Requests, not messages β and why batching cuts the bill 10Γ.
Queues bill requests, not messages, so batching up to 10 per call and long polling slash the count. For streaming-scale throughput instead, see the managed Kafka cost calculator.
The Message Queue (SQS) Cost Calculator estimates your monthly bill for a message-queue service that charges per API request. It multiplies your messages per month by the API requests per message to get total requests, then applies your price per million requests to produce a monthly cost. The main drivers are volume and the request multiplier: each message rarely costs a single request, because sending, receiving, and deleting are typically separate calls. A queue processing tens of millions of messages, each involving several requests, adds up faster than raw message counts suggest.
The key trade-off to watch is the requests-per-message figure, since it can quietly double or triple your bill. Batching multiple messages into a single send or receive call is the most direct lever: fewer API requests for the same throughput means lower cost. Before committing to an architecture, test realistic multipliers here rather than assuming one request per message, and remember that aggressive polling or short visibility timeouts can inflate receive calls well beyond what your message volume alone would imply.
Per API request, not per message. Each message typically involves multiple requests (send, receive, delete), so the request count is a few times the message count. The first million requests each month are free; after that it is about $0.40 per million.
Batch. SQS lets you send, receive and delete up to 10 messages per API call, which cuts your billable request count by up to 10Γ. Also use long polling to avoid empty-receive requests that you still pay for.
Cost at different volumes:
| Volume | Monthly | Yearly |
|---|