When you use the AWS DeepRacer service console you will be charged based on your usage. Your monthly AWS billing statement will show a charge from each of the AWS services used by AWS DeepRacer to train models, evaluate models, or store models and metadata, such as log files. The details of all services used by AWS DeepRacer can be seen in the related product pricing section and in pricing example #1 below.
To get you started, AWS DeepRacer provides a Free Tier to first time AWS DeepRacer users, that should cover the first 7 hours of training. This is enough time to train and tune your first model and enter the AWS DeepRacer League. There is no cost for submitting a model to take part in any AWS DeepRacer League virtual event. Learn more in the Free Tier section below.
You can pre-order your AWS DeepRacer car on Amazon.com for $399 (US only). Availability for selected other countries will be announced later. There is no charge for driving your AWS DeepRacer car.
You can access the simulator from the US East (N. Virginia) region in the AWS console.
Free Tier
To get you started with the AWS DeepRacer service console, we provide credits to cover the first seven hours of usage. This is enough to train your first model, evaluate it, tune it, and then enter it into the AWS DeepRacer League. At the end of the first month that you use AWS DeepRacer, your account will receive the following service credits: 65 Simulation Unit hours ($26) for AWS RoboMaker, 10 hours of training ($5.57) on an ML.C4.2XL Amazon SageMaker instance, and $1.50 to cover Amazon S3, Amazon Kinesis Video Streams, and Amazon CloudWatch costs. The service credits will be loaded at the end of the month, will be valid for a month, and will be visible on your billing statement.
You will also benefit from the Free Tiers for other AWS services. If you have never used AWS RoboMaker, you will receive AWS RoboMaker Free Tier for the first month. Furthermore, on a monthly basis, you will also benefit from the Amazon S3, and Amazon CloudWatch Free Tier.
Please see pricing example #1 below for a breakdown of costs you can expect for a typical AWS DeepRacer simulation, after you have used the Free Tier.
Related product pricing
AWS DeepRacer integrates with Amazon SageMaker for reinforcement learning model training, AWS RoboMaker to provide the racing simulator, Amazon Kinesis Video Streams for video streaming of virtual simulation footage, Amazon S3 for model storage, and Amazon CloudWatch for log capture. The first time that you use the AWS DeepRacer service, the required IAM roles needed to call the integrated services on your behalf, and the required Virtual Private Cloud (VPC) stack to run Amazon SageMaker and AWS RoboMaker, will be created in your account. There is no cost associated with creating and using a VPC, but each service will carry its own charge. For more details on related product pricing please click here.
Pricing example #1 (assuming you have used the Free Tier)
The examples illustrate costs assuming you have used all of the Free Tier, to provide a clear view of the costs of the service. When getting started with AWS DeepRacer, developers need to train a reinforcement learning model. The AWS DeepRacer service console will guide you through creating your first model with a series of sensible defaults, followed by training in the virtual simulator. We recommend that you train your model for at least two hours using the default settings, but training time may change depending on your parameters. During training, AWS DeepRacer will use AWS services on your behalf, including Amazon SageMaker, AWS RoboMaker, Amazon S3, Amazon Kinesis Video Streams, and Amazon CloudWatch.
After training completes, you will use the AWS DeepRacer service console to evaluate the performance of your trained model, by allowing it to drive autonomously around the virtual track. This allows you to benchmark the performance of your model. For evaluations AWS DeepRacer will still use the same AWS services as during training, except Amazon SageMaker, as no machine learning training is needed. Evaluation typically takes 5 minutes. To simplify the example, we aggregate the costs per service across training and evaluation. The cost of these activities is listed below:
Amazon SageMaker
| SageMaker Instance | Hours used | Cost per hour ($) | Total ($) |
ML.C4.2XL |
2.00 |
$0.557 |
$1.11 |
AWS RoboMaker
Estimated Simulation Units per hour are based on the track footprint and compute requirement. They range from 6 RoboMaker Simulation Units for the re:Invent track to 9 Simulation Units for the July 2019 Virtual Circuit track The example below shows pricing based on the July 2019 track.
Estimated Simulation Units per Hour |
Hours used |
Cost per simulation unit per hour ($) |
Total ($) |
|---|---|---|---|
9.00 |
2.08 |
$0.40 |
$7.50 |
Amazon S3
Activity (in unit) |
Estimated usage (in unit) |
Cost per unit in $ |
Total ($) |
|---|---|---|---|
Data Storage (GB) |
1.13 |
$0.023 |
$0.03 |
PUT, COPY, POST LIST (requests) |
2,928.00 |
$0.005 per 1000 |
$0.01 |
GET (requests) |
1,665.53 |
$0.0004 per 1000 |
$0.00 |
Total |
$0.04 |
Amazon Kinesis Video Streams
Activity |
Estimated usage (GB) |
Cost per GB ($) |
Total ($) |
|---|---|---|---|
Data Ingest |
1.18 |
$0.00850 |
$0.01 |
HLS Stream |
0.07 |
$0.01190 |
$0.00 |
Total |
|
|
$0.01 |
Amazon CloudWatch Metrics
Activity (in units) |
Estimated usage (in unit) |
Cost per unit in $ |
Total ($) |
|---|---|---|---|
Metrics (count) |
0.01 |
$0.30 |
$0.00 |
Metric Requests (count) |
644.63 |
$0.01 per 1000 |
$0.01 |
Data Ingest (GB) |
0.01 |
$0.50 |
$0.00 |
Data Stored (GB) |
0.01 |
$0.03 |
$0.00 |
Total |
|
|
$0.01 |
In this example, the total cost is $8.98.
Pricing Example #2 (assuming you have used all of the Free Tier)
While the model you created in Example #1 is a good start, you think you can improve its driving behavior and get a better lap time as a result. You decide to iterate on your model parameters and reward function, and train for 2 hours. The evaluation looks promising but you suspect your model may need more training as it is fast, but can’t yet complete a full lap. You clone this new model and train it for another 2 hours. Your cloned model is driving much better and you decide to test your optimizations in the real-world by deploying the final model to your AWS DeepRacer car. The example below shows pricing based on the July 2019 Virtual Circuit track using 9 Simulation Units. Estimated Simulation Units per hour are based on the track footprint and compute requirement.
Training and Evaluation
Service |
Estimated Cost ($) |
|---|---|
Amazon SageMaker |
$2.23 |
AWS RoboMaker |
$15.60 |
Amazon S3 |
$0.08 |
Amazon Kinesis Video Streams |
$0.02 |
Amazon CloudWatch |
$0.03 |
Total |
$17.96 |
Model deployment
Activity |
Estimated data (GB) |
Cost per GB ($) |
Total ($) |
|---|---|---|---|
Amazon S3 - Retrieve |
0.025 |
$0.09 |
$0.002 |
In this example, the total cost is $17.96.
Pricing Example #3 (assuming you have used all of the Free Tier)
You decide to leave your models from examples #1 and #2 on Amazon S3 and logs in Amazon CloudWatch. The following table shows the cost for a month of storage.
Service |
Estimated Data Stored (GB) |
Cost per GB per month |
Total |
|---|---|---|---|
Amazon S3 |
3.40 |
$0.023 |
$0.08 |
Amazon CloudWatch |
0.02 |
$0.030 |
$0.00 |
Total |
|
|
$0.08 |
The total storage cost for a month is $0.08.
Additional pricing resources
Calculate your total cost of ownership (TCO)
Easily calculator your monthly costs with AWS
Additional resources for switching to AWS
AWS Information on AWS DeepRacer League rules and eligibility requirements.
Get hands-on with RL, experiment, and learn through autonomous driving.
