, 4 listings$4.14per GPU-hour
- 30 days
- up 6.4 percent over 30 days
- 90 days
- up 16.1 percent over 90 days
- 180 days
- up 32.9 percent over 180 days
- 12 months
- up 52.4 percent over 12 months
$2.89 to $6.16 - NVIDIA H200141 GiB, 3 listings$5.40per GPU-hour
- 30 days
- unchanged over 30 days
- 90 days
- up 20.0 percent over 90 days
- 180 days
- up 35.3 percent over 180 days
- 12 months
- up 50.4 percent over 12 months
$5.29 to $6.30 - NVIDIA B200179 GiB, 4 listings$8.24per GPU-hour
- 30 days
- up 7.8 percent over 30 days
- 90 days
- up 19.1 percent over 90 days
- 180 days
- up 35.3 percent over 180 days
- 12 months
- up 47.4 percent over 12 months
$7.69 to $8.60 - NVIDIA B300268 GiB, 2 listings$8.99 to $9.50per GPU-hour
- 30 days
- up 6.3 percent over 30 days
- 90 days
- up 21.3 percent over 90 days
- 180 days
- no data
- 12 months
- no data
$8.99 to $9.50 - NVIDIA A100 80GB80 GiB, 3 listings$2.70per GPU-hour
- 30 days
- unchanged over 30 days
- 90 days
- unchanged over 90 days
- 180 days
- unchanged over 180 days
- 12 months
- up 50.8 percent over 12 months
$2.09 to $3.19 - NVIDIA L40S44.7 GiB, 3 listings$1.55per GPU-hour
- 30 days
- unchanged over 30 days
- 90 days
- unchanged over 90 days
- 180 days
- unchanged over 180 days
- 12 months
- unchanged over 12 months
$1.09 to $2.25 - NVIDIA RTX PRO 600096 GiB, 3 listings$2.50per GPU-hour
- 30 days
- up 19.6 percent over 30 days
- 90 days
- up 25.6 percent over 90 days
- 180 days
- up 18.2 percent over 180 days
- 12 months
- no data
$1.80 to $2.69
Listed by one GPU cloud only: NVIDIA A100 40GB $1.99 on Lambda, NVIDIA H100 NVL $3.19 on RunPod, NVIDIA L4 $0.69 on RunPod, NVIDIA A10 $1.29 on Lambda.
Check your GPU price
Enter what you pay for a GPU to see where it sits against this month's listed prices, and the difference over a month.
Worked out in your browser. Nothing you type is sent unless you ask for the result by email. For a machine with several GPUs, divide its hourly price by the number of GPUs. A full month is 730 hours.
Enter your price to see where it sits against the 4 listings recorded for the H100 on GPU clouds this month ($2.89 to $6.16).
Paying more than these prices for GPUs?
Cloud Cost Audit: We review your GPU machines, schedules and disks against what your models need. From $1,500, fixed. Nothing paid upfront.
AWS, Azure and Google Cloud
Each cloud's lowest on-demand price per GPU-hour, then the lowest Spot and commitment prices any of the three publishes. The GPU comes inside a whole machine, so the price also buys its processors, memory and local disk.
No on-demand price in this table changed over the 12 months to November 2026.
