AWS SQS charges per million requests and caps messages at 1 MiB

The workload

Designing a queue on Amazon SQS means checking two AWS-published numbers first: the per-request price and the maximum message size. AWS's own SQS pricing page, retrieved 16 September 2026, and its message quotas documentation state both. Neither requires provisioning a server; the workload is reading the documentation and, if payloads exceed the size limit, adding a second AWS service to the design.

What the documents show

AWS's pricing page states, verified, that 'all customers can make 1 million Amazon SQS requests for free each month.' Beyond that, AWS's own published on-demand rate data — the data behind the pricing page — prices standard-queue requests at $0.40 per million in the first usage tier (up to 100 billion requests monthly), $0.30 per million in the second tier, and $0.24 per million in the third, each verified and effective from 1 August 2025. FIFO requests cost more: $0.50 per million in the first tier, down to $0.35 per million in the third, verified. AWS's message quotas page states, verified, that the maximum message size is 1,048,576 bytes (1 MiB), not the 256 KB figure that appears elsewhere in AWS's documentation — the Extended Client Library documentation for Java still describes offloading 'particularly for payloads ranging from 256 KB to 2 GB' and offers storing a message in S3 'only when the size of a message exceeds 256 KB.' Both are verified quotes from AWS's current documentation; 256 KB is a configurable default inside the older extended-client tooling, not a queue-wide ceiling, and the two documents describe different layers of one system rather than contradicting each other.

The operating cost

A workload sending 10 million standard-queue requests monthly, an assumed volume for this estimate, would pay for 9 million billable requests after the free allowance, at $0.40 per million: 9 × $0.40 = $3.60 monthly, an editorial estimate using AWS's Tier 1 rate. AWS also charges by payload size: 'each 64 KB chunk of a payload is billed as 1 request,' verified, so a message near the 1 MiB ceiling bills as roughly 16 requests, not one.

The stop condition

The size ceiling is the stop condition AWS itself states: at 1 MiB, a message cannot be sent as-is, and the documented alternative is the Extended Client Library storing the payload in S3, with a stated maximum of 2 GB. The cost-side stop condition is editorial: because payloads bill in 64 KB chunks, small-message designs accumulate per-chunk charges faster than the raw message count suggests, worth checking before assuming the free million requests will cover a given month.

• How many 64 KB chunks does a typical message in this workload actually consume, once measured rather than assumed?
• Would any payload in this system approach the 1 MiB limit, and is the S3 extended-client pattern already wired in before that happens?
• Does the workload's queue type, standard or FIFO, match its actual ordering needs, given FIFO's higher per-million rate?

AWS's own pricing and quotas documentation are both living documents that can change; the rates and limits above reflect what AWS published as retrieved on 16 September 2026, not a permanent commitment.