GCP Free Tier and Billing Overview
How the Google Cloud Free Trial and Free Tier work, how billing accounts connect to projects, and how to set budgets, alerts and spend caps.
Google Cloud gives new users two separate things for free: a trial credit that runs out, and a set of monthly allowances that don’t. People tend to blur them together, and that is where most surprises come from. A VM that was free during the trial can start costing money the day the trial ends, while a small Cloud Run service may stay free for years.
This article explains how the Free Trial and the Free Tier work today, how Google Cloud billing is structured around billing accounts and projects, and which guardrails to set up before you build anything: budgets, anomaly alerts, labels and, for a few services, a hard spend cap.
Two programs: the Free Trial and the Free Tier
The Free Trial is for new customers. You get $300 in Cloud Billing credits to spend on almost any Google Cloud product, and it lasts 90 days. It ends when the credit is used up or the 90 days pass, whichever comes first. The clock starts at sign-up and can’t be paused or extended.
The Free Tier is a list of products with monthly usage allowances that apply to every billing account, during the trial and after it, on paid accounts too. It has no end date, although Google can change the limits with 30 days’ notice. Usage within those limits is never charged, and during the trial it doesn’t draw down your credit either.
| Free Trial | Free Tier | |
|---|---|---|
| Who gets it | New Google Cloud customers, once | Every billing account |
| What you get | $300 in credits | Fixed monthly allowances on selected products |
| How long | 90 days or until the credit is spent | No end date |
| Products covered | Most products, with some restrictions | A specific list, each with its own limit |
| Unused amount | Expires at the end of the trial | Resets monthly, doesn’t roll over |
| Can produce a bill | No, unless you activate a paid account | Yes, for usage above the limits on a paid account |
What the Free Trial restricts
Trial accounts can use most of Google Cloud, but not all of it. You can’t attach GPUs to VMs, you can’t run Windows Server VMs, and you can’t request quota increases. If you need any of those, you have to activate a full account, which turns the trial billing account into a normal paid one. Any unexpired trial credit stays on the account when you do.
Sign-up asks for a payment method to verify your identity. During the trial it isn’t charged; Google only bills it after you activate a full account.
What happens when the trial ends
When the credit runs out or the 90 days pass, everything you created is stopped. You then have a 30-day grace period in which you can activate a full account and recover your resources. Data in Compute Engine is marked for deletion at the end of the trial, so recovery is likely but not guaranteed. After the 30 days, the trial’s resources are gone.
flowchart TD
A["Sign up"] --> B["$300 credit, 90-day trial"]
B --> C{"Activate full account?"}
C -->|"Yes, any time"| P["Paid billing account"]
C -->|No| D{"Credit spent or 90 days passed?"}
D -->|No| B
D -->|Yes| E["Resources stopped"]
E --> F{"Activate within 30 days?"}
F -->|Yes| P
F -->|No| G["Trial resources deleted"]
P --> H["Free Tier still applies"]
The useful consequence: a trial account can’t produce a bill. The risk on a trial is losing your work, not losing money. The risk shifts once you activate, because from then on anything above the Free Tier is charged to your card.
You started the Free Trial 40 days ago and have already spent all $300 of the credit. What happens?
Correct! The trial ends when the credit is used or 90 days pass, whichever comes first. Your card is only charged after you activate a full account.
What the Free Tier includes
The Free Tier list is long, and Google updates it. These are the allowances most people use when learning, as published at the time of writing. All limits are per month and calculated per billing account, not per project.
| Product | Monthly allowance | Notes |
|---|---|---|
| Compute Engine | 1 e2-micro VM, 30 GB-months of standard persistent disk, 1 GB of outbound data transfer | Only in us-west1, us-central1 and us-east1 |
| Cloud Storage | 5 GB-months of Standard storage | Only in us-west1, us-central1 and us-east1, combined |
| Cloud Run | 2 million requests, 180,000 vCPU-seconds, 360,000 GiB-seconds | Request-based billing |
| Cloud Run functions | 2 million invocations, plus a compute allowance | HTTP and event-driven invocations |
| BigQuery | 1 TiB of queries, 10 GiB of storage | Query allowance is on-demand pricing |
| Pub/Sub | 10 GiB of messages | |
| Firestore | 1 GiB stored, 50,000 reads, 20,000 writes and 20,000 deletes per day | One database per project |
| Artifact Registry | 0.5 GB of storage | |
| Secret Manager | 6 active secret versions, 10,000 access operations |
Check the Free Tier usage limits before you rely on any of these numbers. The point of the table is the shape of the offer, not the exact figures.
The fine print on the free VM
The e2-micro allowance is the one people most often get wrong, so it is worth reading closely.
- It is measured in hours, not instances. Each month, e2-micro usage is free until it adds up to the number of hours in that month. One VM running all month uses the whole allowance. Two VMs running all month means you pay for one of them.
