GCP monitoring cost 2026: the Cloud Operations bill explained

Google Cloud bundles observability into the Cloud Operations suite, formerly Stackdriver: Cloud Monitoring for metrics, Cloud Logging for logs, and Cloud Trace for distributed traces. Unlike AWS, which meters CloudWatch per custom metric and per alarm, or Azure, which folds most telemetry into a single Log Analytics ingestion charge, GCP prices each pillar on its own consumption unit. Understanding the bill means understanding three meters, and which one dominates depends entirely on the shape of your workload.

Cloud Logging is the meter most teams meet first. It charges $0.50 per GiB of logs ingested, identical to AWS CloudWatch Logs and far below Azure Monitor's $2.30 per GiB default plan, after a free allowance of 50 GiB per project per month, the most generous of the three hyperscalers. That $0.50 charge includes 30 days of storage in the destination bucket, so for most operational logging there is no separate retention cost at all. Retention only bills, at $0.01 per GiB per month, when you deliberately hold a bucket's logs longer than the default 30 days.

Cloud Monitoring is the meter that surprises people. It bills metrics by ingested volume in MiB, not per metric or per host: $0.2580 per MiB for the first 150 MiB to 100,000 MiB each month, stepping down to $0.1510 per MiB up to 250,000 MiB and $0.0610 per MiB beyond. The first 150 MiB per billing account is free, and Google Cloud's own system metrics are never chargeable. The charge is on user-defined and agent-collected metrics, and because every label value multiplies the time series count, high-cardinality workloads (Kubernetes especially) drive this meter far harder than a host count would suggest.

Cloud Trace is usually the smallest line item. It charges $0.20 per million spans ingested after the first 2.5 million free each month, so even a busy microservice estate tracing tens of millions of spans pays single-digit to low-tens of dollars. All of these rates are list pay-as-you-go figures; committed-use and Enterprise Agreement discounts apply on top, and prices can vary by region.

A small project of 10 VMs sending modest application and platform logs often stays inside the free tiers entirely. Fifty GiB of free logs per project per month covers a light workload, the first 150 MiB of chargeable metrics is free, and 2.5 million free trace spans absorb a low-traffic service. This is the scenario where GCP is effectively free for observability, and it is genuinely common for early-stage projects, because the free allowances are the most generous of the three major clouds.

A mid-market project of 100 VMs pushing around 100 GiB per day of logs ingests roughly 3,000 GiB per month. After the 50 GiB free allowance, that is about 2,950 GiB at $0.50, near $1,475 per month for logging alone, with the included 30-day retention covering most needs. Metrics add on top: a GKE-heavy project exporting custom and Prometheus-style series can push metrics ingestion well past the free 150 MiB into the $0.2580 per MiB tier, often adding hundreds of dollars depending on cardinality. Trace stays small, tens of millions of spans landing in the tens of dollars.

An enterprise or platform-team workload is where routing discipline decides the bill. The lever is the Logs Router: excluding high-volume, low-value logs (load balancer access logs, VPC flow logs, verbose debug output) from the _Default bucket and exporting them to Cloud Storage or BigQuery via a sink avoids the $0.50 per GiB Logging ingestion charge entirely, leaving only the destination's far cheaper storage rate. Combined with cardinality discipline on metrics and a committed-use discount, a tuned configuration routinely runs at a fraction of a naive all-_Default, all-metrics setup.