GCP Pricing and Cost Calculation Complete Guide: Mastering Google Cloud Cost Optimization Strategies

GCP Pricing and Cost Calculation Complete Guide: Mastering Google Cloud Cost Optimization Strategies

"Why did this month's cloud bill explode again?"

If you've used GCP, you've probably had this experience. You didn't think you used that much, but the bill was higher than expected.

The problem usually is: you don't understand GCP's billing model well enough.

This article will completely break down GCP's pricing model, from basic billing to various money-saving discounts, to actually using the pricing calculator. After reading, you'll know where your money goes and how to spend less.

Want to understand GCP basics first? Please refer to "GCP Complete Guide: From Beginner Concepts to Enterprise Practice."

GCP Pricing Model Overview

💡 Key Takeaway: GCP's billing logic is actually simple: pay for what you use. But the devil is in the details.

Pay-as-you-go Explanation

This is the most basic billing method.

Features:

• Billed per second (minimum 1-minute charge)
• No upfront payment, no contract
• Stop when done, no charge when stopped
• Highest unit price among all options

Who is it for?

• New users just starting to test GCP
• Projects with unstable, unpredictable usage
• Temporary development or test environments

Note: Even if a VM is shut down, as long as it's not deleted, disks and reserved IPs continue to be billed. This is the most common trap for beginners.

Sustained Use Discounts

This is GCP's unique automatic discount mechanism—AWS doesn't have this.

How it works:

• Within the same month, the longer you use resources of the same spec, the higher the discount
• Completely automatic calculation, no setup or commitment required
• Up to 30% discount

Discount Calculation:

Actual discount is calculated by weighted average—using all month saves about 30%.

Applies to:

• Compute Engine (VMs)
• GKE nodes
• Cloud SQL

Does not apply to:

• E2 and A2 machine types
• Preemptible VMs
• Cloud Run, App Engine, and other serverless services

Committed Use Discounts

Willing to commit to long-term use? Bigger discounts.

Plan Options:

CUD Types:

1. Resource-based CUD

• Commit to specific amounts of vCPU and memory
• Can be used across projects and regions
• More flexible

2. Machine-type CUD

• Commit to specific machine types
• Slightly higher discount
• Less flexible

Purchase Recommendations:

• Observe 3-6 months of actual usage first
• Only buy CUD for stable baseline loads
• Keep 20-30% flexible capacity for traffic fluctuations

Spot VM (Preemptible) Cost Savings

Want extreme savings? Spot VM is the cheapest option.

Price Advantage:

• Only 60-91% off standard VM prices
• Some machine types can save up to 91%

What's the cost?

• Google can terminate your VM at any time (usually 30 seconds notice)
• Maximum 24-hour runtime
• No guarantee resources will be available

Suitable scenarios:

• Batch computing (can interrupt and restart)
• CI/CD pipelines
• Large-scale data processing
• Machine learning training (with checkpoint mechanism)

Not suitable for:

• Production environment web services
• Databases
• Any service that can't be interrupted

GCP Free Tier Complete Contents

GCP's free resources are quite generous—use them well and you can run many small projects.

Always Free Products List

These are monthly free quotas that don't expire:

Compute Services:

• 1 e2-micro VM (limited to us-west1, us-central1, us-east1)
• 720 hours per month (exactly one 24/7 running machine)

Storage Services:

• 5 GB Cloud Storage (Regional)
• 1 GB Cloud Firestore
• 10 GB Cloud Storage for Firebase

Database:

• Cloud Firestore: 50,000 reads, 20,000 writes, 20,000 deletes (per day)

Serverless:

• Cloud Functions: 2 million invocations, 400,000 GB-seconds compute
• Cloud Run: 2 million requests, 360,000 GB-seconds compute

BigQuery:

• 1 TB queries per month
• 10 GB storage

Other:

• Cloud Build: 120 minutes per day
• Container Registry: 500 MB storage
• Cloud Vision API: 1,000 calls per month

Free Trial Credits ($300) Usage Rules

New users get additional trial credits.