- Region matters. An e2-micro in us-west1 (Oregon), us-central1 (Iowa) or us-east1 (South Carolina) qualifies. The same VM in any other region is billed in full.
- Disk type matters. The 30 GB covers standard persistent disk. Compute Engine often suggests a balanced persistent disk as the boot disk, which is not covered, so change it when you create the VM.
- Spot VMs don’t qualify. The allowance is for a regular, non-preemptible e2-micro.
- Network egress is limited. You get 1 GB a month of outbound data transfer from North America, excluding destinations in China and Australia. A site that gets real traffic will use more than that.
Older guides, including an earlier version of this article, mention a free f1-micro instance. That machine type belongs to a previous generation of the offer; the free VM is now an e2-micro. Launching Your First GCP Compute Engine Instance walks through creating one with the right region and disk settings.
You run two e2-micro VMs in us-central1 for a full month on a paid account. What do you pay for?
Correct! The allowance is a month's worth of e2-micro hours per billing account, so two full-time VMs use it twice over.
How Google Cloud billing is structured
Everything you create in Google Cloud lives in a project. Projects don’t hold a payment method themselves. Each project is linked to one Cloud Billing account, which holds the payment method and receives the charges. One billing account can pay for many projects.
flowchart TD
PAY["Payments profile"] --> BA["Cloud Billing account"]
BA --> P1["Project: blog-prod"]
BA --> P2["Project: blog-dev"]
BA --> P3["Project: experiments"]
P1 --> R1["Cloud Run service"]
P2 --> R2["e2-micro VM"]
P3 --> R3["BigQuery dataset"]
This split has a few practical effects:
- Free Tier limits are shared across the billing account. Three projects on one billing account share a single e2-micro allowance, not three.
- Projects are the unit for cleanup. Shutting down a project stops its billing for everything inside it. That makes a separate project per experiment a simple way to keep costs contained, since you can remove the whole thing in one step when you’re done.
- Billing access is separate from project access. Being an Owner on a project doesn’t let you see the bill. The main billing roles are Billing Account Administrator (full control), Billing Account User (can link projects to the account), Billing Account Viewer (read-only cost data) and Project Billing Manager (can link or unlink one project). GCP Identity and Access Management (IAM) explains how roles and bindings work.
If you unlink a project from its billing account, or the billing account is closed, paid services in that project stop. That includes resources that would otherwise fit inside the Free Tier, because Free Tier usage still needs an active billing account.
Where charges come from
Google Cloud bills monthly for metered usage: compute per second, storage per GB-month, requests by count, and network egress per GB. Prices differ between regions. A few patterns cause most unexpected charges:
- Stopped VMs still have disks. Stopping a VM ends compute charges, but its persistent disks keep billing until you delete them.
- Reserved static IP addresses cost money, and an address that isn’t attached to anything is still billed.
- Egress adds up. Data coming into Google Cloud is generally free. Data going out to the internet, and in many cases between regions, is not.
- Some services bill while idle. A Cloud SQL instance, a GKE cluster or a load balancer is billed for the time it exists, not for the traffic it serves.
Seeing what you spend
The billing tools live in the Google Cloud console under Billing. If you’re new to the console, Navigating the GCP Console covers how to find your way around.
| Tool | What it shows or does |
|---|---|
| Overview | Current month’s costs, forecast and trend |
| Reports | Costs over time, grouped and filtered by project, service, SKU, region or label |
| Cost table | Line-item detail for an invoice month, exportable to CSV |
| Credits | Remaining trial and promotional credits and when they expire |
| Budgets and alerts | Thresholds that send notifications when spend or forecast crosses them |
| Anomalies | Automatically detected spending spikes, with root-cause details |
| Billing export | Detailed usage and cost data streamed into BigQuery |
Reports are the place to start. Grouping by service for the last 30 days answers “what am I paying for?”, and grouping by project answers “which experiment is it?”. Billing data isn’t instant: charges can take up to a day to appear, so a report shows where you were, not where you are.
For anything beyond the built-in reports, enable the billing export to BigQuery. It doesn’t hold your full history, only data from around when you turn it on, so enable it early even if you don’t plan to query it yet. Querying a small export usually fits easily inside BigQuery’s free 1 TiB a month.
Guardrails to set up on day one
None of these take long, and together they cover most mistakes.
flowchart LR
U["Usage in projects"] --> D["Cloud Billing data"]
D --> B["Budget thresholds"]
D --> A["Anomaly detection"]
D --> SC["Spend cap (eligible services)"]
B -->|email| Y["You"]
A -->|email| Y
B -->|Pub/Sub| AUT["Automation"]
SC -->|"cap reached"| PZ["Requests paused"]
Create a budget
A budget compares your actual or forecast spend with an amount you set and sends alerts at thresholds you choose. By default, alert emails go to the billing account’s administrators and users. A sensible starting point for a personal account is a small monthly amount with alerts at 50 and 90 percent of actual spend, plus one at 100 percent of forecast spend. The forecast alert matters most, because it can warn you mid-month before the money is spent.