- NVIDIA H100unchanged over 12 months
- AWS, on demand
- $6.88
- Azure, on demand
- $12.29
- Google Cloud, on demand
- $11.06
- Lowest Spot
- $0.81 AWS
- Lowest 1-year
- $5.42 AWS
- Lowest 3-year
- $4.86 Google Cloud
- NVIDIA H200unchanged over 12 months
- AWS, on demand
- $7.91
- Azure, on demand
- $10.60
- Lowest Spot
- $0.85 AWS
- Lowest 1-year
- $5.82 Azure
- Lowest 3-year
- $5.28 Azure
- NVIDIA B200unchanged over 12 months
- AWS, on demand
- $14.24
- Lowest Spot
- $2.29 AWS
- Lowest 1-year
- $11.22 AWS
- Lowest 3-year
- $10.46 AWS
- NVIDIA B300no data
- AWS, on demand
- $17.80
- Lowest Spot
- $7.53 AWS
- Lowest 1-year
- $14.02 AWS
- Lowest 3-year
- $13.07 AWS
- NVIDIA GB200unchanged over 12 months
- Azure, on demand
- $27.04
- Lowest Spot
- $27.04 Azure
- Lowest 1-year
- $17.31 Azure
- Lowest 3-year
- $11.90 Azure
- NVIDIA A100 80GBunchanged over 12 months
- AWS, on demand
- $3.43
- Azure, on demand
- $3.67
- Google Cloud, on demand
- $5.07
- Lowest Spot
- $0.75 Azure
- Lowest 1-year
- $2.61 AWS
- Lowest 3-year
- $1.77 Azure
- NVIDIA A100 40GBunchanged over 12 months
- AWS, on demand
- $2.74
- Google Cloud, on demand
- $3.67
- Lowest Spot
- $0.60 AWS
- Lowest 1-year
- $2.09 AWS
- Lowest 3-year
- $1.29 Google Cloud
- NVIDIA L40Sunchanged over 12 months
- AWS, on demand
- $1.86
- Lowest Spot
- $0.57 AWS
- Lowest 1-year
- $1.41 AWS
- Lowest 3-year
- $0.96 AWS
- NVIDIA RTX PRO 6000no data
- AWS, on demand
- $3.36
- Azure, on demand
- $5.50
- Google Cloud, on demand
- $4.50
- Lowest Spot
- $1.02 Azure
- Lowest 1-year
- $2.65 AWS
- Lowest 3-year
- $1.96 AWS
- NVIDIA RTX PRO 4500no data
- AWS, on demand
- $2.52
- Lowest Spot
- $0.32 AWS
- Lowest 1-year
- $1.98 AWS
- Lowest 3-year
- $1.36 AWS
- NVIDIA H100 NVLunchanged over 12 months
- Azure, on demand
- $6.98
- Lowest Spot
- $1.68 Azure
- Lowest 1-year
- $5.81 Azure
- Lowest 3-year
- $4.31 Azure
- NVIDIA L4unchanged over 12 months
- AWS, on demand
- $0.80
- Google Cloud, on demand
- $0.71
- Lowest Spot
- $0.21 AWS
- Lowest 1-year
- $0.56 Google Cloud
- Lowest 3-year
- $0.40 Google Cloud
- NVIDIA A10Gunchanged over 12 months
- AWS, on demand
- $1.01
- Lowest Spot
- $0.24 AWS
- Lowest 1-year
- $0.79 AWS
- Lowest 3-year
- $0.54 AWS
- NVIDIA A10unchanged over 12 months
- Azure, on demand
- $3.20
- Lowest Spot
- $0.65 Azure
- Lowest 1-year
- $2.66 Azure
- Lowest 3-year
- $1.97 Azure
- NVIDIA T4unchanged over 12 months
- AWS, on demand
- $0.53
- Azure, on demand
- $0.53
- Lowest Spot
- $0.11 AWS
- Lowest 1-year
- $0.38 Azure
- Lowest 3-year
- $0.26 Azure
A commitment is billed for every hour of its term, used or not. Spot capacity can be taken back at short notice.
