Paste a line item from your invoice and find what it bills for, across the Google Cloud services covered so far.
Cloud Run CPU and Memory Allocation
Cloud Run charges for the CPU and memory allocated to your container.
Request-based billing charges while it handles requests; instance-based billing charges for the whole time an instance exists. Which mode you are in is the single biggest factor in a Cloud Run bill, and minimum instances add idle-time charges even in request-based mode.
Billed: Per vCPU-second and gibibyte-second, rounded up to the nearest 100 milliseconds, plus a per-request charge in request-based mode, each after a monthly free tier per billing account. Billing differs sharply between request-based and instance-based modes.
What triggers this charge
- Request-based billing charges CPU and memory only while an instance starts, handles requests, or shuts down, plus a per-request fee, and scales to zero when idle.
- Instance-based billing (formerly "CPU always allocated"), needed for background work, charges for the whole instance lifetime regardless of traffic, with a 1-minute minimum, at a lower rate and with no per-request fee. Cloud Run jobs always bill this way, and worker pools bill for every instance they keep running.
- Minimum instances keep containers warm around the clock. Under request-based billing their idle time is billed at a reduced CPU rate and the full memory rate, which is a latency purchase, not a scaling one.
How to reduce it
Set minimum instances to zero where cold starts are tolerable
Minimum instances is often set during a launch to smooth latency and never revisited. For internal and batch services the cold start rarely matters.
Check the concurrency setting
A new service gets 80 concurrent requests per instance from the console, or 80 per vCPU from gcloud or Terraform, up to a maximum of 1,000. Services set to 1 or a low value often handle far more. Higher concurrency means fewer instances for the same traffic.
Right-size the container
CPU and memory are allocated per instance and billed in 100-millisecond steps. A service given 4 vCPU because someone once saw a spike is paying for it on every request.
Rates
Sources and review
- Explanation reviewed:
- 1 Nov 2026, against Cloud Run pricing and the Cloud Run concurrency docs.
- Rates verified:
- not yet. No rate is shown on this page until one is loaded from the Cloud Billing Catalog API.
Now find out how much of this you are actually paying.
Finitizer breaks your Google Cloud bill down to the SKU, shows how each one has moved since last week, and separates rate changes from usage changes, so you know whether to talk to engineering or to procurement.
See how Finitizer works