You can create a budget in the console or with gcloud:
gcloud billing budgets create \
--billing-account=000000-AAAAAA-BBBBBB \
--display-name="monthly-cap" \
--budget-amount=10USD \
--threshold-rule=percent=0.5 \
--threshold-rule=percent=0.9 \
--threshold-rule=percent=1.0,basis=forecasted-spend
Find your billing account ID with gcloud billing accounts list. Without a filter, the budget covers every project on the billing account, which is usually what you want for a personal account. Forecast thresholds only work with calendar periods such as monthly, not custom date ranges.
The important limitation: a standard budget does not stop anything. It sends notifications, and your resources keep running. If you need it to act, you can connect the budget to a Pub/Sub topic and have a function react to the message, for example by disabling billing on a project. Google documents this pattern, but treat it as a last resort: disabling billing shuts down every service in the project, including Free Tier ones, and because billing data lags, it won’t react instantly.
Use a spend cap where it is available
Google Cloud has added spend cap budgets for a small set of services: currently Cloud Run, Cloud Run functions, the Gemini API and Gemini Enterprise Agent Platform (formerly Vertex AI). A spend cap applies to one service in one project. When it is reached, new requests to that service in that project are paused until you lift the cap or the next month starts, while other projects and services carry on.
Spend caps use estimated costs, so they react faster than regular budgets, but enforcement still isn’t instant and any overage is billed. They are most useful for a public endpoint where a traffic spike or a runaway loop could run up a bill. Check Manage spend cap budgets for the current list of eligible services.
Your monthly budget is $10 with an alert at 100 percent. A forgotten VM pushes spend to $15. What did the budget do?
Correct! Standard budgets only notify. Stopping resources needs a spend cap on an eligible service or your own automation on budget notifications.
Check anomaly alerts are on
Cost anomaly detection watches your spending pattern and flags unusual spikes, such as a service you’ve never used suddenly costing money. It is free and enabled by default, with alerts sent to billing administrators. It also works on new accounts without spending history. Open the Anomalies page once to confirm who receives the alerts and adjust the threshold if the defaults are too noisy.
Label your resources
Once you run more than one thing, the bill’s breakdown by service stops being enough. Labels are key-value pairs on resources, such as project=blog or env=dev, and they show up in billing reports and the BigQuery export, so you can group costs by them.
gcloud compute instances add-labels my-vm \
--zone=us-central1-a \
--labels=project=blog,env=dev
Unlike tags on some other clouds, labels don’t need to be activated for billing; they appear in reports as soon as charges with those labels are recorded. Keep the scheme small: two or three keys applied consistently beat a dozen applied at random. Separate projects often do the same job more simply, since every report can already group by project.
A cleanup habit
Most surprise charges come from resources left behind after a tutorial. A short routine after each experiment prevents them:
- If the experiment had its own project, shut the project down. That stops billing for everything in it. The project stays recoverable for about 30 days, then is deleted.
- Otherwise, delete VMs rather than stopping them, and check the Disks page for anything left over.
- Release static IP addresses you reserved, and delete load balancers, Cloud SQL instances and GKE clusters you no longer need.
- Look at billing reports grouped by project and region for the last few days. Anything in a region you weren’t working in deserves a look.
How this compares with AWS and Azure
All three providers combine a short-lived credit with permanent allowances, but the details differ. Google’s trial gives a larger single credit than AWS’s sign-up credit and never bills you, and its Free Tier is shared across every project on a billing account. AWS Free Tier and Billing Overview and Azure Free Tier and Billing Overview cover the other two, and Setting Up Free Accounts on AWS, GCP and Azure walks through signing up for all three.
If you get stuck on a billing question, billing support is available at no cost to every Google Cloud customer. GCP Support and Documentation Resources explains where to ask.
Further reading
- Free Google Cloud features and trial offer
- Cloud Billing overview
- Create, edit, or delete budgets and budget alerts
- Manage spend cap budgets
- Export Cloud Billing data to BigQuery
- gcloud billing budgets create reference
Key takeaways
- The Free Trial gives new customers $300 in credits for 90 days, ends at whichever limit comes first, and never bills you unless you activate a full account.
- When the trial ends, resources stop and you have 30 days to activate a paid account before they are deleted.
- The Free Tier is separate and permanent: monthly allowances such as one e2-micro VM in three US regions, 2 million Cloud Run requests and 1 TiB of BigQuery queries.
- Free Tier limits are counted per billing account, so every project on that account shares them.
- Standard budgets only send alerts. Spend caps can pause a few services, such as Cloud Run and the Gemini API, when a limit is reached.
- Set up a budget with a forecast alert, confirm anomaly alerts, enable the BigQuery billing export, and use a project per experiment so cleanup is one step.
Part 5 of 7 · Exploring Google Cloud Platform (GCP)