CloudConsole

DynamoDB on-demand vs provisioned: which is cheaper?

Updated 17 November 2026 · How we write and check guides

Short answer

On-demand bills per read and write request unit you actually use, so idle time costs nothing; provisioned bills per read and write capacity unit-hour you configure, whether or not you use it, at a lower price per unit. Provisioned is cheaper for steady, predictable traffic that keeps capacity well utilised; on-demand is cheaper for spiky, low or unknown traffic. You can switch from provisioned to on-demand up to four times in a rolling 24 hours, and back at any time.

Diagram: What each capacity mode bills. On-demand: spiky, low or unknown traffic: Requests you make → Per request unit…
Where this charge comes from, using the dimensions and usage types named on this page.

How each mode appears on the bill

ModeUsage typesYou pay for
On-demandReadRequestUnits, WriteRequestUnitsRequests made
ProvisionedReadCapacityUnit-Hrs, WriteCapacityUnit-HrsCapacity configured, every hour
BothTimedStorage-ByteHrsTable and index storage

Usage types have an optional Region prefix (for example USW2-ReadRequestUnits); no prefix means us-east-1. Global secondary indexes consume and bill their own capacity in both modes.

Reserved capacity, bought for one or three years, applies only to provisioned capacity. A table you might move to on-demand later is a poor candidate for a reservation, and on-demand tables cannot use it.

Compare on your own numbers

  1. Pull ConsumedReadCapacityUnits and ConsumedWriteCapacityUnits for the table (and each GSI) over two to four weeks.
  2. On-demand estimate: total consumed units x on-demand request price.
  3. Provisioned estimate: the capacity you would need to provision (peak plus headroom, or what auto scaling would track) x hours x capacity-hour price.
  4. Compare with your Region's prices from the DynamoDB pricing page.
# Consumed write capacity, hourly sums for the last 14 days
aws cloudwatch get-metric-statistics --namespace AWS/DynamoDB \
  --metric-name ConsumedWriteCapacityUnits \
  --dimensions Name=TableName,Value=my-table \
  --start-time 2026-11-03T00:00:00Z --end-time 2026-11-17T00:00:00Z \
  --period 3600 --statistics Sum --output text

# Current mode and provisioned throughput
aws dynamodb describe-table --table-name my-table \
  --query 'Table.[BillingModeSummary.BillingMode,ProvisionedThroughput]'

A flat consumption line with a modest peak-to-average ratio favours provisioned; long idle periods with sharp bursts favour on-demand.

Switch and tune

# To on-demand (allowed up to four times in a rolling 24 hours)
aws dynamodb update-table --table-name my-table --billing-mode PAY_PER_REQUEST

# To provisioned, with explicit capacity (then attach auto scaling)
aws dynamodb update-table --table-name my-table --billing-mode PROVISIONED \
  --provisioned-throughput ReadCapacityUnits=200,WriteCapacityUnits=50
  • With provisioned, use auto scaling with a target utilisation, and buy reserved capacity only for the steady floor.
  • With on-demand, set a maximum throughput per table to bound cost from runaway clients.
  • The Standard-Infrequent Access table class lowers storage cost for tables where storage dominates throughput.

What not to do

  • Do not switch to provisioned with capacity set to today's average; bursts above it are throttled.
  • Do not forget GSIs when estimating; a busy index can cost as much as the table.
  • Do not use scans for routine reads; they consume capacity for every item read, in either mode.
  • Do not delete expired items with your own batch job when Time to Live can do it: TTL deletions do not consume write capacity, while a cleanup job pays for every delete.

Questions

Is DynamoDB on-demand more expensive than provisioned?

Per unit, yes. In total it is often cheaper for spiky or low traffic because you pay nothing for idle capacity.

How often can I switch DynamoDB capacity modes?

From provisioned to on-demand up to four times in a rolling 24-hour window; from on-demand to provisioned at any time.

What is ReadCapacityUnit-Hrs on my bill?

Provisioned read capacity, billed per unit per hour whether or not it was used.

When this happens away from your desk

Capacity mode decisions age as traffic changes. CloudConsole, a phone app for AWS, shows DynamoDB spend by usage type, with request-unit and capacity-hour lines explained, so you can tell which mode a table is paying for from your phone.

About the app → · Download on the App Store

Sources