GPU rental by the hour,
compared for you.

172 machines from 10 providers over 32 GPU models, read every day from public price pages. Every rate is what one GPU costs for one hour, so a node of eight is compared at what one of its GPUs costs, and the spot price sits beside it wherever a provider publishes one.

Cheapest H100 PCIe today

RunPod
H100 PCIe · Community Cloud

1x H100 PCIe · 80 GB · 16 vCPU · 188 GB RAM

$1.99/GPU-hr

on demand · $1.99/hr for the whole machine

$2.89median of 3 rentals

2 providers rent the NVIDIA H100 PCIe, and this is the cheapest of them, 31% under the middle of that spread. It is a marketplace row, so the price is orderable and the machine belongs to a third party. Storage included: none, volumes are billed separately.

prices as of · re-ranked daily

Top 10 cheapest GPUs by the hour

browse every rental →

boards rented by 2 providers or more · prices as of

  1. 11V100.6V · 1x Tesla V100 16GBVerda (DataCrunch) · 1x V100 16GB · 16 GB · 6 vCPU · 23 GB RAM$0.173/GPU-hr
  2. 22V100.10V · 2x Tesla V100 16GBVerda (DataCrunch) · 2x V100 16GB · 16 GB each · 10 vCPU · 45 GB RAM$0.173/GPU-hr
  3. 34V100.20V · 4x Tesla V100 16GBVerda (DataCrunch) · 4x V100 16GB · 16 GB each · 20 vCPU · 90 GB RAM$0.173/GPU-hr
  4. 48V100.48V · 8x Tesla V100 16GBVerda (DataCrunch) · 8x V100 16GB · 16 GB each · 48 vCPU · 180 GB RAM$0.173/GPU-hr
  5. 5RTX A6000 · Community CloudRunPod · 1x RTX A6000 · 48 GB · 9 vCPU · 50 GB RAM$0.330/GPU-hr
  6. 6RTX A6000 · Secure CloudRunPod · 1x RTX A6000 · 48 GB · 9 vCPU · 50 GB RAM$0.530/GPU-hr
  7. 71A6000.10V · 1x RTX A6000 48GBVerda (DataCrunch) · 1x RTX A6000 · 48 GB · 10 vCPU · 60 GB RAM$0.592/GPU-hr
  8. 82A6000.20V · 2x RTX A6000 48GBVerda (DataCrunch) · 2x RTX A6000 · 48 GB each · 20 vCPU · 120 GB RAM$0.592/GPU-hr
  9. 94A6000.40V · 4x RTX A6000 48GBVerda (DataCrunch) · 4x RTX A6000 · 48 GB each · 40 vCPU · 240 GB RAM$0.592/GPU-hr
  10. 108A6000.80V · 8x RTX A6000 48GBVerda (DataCrunch) · 8x RTX A6000 · 48 GB each · 80 vCPU · 480 GB RAM$0.592/GPU-hr

Top 10 cheapest GPU memory

full ranking →

per GB of VRAM an hour, same floor · prices as of

  1. 1RTX A6000 · Community CloudRunPod · 48 GB per GPU · 1x RTX A6000 · 48 GB · 9 vCPU · 50 GB RAM$0.330/GPU-hr$0.0069/GB-hr
  2. 21V100.6V · 1x Tesla V100 16GBVerda (DataCrunch) · 16 GB per GPU · 1x V100 16GB · 16 GB · 6 vCPU · 23 GB RAM$0.173/GPU-hr$0.0108/GB-hr
  3. 32V100.10V · 2x Tesla V100 16GBVerda (DataCrunch) · 16 GB per GPU · 2x V100 16GB · 16 GB each · 10 vCPU · 45 GB RAM$0.173/GPU-hr$0.0108/GB-hr
  4. 44V100.20V · 4x Tesla V100 16GBVerda (DataCrunch) · 16 GB per GPU · 4x V100 16GB · 16 GB each · 20 vCPU · 90 GB RAM$0.173/GPU-hr$0.0108/GB-hr
  5. 58V100.48V · 8x Tesla V100 16GBVerda (DataCrunch) · 16 GB per GPU · 8x V100 16GB · 16 GB each · 48 vCPU · 180 GB RAM$0.173/GPU-hr$0.0108/GB-hr
  6. 6RTX A6000 · Secure CloudRunPod · 48 GB per GPU · 1x RTX A6000 · 48 GB · 9 vCPU · 50 GB RAM$0.530/GPU-hr$0.0110/GB-hr
  7. 71A6000.10V · 1x RTX A6000 48GBVerda (DataCrunch) · 48 GB per GPU · 1x RTX A6000 · 48 GB · 10 vCPU · 60 GB RAM$0.592/GPU-hr$0.0123/GB-hr
  8. 82A6000.20V · 2x RTX A6000 48GBVerda (DataCrunch) · 48 GB per GPU · 2x RTX A6000 · 48 GB each · 20 vCPU · 120 GB RAM$0.592/GPU-hr$0.0123/GB-hr
  9. 94A6000.40V · 4x RTX A6000 48GBVerda (DataCrunch) · 48 GB per GPU · 4x RTX A6000 · 48 GB each · 40 vCPU · 240 GB RAM$0.592/GPU-hr$0.0123/GB-hr
  10. 108A6000.80V · 8x RTX A6000 48GBVerda (DataCrunch) · 48 GB per GPU · 8x RTX A6000 · 48 GB each · 80 vCPU · 480 GB RAM$0.592/GPU-hr$0.0123/GB-hr

Cheapest eight-GPU node

8V100.48V · 8x Tesla V100 16GB

Verda (DataCrunch)

$0.173/GPU-hr

8x V100 16GB · 16 GB each · 48 vCPU · 180 GB RAM

one machine, GPUs on the same board

Cheapest RTX board

RTX A5000 · Community Cloud

RunPod

$0.160/GPU-hr

1x RTX A5000 · 24 GB · 9 vCPU · 25 GB RAM

the cheap end of this market

Index

172

machines · median $1.84/GPU-hr

10

providers

32

GPU models

3

publish spot

What a GPU costs for an hour, on demand (USD per GPU-hour)

16
<$0.50
24
0.50–1
47
1–2
28
2–4
34
4–8
7
$8+

This week

Every rate that
moved, dated.

what changed →

Today's rankings, recomputed daily

Every ranking uses the on-demand rate for one GPU for one hour. Reserved and committed rates are not indexed, because a price that needs a month's commitment is not the standard recurring rate this site compares. Serverless rates are listed and browsable and ranked nowhere: they buy a container that runs when it is called rather than a machine you hold.