To reduce Pub/Sub and Eventarc cost on Google Cloud, rank topics by throughput, trim message retention, batch and compress payloads, delete idle subscriptions and snapshots, and prefer standard delivery over BigQuery or Cloud Storage subscriptions unless you need the export. Throughput is the main charge, at 40 US dollars per TiB after a 10 GiB monthly free tier, and fan-out multiplies it because each subscription receives a full copy of every message. Eventarc adds no separate per-event charge for most paths; its cost surfaces as the Pub/Sub throughput and the Cloud Run or function invocations it triggers, so optimizing those optimizes Eventarc.
Last updated: June 2026. Written by Fredrik Filipsson and reviewed by Morten Andersen, built on our See, Cut, Lock, Run method.
This guide sits in our complete guide to Google Cloud cost optimization, the cluster pillar it links up to, and connects to the broader cloud cost optimization playbook. Messaging and logging waste tend to appear together, so read it alongside our companion guide on how to cut Cloud Logging and Log Router costs on GCP. Trimming message throughput is a Cut step move against a bill that grows with traffic.
Fan-out. Pub/Sub bills throughput on both publish and delivery, so a topic with five subscriptions delivers each message five times and bills accordingly. A single high-volume topic feeding several consumers can cost far more than the publish volume alone suggests. Before tuning message size, count subscriptions per topic and remove the ones nobody consumes, then batch and compress what remains. The cheapest message is one that is never delivered to a subscription nobody reads.
How is Pub/Sub priced on Google Cloud?
Pub/Sub is priced mainly on throughput, the volume of message data published and delivered. Per the Pub/Sub pricing documentation, throughput is 40 US dollars per TiB in all regions after a 10 GiB monthly free tier, retained message storage is billed separately per GiB-month, and BigQuery or Cloud Storage subscriptions carry a higher read-side rate of around 50 US dollars per TiB. Because throughput dominates most bills and is charged on both publish and delivery, reducing byte volume and unnecessary fan-out is the primary lever. Verify current per-TiB rates in the linked pricing page before modeling, because Google Cloud prices change.
Does Eventarc cost extra on top of Pub/Sub?
Eventarc has no separate per-event charge for most paths; you pay for the underlying resources it uses, principally Pub/Sub throughput and the compute target that receives the events. An Eventarc trigger provisions Pub/Sub topics and subscriptions under the hood, so its cost shows up as Pub/Sub throughput plus the Cloud Run or Cloud Functions invocations it triggers. That means the way to optimize Eventarc is to optimize what it sits on: reduce the event volume flowing through Pub/Sub, and make sure the triggered compute is right-sized and not over-invoked. Treat Eventarc as a routing layer whose bill lives in the services beneath it.
How do I reduce the cost, step by step?
Reduce it by cutting throughput, retention, and unnecessary delivery. Five steps:
- Find the high-throughput topics. Rank topics and subscriptions by published and delivered bytes in monitoring. Result: the volume to target first.
- Trim message retention. Lower topic and subscription retention to what consumers need. Result: less retained-byte storage cost.
- Batch and compress messages. Combine small messages and compress payloads. Result: fewer billable bytes and less request overhead.
- Remove unused snapshots and subscriptions. Delete idle subscriptions, snapshots, and seek retention. Result: no fan-out or storage for consumers nobody reads.
- Choose the cheapest delivery path. Prefer standard delivery unless a BigQuery or Cloud Storage subscription is genuinely needed. Result: avoid the higher per-TiB export rate.
Is fan-out quietly multiplying your messaging bill?
Our Google Cloud cost audit maps your topics and subscriptions, finds redundant fan-out and oversized messages, trims retention, and quantifies the Pub/Sub and Eventarc spend you can reclaim. On the performance model, you pay only from realized savings. No savings, no fee.
Book a Google Cloud cost audit →Standard subscription or BigQuery subscription?
Choose a BigQuery or Cloud Storage subscription only when you genuinely need messages written directly to those destinations, because they cost more per TiB than standard pull or push delivery. The native export can still be the cheaper option if a consumer would otherwise run its own pipeline to load the same data, since you remove that compute and code; but if you do not need the data in BigQuery or Cloud Storage, standard delivery is the lower-cost path. The decision is whether the built-in export actually saves you work, not whether it is convenient. Match each subscription type to the consumer's real need, and keep the high-rate export subscriptions only where they replace something more expensive.
The Google Cloud Cost Optimization Field Guide includes the Pub/Sub throughput and fan-out worksheet and the subscription selection matrix referenced above. It is the downloadable companion to this guide.
Frequently asked questions
How is Pub/Sub priced on Google Cloud?
Pub/Sub is priced mainly on throughput, the volume of message data published and delivered, at 40 US dollars per TiB in all regions after a 10 GiB monthly free tier. Retained message storage is billed separately per GiB-month, and BigQuery or Cloud Storage subscriptions carry a higher rate of around 50 US dollars per TiB for the read side. Throughput dominates most bills, so reducing byte volume is the primary lever. Verify current rates in the Pub/Sub pricing page.
Does Eventarc cost extra on top of Pub/Sub?
Eventarc itself has no separate per-event charge for most paths; you pay for the underlying resources it uses, principally Pub/Sub throughput for the events it routes and the compute target that receives them. In practice an Eventarc trigger creates Pub/Sub topics and subscriptions under the hood, so Eventarc cost shows up as Pub/Sub throughput and as the Cloud Run or function invocations it triggers. Optimizing Pub/Sub and the target compute optimizes Eventarc.
What is the biggest Pub/Sub cost driver?
The biggest driver is throughput, the total bytes published and delivered, especially when many subscriptions each receive a full copy of every message. Fan-out multiplies delivery volume, so a topic with several subscriptions bills delivery for each. Reducing message size, batching, and removing unnecessary subscriptions are the most effective moves. Retained-byte storage from long retention and idle snapshots is a common secondary driver.
Should I use a BigQuery subscription or a standard subscription?
Use a BigQuery or Cloud Storage subscription only when you genuinely need messages written directly to those destinations, because they cost more per TiB than standard pull or push delivery. If a consumer already loads data into BigQuery, the native subscription can still be cheaper than running your own pipeline; if not, standard delivery is the lower-cost path. Match the subscription type to whether the built-in export actually saves you work.
The short version
Pub/Sub bills throughput on publish and delivery, so cut bytes and fan-out first: rank topics, trim retention, batch and compress, and drop idle subscriptions, then pick standard delivery unless an export subscription earns its higher rate. Eventarc rides on those same resources, so optimizing Pub/Sub and the triggered compute optimizes both. When you want fan-out mapped and message volume trimmed across your projects, that is what our Google Cloud cost optimization service delivers.
Cloud pricing and service behavior change frequently. Verify the specifics in this guide against the providers’ own current documentation and the FinOps Foundation: Google Cloud pricing ↗, Google Cloud documentation ↗ and FinOps Foundation Framework ↗. This article also reflects Cloud Cost Room’s hands-on, vendor-neutral engagement experience.