Lowest listed price for each GPU
The cheapest on-demand listing recorded this month across all 7 providers, with the page it was read from.
| GPU | Price per GPU-hour | Provider | Listing |
|---|---|---|---|
| NVIDIA H100 | $2.89 | RunPod | H100 PCIe |
| NVIDIA H200 | $5.29 | RunPod | H200 SXM |
| NVIDIA B200 | $7.69 | Lambda | B200 HGX, 8 GPUs in the machine |
| NVIDIA B300 | $8.99 | RunPod | B300 |
| NVIDIA GB200 | $27.04 | Azure | Standard_ND128isr_NDR_GB200_v6, 4 GPUs in the machine |
| NVIDIA A100 80GB | $2.09 | RunPod | A100 PCIe |
| NVIDIA A100 40GB | $1.99 | Lambda | A100 40GB |
| NVIDIA L40S | $1.09 | RunPod | L40S |
| NVIDIA RTX PRO 6000 | $1.80 | Nebius | RTX PRO 6000 |
| NVIDIA RTX PRO 4500 | $2.52 | AWS | g7.2xlarge |
| NVIDIA H100 NVL | $3.19 | RunPod | H100 NVL |
| NVIDIA L4 | $0.69 | RunPod | L4 |
| NVIDIA A10G | $1.01 | AWS | g5.xlarge |
| NVIDIA A10 | $1.29 | Lambda | A10 |
| NVIDIA T4 | $0.53 | AWS | g4dn.xlarge |
How these prices are measured
- Each price is what a provider lists on its own pricing page or price feed for one GPU for one hour, on demand, before tax.
- A machine with several GPUs is divided by its GPU count. Storage and data transfer are not included.
- A provider's price for a GPU is its lowest listing for that model, across variants and machine sizes.
- A change compares two months over the providers present in both, so a provider joining or leaving is not read as a price move.
- Where the month to compare with was not read, or lacks the reading of a provider that lists the GPU now, the cell says no data. Nothing is estimated or filled in.
- The months before November 2026 were read from dated sources after the fact: web archive captures of each pricing page, the published versions of the AWS price list, and the effective dates Azure gives its prices.
Sources, read 14 Nov 2026
Use the data
Every recorded row is free to download as CSV or JSON, under CC BY 4.0. Credit it as: GPU rental price data by Finitizer (tools.finitizer.com/gpu-rental-prices), CC BY 4.0.
Questions people ask
How much does an NVIDIA H100 cost per hour?
An NVIDIA H100 rents for a median $4.14 per GPU-hour on demand across 4 GPU clouds (November 2026). The lowest listed price is $2.89 on RunPod, and the highest among GPU clouds is $6.16. These are list prices for one GPU, before storage, data transfer and tax.
Are GPU rental prices going up?
On GPU clouds, yes for most models. Over 12 months the like-for-like price rose for H100 (+52.4%), H200 (+50.4%), B200 (+47.4%) and A100 80GB (+50.8%). On-demand list prices on AWS, Azure and Google Cloud did not change over the same 12 months.
Why do prices for the same GPU differ so much between providers?
A price covers different things at each provider: the machine around the GPU, the network, the support and how long you commit for. The big clouds sell a GPU inside a whole machine at a higher on-demand rate and discount it heavily for Spot or a 1-year or 3-year commitment. GPU clouds list a lower hourly price and sell less around it.
What is the difference between on-demand, Spot and committed prices?
On demand is the price with no commitment: you can stop at any time. Spot is spare capacity at a discount that the provider can take back at short notice. A commitment is a lower hourly rate in exchange for paying for one or three years. This page compares on-demand prices, and lists the Spot and commitment prices the three big clouds publish.
How often are these prices updated?
Once a month. The prices on this page were read on 14 Nov 2026, and the next reading is due on 1 Dec 2026. The history starts in November 2025. A month, once recorded, is never changed.
Why does this figure differ from a GPU price index?
Market indexes measure prices actually paid, often on spot marketplaces or multi-year contracts, and move daily. This page records the on-demand price each provider lists on its own pricing page, once a month. A list price is what you pay if you rent today with no contract, and it moves in steps.
The GPU price is one line of an AI bill.
Finitizer Tokenomics shows what your AI actually costs, read from your own bills and exports: by workload, team, model and the key carrying it.
See Finitizer Tokenomics