Explanation
Why the waste happens and who it affects.
In such cases, provisioned concurrency remains allocated and billed during idle periods, creating ongoing cost without proportional performance or business benefit. This is distinct from standard Lambda execution charges, which are purely usage-based.
Billing model
The pricing dimensions that drive this cost.
Lambda Provisioned Concurrency is billed per GB-second, based on configured memory multiplied by the amount of concurrency provisioned, from when it is enabled until it is disabled, rounded up to the nearest 5 minutes. Charges accrue regardless of invocation volume, even when functions are idle.
How to detect
3 checks to find it in your estate.
- Review whether functions with Provisioned Concurrency have sustained or predictable traffic patterns
- Assess whether cold-start latency materially impacts user experience or system behavior
- Identify functions where provisioned concurrency appears idle for extended periods (for example, using the ProvisionedConcurrencyUtilization metric); reserved concurrency is a separate, free setting
How to fix
2 ways to remove the waste.
- Remove Provisioned Concurrency from functions without strict latency or availability requirements
- Reserve Provisioned Concurrency only for workloads with consistent traffic or critical cold-start sensitivity
Documentation
Vendor references for pricing and configuration.