The Ultimate Guide to Cloud GPU Providers
Cloud GPUs available at each GPU cloud provider
Cloud GPUs available at each GPU cloud provider
Select by Cloud Provider
Showing 55 results of 55 items.
GPU Cloud Comparison
| GPU Cloud | Marketing Name | Multi-GPU | GPU Memory (GiB) | TFLOPS (SP) | Tensor Cores | CPUs | RAM (GiB) | $ per hour | $ per month |
|---|---|---|---|---|---|---|---|---|---|
| AWS EC2 | M60 | 1x, 2x, 4x | 8 | 9.6 | 0 | 16 | 122 | $1.14 | -- |
| AWS EC2 | T4 | 1x, 2x, 4x, 8x | 16 | 8.1 | 320 | 16 | 64 | $1.20 | -- |
| AWS EC2 | A10G | 1x, 4x, 8x | 24 | 31.0 | 320 | 16 | 64 | $1.62 | -- |
| AWS EC2 | V100 16 GB | 1x, 4x, 8x | 16 | 14.0 | 640 | 8 | 61 | $3.06 | -- |
| AWS EC2 | V100 32 GB | 8x | 32 | 14.0 | 640 | 12* | 96* | $3.90* | -- |
| AWS EC2 | A100 40 GB | 8x | 40 | 19.5 | 432 | 12* | 144* | $4.10* | -- |
| CoreWeave | Quadro RTX 4000 | 1x, 2x, 3x, 4x, 5x, 6x, 7x | 8 | 7.1 | 288 | 3 | 16 | $0.35 | -- |
| CoreWeave | RTX A4000 | 1x, 2x, 3x, 4x, 5x, 6x, 7x | 16 | 19.2 | 192 | 4 | 20 | $0.75 | -- |
| Google Cloud | T4 | 1x, 2x, 4x | 16 | 8.1 | 320 | 8 | 30 | $0.75 | $383 |
| Google Cloud | K80 | 1x, 2x, 4x, 8x | 12 | 8.7 | 0 | 8 | 30 | $0.85 | $434 |
| Google Cloud | P4 | 1x, 2x, 4x | 8 | 5.5 | 0 | 8 | 30 | $1.00 | $510 |
| Google Cloud | P100 | 1x, 2x, 4x | 16 | 9.3 | 0 | 8 | 30 | $1.86 | $950 |
| Jarvis Labs | Quadro RTX 5000 | 1x | 16 | 11.2 | 384 | 7 | 32 | $0.49 | $245 |
| Microsoft Azure | K80 | 1x, 2x, 4x | 12 | 8.7 | 0 | 6 | 56 | $0.90 | $657 |
| Microsoft Azure | T4 | 1x, 4x | 16 | 8.1 | 320 | 16 | 110 | $1.20 | $879 |
Introduction
In this guide we’ll take a look at various GPU cloud services offerings and we’ll compare the GPU cloud offerings of the large public clouds to independent alternatives, startups, and everything in between. By the end of this guide you’ll have a good understanding of GPU cloud options available today – especially for machine learning and deep learning applications.
Methodology
GPU cloud providers often use different units of measurement with different sensible defaults. GPU machine specs can vary wildly from cloud to cloud – with different instance or machine sub-groupings and different pricing conventions.
Our mission in this guide is to provide a simplified overview of the GPU cloud providers to make it easier to understand what GPU cloud providers are selling in comparative terms and what various services offer as part of their GPU cloud offerings.
Let’s jump into the analysis!
Paperspace
Best GPU variety
Best GPU cloud providers to get up and running
About the product
Paperspace is a Series B cloud infrastructure company based in New York City focused on accelerated computing applications. With more than 500,000 users, Paperspace operates data centers across the US and Europe.
GPUs available
- A100 40 GB
- A100 80 GB
- Quadro M4000
Pros
- Widest selection of GPUs
- Competitive price per hour
Cons
- Limited free availability
AWS EC2
Best GPU cloud providers for infinite scale
About the product
Amazon Elastic Compute Cloud or EC2 is one of the oldest products in the AWS portfolio.
GPUs available
- A100 40 GB
- A10G
Pros
- GPU machines are part of highly popular EC2 product
Cons
- Very long time to get setup
Google Cloud
Best GPU cloud providers for infinite scale
About the product
Google’s GCP offers six different GPU types which are available to add on to new or existing VMs.
GPUs available
- A100 40 GB
- K80
Pros
- Google products Kaggle and Colab feature similar GPUs
Cons
- Few GPU options
Microsoft Azure
Best GPU cloud providers for infinite scale
About the product
Microsoft Azure has the best selection of GPU instances among the big public cloud providers.
GPUs available
- A100 40 GB
- A100 80 GB
Pros
- Best variety of GPUs among big public cloud providers
Cons
- Difficult to get help when needed
OVH Cloud
Best GPU cloud providers for long-term reserved pricing
About the product
OVH is a French cloud computing company founded in 1999 and is Europe’s largest hosting provider.
GPUs available
- V100 16 GB
- V100 32 GB
Pros
- Extremely good pricing
Cons
- Only offering V100 which lags behind state of the art A100