Amazon Personalize Pricing

With Amazon Personalize, you pay only for what you use, and there are no minimum fees and no upfront commitments. You are charged based on the amount of data processed and stored, the compute hours used to train your models, and for the throughput of recommendations.

Data ingestion
You are charged per GB of data uploaded to Amazon Personalize. This includes real-time data streamed to Amazon Personalize and batch data uploaded via Amazon S3.

Training
You are charged for the training hours consumed to train a custom model with your data. Note: A training hour represents 1 hour of compute capacity using 4v CPUs and 8 GiB memory. Amazon Personalize automatically chooses the most efficient instance types to train your data, which may be an instance that exceeds the baseline specifications in order to complete your job more quickly. Therefore, the number of training hours billed may be greater than the number of elapsed hours.

RECOMMENDATION (INFERENCE)

Real Time recommendations

You are charged for the personalization requests processed by Amazon Personalize. The service supports real-time recommendations, which is measured in transactions per second (TPS). Developers need to specify the minimum limits of the throughput, with Amazon Personalize guaranteeing low latency response for requests up to the provisioned throughput. If your requested throughput is more than the minimum provisioned TPS, Amazon Personalize will scale up to serve the additional requests and then scale down up to the minimum if the traffic reduces. The actual TPS used is calculated as the average requests/second within a 5-minute window. You pay for maximum of either the minimum provisioned TPS or the actual TPS.
When serving real-time recommendations, you are charged for throughput capacity per hour in units of TPS-hour (rounded up to the nearest hour). This is calculated as the maximum of either the minimum provisioned TPS or the actual TPS multiplied by the total time (in 5 minute increments within each hour) that requests are processed. These are then aggregated for the month’s usage and billed according to the pricing tiers.

TPS-hours = Maximum of (minimum provisioned TPS, actual TPS) x (5/60 minutes)

Batch recommendations

You are charged for number of users processed when using ‘USER_PERSONALIZATION’ and ‘PERSONALIED_RANKING’ recipes and for items processed when using ‘RELATED_ITEMS’ recipe for a batch inference job.