DynamoDB cost calculator: on-demand vs provisioned

This DynamoDB cost calculator estimates your monthly bill for the Standard table class in both capacity modes, on-demand and provisioned, from reads, writes, item size and stored data. Prices are the current US East (N. Virginia) rates from the AWS price list published on 2026-09-11, and the on-demand maths reproduces AWS's own pricing example to the cent.

Running it behind Lambda? AWS Lambda cost calculator →

Calculator

The defaults are AWS's own example (3.55 million 1 KB writes and strongly consistent reads, 27 GB stored), which costs $3.16 a month on-demand. For provisioned mode, set how much higher your peak traffic is than your average, because capacity has to cover the peak. The default of 21 matches the example's busiest day.

How DynamoDB counts reads and writes

Every read or write is converted into request units before it is priced. Two things change the count: the size of the item, and the kind of read or write. A 3 KB item and a 4 KB item cost the same to read; a 4 KB item and a 5 KB item do not. Keeping items small, or reading only the attributes you need through a projection, is the cheapest optimisation there is.

Current prices (US East)

Source: AWS Price List for Amazon DynamoDB, US East (N. Virginia), published 2026-09-11. The provisioned free tier covers 25 WCU and 25 RCU for a whole month; on-demand request units are not part of it.

On-demand vs provisioned

A useful way to think about it: provisioned capacity bills every hour of the month, so it only pays off when you actually use most of what you provision. Traffic that is high during working hours and near zero at night usually favours on-demand, even at large volumes. Try both in the calculator with your real peak-to-average ratio.

Worked example

AWS's example application writes and reads 1 KB items with strong consistency: 10,000 of each per day for ten days, 2.5 million on one busy day, then 50,000 a day for the rest of the month. That totals 3.55 million writes and 3.55 million reads, with 27 GB stored.

Now compare provisioned capacity sized for the busy day (about 21 times the average rate, so 29 write and 29 read units):

The free capacity is shared across the account, so the second row is the realistic one once you have a few tables. Paying all month for capacity that one day needs is what makes spiky traffic expensive in provisioned mode.

Costs the calculator leaves out

• Global secondary indexes. Every write to a table also writes to each index that includes the item, and those writes are billed too.
• Backups and restores. On-demand backups and point-in-time recovery are billed per GB.
• Streams and global tables. Reading change streams and replicating to other regions add their own charges.
• Data transfer. Data leaving the region is billed per GB.

How to reduce DynamoDB costs

1. 1
   Use eventually consistent reads wherever a few hundred milliseconds of staleness is fine. They cost half.
2. 2
   Keep items small. Move large blobs to S3 and store the key. Writes are billed per KB, so a 6 KB item costs six times a 1 KB one.
3. 3
   Project only what you need into secondary indexes, and drop indexes nobody queries.
4. 4
   Switch steady tables to provisioned with auto scaling once traffic is predictable, and keep spiky ones on-demand.
5. 5
   Expire old data with TTL. Deletes by time-to-live are free and shrink storage.
6. 6
   Cache hot reads in the application or a cache layer when the same items are read thousands of times.

Common questions

Need help with DynamoDB or the code around it?

Table design, access patterns and the serverless functions that use them are a big part of AWS serverless development. To estimate the compute side, use the AWS Lambda cost calculator.