Credit Contents:

• $300 USD
• Must use within 90 days

What can it be used for?

• Almost all GCP services
• Can run larger machines than free tier
• Can test GPU/TPU

Limitations:

• Cannot be used for cryptocurrency mining
• Cannot create VMs with more than 8 vCPUs
• Some advanced services have restrictions

Free Tier Limitations and Considerations

Free doesn't mean completely free.

Common Overspending Situations:

1. Wrong Region: e2-micro free is limited to specific regions; choosing other regions incurs charges

2. Network Traffic: Free tier doesn't include network egress; transferring data out costs money

3. Additional Services: Enabling Cloud Armor, Cloud CDN, etc. incurs fees

4. Snapshots and Images: Created snapshots and custom images take up storage

Recommended Practices:

• Set budget alerts on day one
• Regularly check billing details
• Delete unused resources immediately

GCP Pricing Calculator Tutorial

Want to know how much a project will cost? Use the Pricing Calculator first.

Google Cloud Pricing Calculator Steps

Step 1: Access the Calculator

Go to cloud.google.com/products/calculator

Step 2: Select Services

Left menu to select services to estimate, for example:

• Compute Engine
• Cloud Storage
• BigQuery
• Cloud SQL

Step 3: Enter Specifications

Using Compute Engine as example:

• Select region (asia-east1 = Taiwan)
• Select machine type (n2-standard-2)
• Set quantity
• Select operating system
• Set disk size

Step 4: View Estimate

Right side shows:

• Monthly cost estimate
• Item breakdown
• Applicable discounts

Step 5: Export or Share

You can:

• Export as JSON
• Generate shareable link
• Download PDF report

Common Service Cost Estimation Examples

Here are reference prices for asia-east1 region:

Compute Engine VMs:

Cloud Storage:

BigQuery:

Interpreting Results and Optimization Recommendations

After getting estimates, how to optimize?

If costs are too high:

1. Downsize: Many projects start too big, don't actually need that many resources

2. Switch Machine Types: E2 series is usually 20-30% cheaper than N2

3. Consider Spot VMs: Batch jobs using Spot can save 60%+

4. Adjust Storage Class: Infrequently accessed data should use Nearline or Coldline

5. Evaluate CUD: If usage is stable, purchase committed discounts

Common Cost Estimation Mistakes:

• Forgetting to calculate network egress
• Not considering IP address costs
• Missing snapshot and backup storage
• Ignoring log storage costs

Service Pricing Details

Each service has different billing logic. Here are the most commonly used services' cost structures.

Compute Engine (VM) Cost Structure

VM Cost = Compute Cost + Storage Cost + Network Cost

Compute Cost:

• Charged separately by vCPU and memory
• Different machine series have different prices (E2 < N2 < C2)
• Premium OS (like Windows, RHEL) charges extra

Storage Cost:

• Standard HDD: $0.04/GB/month
• Balanced SSD: $0.10/GB/month
• Performance SSD: $0.17/GB/month

Network Cost:

• Same-region transfer: Free
• Cross-region transfer: $0.01-0.02/GB
• External network egress: Starting $0.12/GB

Cloud Storage Costs

Cloud Storage has three cost components:

1. Storage Cost Depends on storage class and region, listed earlier.

2. Operation Costs

3. Data Retrieval Costs Nearline, Coldline, Archive have retrieval costs in addition to operation costs when reading data.

Selection Recommendations:

• Frequently accessed → Standard
• Backup data (may access monthly) → Nearline
• Compliance storage (rarely accessed) → Coldline
• Long-term archive (almost never accessed) → Archive

BigQuery Query and Storage Costs

BigQuery is one of GCP's best value services, but pay attention to usage.

