Redshift concurrency scaling cost: the free credits and what runs over

Quick answer

Redshift concurrency scaling adds transient clusters automatically when query concurrency spikes, and you earn one hour of free concurrency scaling credit per day for each active main cluster, which accrues up to 30 hours. Usage within the free credit costs nothing, and many workloads never exceed it. Beyond the credit, concurrency scaling is billed per second at your cluster's on-demand rate. The cost risk is a bursty or poorly-tuned workload that constantly overflows, so the controls are capping the number of concurrency scaling clusters via WLM and tuning queries so spikes are rarer.

Redshift concurrency scaling solves a real problem: when many queries arrive at once, they queue behind each other on a fixed-size cluster. Concurrency scaling spins up transient clusters to run the queued queries in parallel, then shuts them down when the spike passes. The clever part of its pricing is a daily free credit that covers a surprising amount of real-world bursting for nothing, so the cost only appears when usage runs well past it.

How the free credit works

For every 24 hours that your main cluster is running, you accrue one hour of free concurrency scaling credit. That credit banks up to a maximum of 30 hours per cluster. Because concurrency scaling is only active during actual spikes, and a single free hour per day covers a lot of intermittent bursting, many clusters run concurrency scaling entirely within the free allotment and pay nothing for it. The feature is effectively free for workloads whose concurrency spikes are occasional rather than constant.

Usage	Cost
Within daily free credit	$0
Beyond free credit	Per-second at cluster on-demand rate

What overflow costs

Once you exhaust the accrued free credit, concurrency scaling is billed per second at the same on-demand rate as your main cluster's nodes. So if your main cluster costs $10 per hour, an hour of concurrency scaling beyond the credit adds about $10, prorated per second. This is usually modest for a workload that occasionally overshoots its free hour, but it can add up for a warehouse where heavy concurrency is the norm all day, because then concurrency scaling is running for many hours beyond the single daily free hour.

When it becomes a cost problem

The trouble arises when concurrency scaling masks an undersized or untuned cluster. If your main cluster is too small for your steady concurrency, concurrency scaling fires constantly, blows through the free credit early each day, and bills per-second for hours. In that case you are renting burst capacity continuously, which is more expensive than simply sizing the main cluster correctly or moving to a model that fits variable load. Persistent heavy overflow is a signal to revisit cluster sizing, covered in RA3 versus DC2, or to consider Redshift Serverless.

Keeping it in check

Two controls matter. First, the workload management (WLM) setting max_concurrency_scaling_clusters caps how many transient clusters can spin up, bounding the maximum spend from concurrency scaling per unit time. Set it to a deliberate number rather than leaving it wide open. Second, tune the queries and WLM queues so spikes are rarer: better queue configuration, query optimization, and result caching reduce how often concurrency scaling is needed at all. Reducing the frequency of spikes is cheaper than paying to absorb them.

Let the credit do the work

For most warehouses, concurrency scaling is a free or near-free feature that smooths out occasional query bursts, and it should stay enabled. The cost only matters when a workload overflows constantly, which is really a sizing or tuning signal in disguise. Cap the cluster count via WLM, watch the concurrency scaling usage metrics, and price your Redshift configuration against the resource catalog so persistent overflow prompts a sizing decision rather than an uncapped monthly charge.

FAQ

How much does Redshift concurrency scaling cost?

You earn one hour of free concurrency scaling credit per day for each active main cluster, accruing up to 30 hours, and usage within that credit is free. Beyond the credit, concurrency scaling is billed per second at your cluster's on-demand node rate, so an hour of overflow on a $10-per-hour cluster adds about $10, prorated per second.

Is Redshift concurrency scaling free?

Often, yes. The one-hour-per-day free credit covers a lot of intermittent bursting because concurrency scaling only runs during actual query spikes. Many clusters stay entirely within the free allotment and pay nothing. Cost appears only when a workload's concurrency spikes are constant enough to exhaust the daily credit and run into per-second billing.

Why is my concurrency scaling cost high?

Usually because the main cluster is undersized or untuned for its steady concurrency, so concurrency scaling fires constantly, exhausts the free credit early each day, and bills per-second for hours. That is effectively renting burst capacity all day, which is more expensive than sizing the main cluster correctly or moving to a model like Redshift Serverless that fits variable load.

How do I limit Redshift concurrency scaling spend?

Set the WLM parameter max_concurrency_scaling_clusters to cap how many transient clusters can spin up, bounding maximum spend per unit time. Then reduce how often spikes happen through query optimization, better WLM queue configuration, and result caching. Cutting the frequency of concurrency spikes is cheaper than paying per-second to absorb them.

How does C3X help with Redshift concurrency scaling cost?

C3X prices Redshift clusters from Terraform including node type, count, and configuration, so the baseline cluster cost is clear before deploy. Combined with monitoring concurrency scaling usage, that helps teams tell whether persistent overflow is a sizing problem worth solving with a bigger cluster or serverless, rather than an uncapped recurring charge they simply absorb.

What to do next

Tell burst smoothing from an undersized cluster. C3X prices Redshift from Terraform against a live resource catalog. Start with the quickstart.