Practical guide
How Cloud Run concurrency changes cost
Compare Cloud Run concurrency and resource settings using billable instance time, latency, and safe throughput.
Updated · Sources checked
Concurrency changes instance time
Cloud Run cost is driven by billed CPU and memory time, not request count alone. Raising safe concurrency can allow more requests to share an instance; lowering it can create more concurrent instances. The result depends on application thread safety, CPU contention, memory, latency, and cold starts.
At 10 million requests of 0.4 seconds, total request work is 4 million request-seconds. At an ideal concurrency of 20, that suggests 200,000 instance-seconds before startup, overlap, and traffic shape. It is planning math, not a performance guarantee. Measure billable instance time after changing concurrency.
Use the Cloud Run Cost Calculator with observed CPU-, GiB-, and request quantities. Keep minimum-instance and networking costs separate.
FAQ
Is maximum concurrency always cheaper?
No. It can harm latency or exceed the application’s safe resource limits.