Model Cloud Run vCPU-second and GiB-second billing with concurrency, minimum instances and request-based pricing.
Cloud Run bills the lifetime of an instance, not the sum of request durations, so raising concurrency divides the same request-seconds across fewer instance-seconds and cuts cost almost linearly. Instance-based billing removes the per-request charge but bills the full instance lifetime including idle time, which is why the two models are compared side by side here. Concurrency is the single most powerful cost lever on Cloud Run and it is free to change, while minimum instances buy away cold starts with a fixed monthly floor you should size deliberately. Confirm the per-vCPU-second and per-GiB-second prices for your region in the Google Cloud pricing calculator, since tier-2 regions cost more.
Cloud Run Cost
instance-seconds = requests × duration ÷ concurrency; cost = instance-seconds × (vCPU × vCPU-second price + GiB × GiB-second price) + idle warm seconds × idle rate + requests ÷ 1,000,000 × request price.
instance-seconds = requests × duration ÷ concurrency; cost = instance-seconds × (vCPU × vCPU-second price + GiB × GiB-second price) + idle warm seconds × idle rate + requests ÷ 1,000,000 × request price. Cloud Run bills the lifetime of an instance, not the sum of request durations, so raising concurrency divides the same request-seconds across fewer instance-seconds and cuts cost almost linearly. Instance-based billing removes the per-request charge but bills the full instance lifetime including idle time, which is why the two models are compared side by side here.
Concurrency is the single most powerful cost lever on Cloud Run and it is free to change, while minimum instances buy away cold starts with a fixed monthly floor you should size deliberately. Confirm the per-vCPU-second and per-GiB-second prices for your region in the Google Cloud pricing calculator, since tier-2 regions cost more.
This calculator takes 13 inputs: Requests per month, Average request duration, Concurrent requests per instance, vCPU per instance, Memory per instance, Price per vCPU-second, Price per GiB-second, Price per million requests, Billing model, Minimum instances kept warm, Idle CPU discount under request billing, Hours in the billing month, Peak requests per second. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.
When your container does useful work outside a request — background processing, streaming, or keeping connections open — or when traffic is steady enough that instances never idle. For spiky traffic, request-based billing with a low minimum instance count almost always wins.