Kinesis Data Streams can now scale ingest capacity down, not just up
Amazon extends warm throughput in On-demand Advantage mode to support write capacity reduction, giving operators bidirectional control over stream sizing.
Can you scale down write throughput on Amazon Kinesis Data Streams?
Yes, as of July 24, 2026. Amazon has extended warm throughput in On-demand Advantage mode to support scaling ingest capacity downward, not just upward. Operators can now set a lower warm throughput value on any qualifying on-demand stream to shed excess write capacity they no longer need.
How the scale-down works
When you reduce the warm throughput value on an on-demand stream, the service targets that requested capacity with one important floor: it will not drop below whatever peak ingest usage the stream recorded in the previous hour. This means the stream will not be undersized for traffic that was active moments ago, but it will release capacity that has genuinely gone idle.
The behavior is intentionally conservative. If your traffic dropped sharply an hour ago and has stayed low, the stream will shed the surplus. If there was a burst within that window, the stream retains enough headroom to cover it.
What changes for operators
Before this update, warm throughput was a one-directional tool. On-demand streams would expand automatically in response to rising ingest load, and warm throughput let you proactively pre-provision capacity ahead of anticipated spikes. Scaling back required waiting for the platform to reclaim unused capacity on its own schedule.
Now both directions are explicit. Setting a lower warm throughput value is the trigger for a downward adjustment, giving teams a straightforward way to right-size streams after a campaign, batch job, or traffic event ends.
Pricing and availability
Amazon is not charging extra for the scale-down capability. It is included at no additional cost for all on-demand streams that have On-demand Advantage mode enabled. The feature is available in every AWS Region where On-demand Advantage is currently supported.
| Attribute | Detail |
|---|---|
| Mode required | On-demand with On-demand Advantage enabled |
| Scale-down trigger | Set a lower warm throughput value on the stream |
| Capacity floor | Higher of: requested value or peak ingest in last hour |
| Additional cost | None |
| Regional availability | All regions supporting On-demand Advantage |
Getting started
To adjust warm throughput on an existing stream, the relevant entry point in the AWS documentation is the "Update a stream" guide in the Kinesis Data Streams Developer Guide. Pricing details are on the Kinesis Data Streams pricing page.
For teams running event-driven architectures where ingest volume is uneven, the practical upside is tighter cost control without manual capacity math. You define the floor, the service enforces safety against recent traffic, and excess capacity is released.