Two Billing Models:

1. On-demand Billing

• $5 per TB queried
• No charge when not querying
• Suitable for occasional analysis

2. Capacity Billing

• Purchase fixed query capacity (Slots)
• Suitable for high-volume, frequent queries

Cost-Saving Tips:

• Only select needed columns (SELECT * is a big mistake)
• Use partitioned tables
• Set query limits (max X TB per day)
• Regularly clean unused tables and datasets

For detailed service operations, see "GCP Core Services Hands-on Tutorial."

Cloud Run / GKE Container Service Costs

Cloud Run:

• CPU: $0.00002400/vCPU-second
• Memory: $0.00000250/GB-second
• Requests: $0.40/million
• Can scale to 0 when no requests, completely free

GKE:

• Cluster management fee: $0.10/hour (~$72/month)
• Autopilot mode: Billed by Pod resources
• Standard mode: Billed by node VMs

Selection Recommendations:

• Small services, unstable traffic → Cloud Run
• Large-scale, complex microservice architecture → GKE

GCP Cost Optimization Practical Tips

Now that you know how billing works, let's look at how to save money.

Setting Budget Alerts and Quotas

Budget Alert Setup:

1. Go to Billing → Budgets & alerts
2. Create new budget
3. Set amount and alert thresholds (recommend 50%, 80%, 100%)
4. Choose notification method (Email, Pub/Sub)

Quota Limits:

• Can set daily query quotas
• Limit number of VMs that can be created
• Prevent accidentally spinning up many resources

Using Recommender Suggestions

GCP automatically analyzes your usage and gives optimization recommendations.

Where to find them?

• Console → Recommendations Hub
• Or each service page's "Recommendations" tab

Common Recommendation Types:

• Idle resource alerts (this VM isn't being used)
• Sizing recommendations (this machine is oversized)
• CUD purchase recommendations (your usage fits committed discounts)
• Over-provisioned disks

Resource Labels and Cost Allocation

Why need labels?

• Know which project money is spent on
• Department cost allocation
• Identify which resources can be cut

Recommended Label Strategy:

Cost Report Usage:

• Billing → Reports
• Can filter by labels
• Export to BigQuery for more detailed analysis

Regular Idle Resource Review

Spend 30 minutes each month checking these:

Checklist:

• Idle VMs (CPU usage < 5%)
• Unattached disks
• Expired snapshots
• Unused static IPs (each $7/month)
• Empty GKE clusters
• Load Balancers with no traffic

Common Cost Traps and Avoidance Guide

Here are the most common pitfalls.

Network Egress Costs

This is the most commonly overlooked cost.

GCP ingress is free, but egress costs money:

• To internet: Starting $0.12/GB
• To other clouds: $0.12/GB
• Cross-region: $0.01-0.02/GB

Ways to Save:

• Use Cloud CDN to cache static content
• Choose regions close to users
• Compress transferred data
• Consider Premium vs Standard network tier

Cross-Region Data Transfer

Scenario: Your VM is in asia-east1 (Taiwan), database is in us-central1 (US).

Every database query crosses the Pacific, and traffic costs can be substantial.

Solutions:

• Keep related services in the same region as much as possible
• If cross-region is necessary, consider data replication strategies
• Evaluate latency and cost tradeoffs

Uncleaned Snapshots and Disks

Common Situations:

• Created many snapshots "just in case," never used them
• Deleted VM without checking "Delete boot disk"
• Forgot to clean up after testing

Cost Accumulation:

• Snapshots: $0.026/GB/month
• Unattached SSD: $0.17/GB/month

100GB of snapshots over a year is $31. 10 snapshots is $310.

Recommended Practices:

• Set snapshot retention policies (auto-delete after X days)
• Run cleanup scripts regularly
• Use labels to mark temporary resources

Cloud Bills Giving You Headaches?

Many enterprises can actually save 20-40% on cloud spending.

Free Bill Health Check—let us help you find hidden cost traps.

CloudSwap's Cost Optimization Services:

• Complete bill analysis to find wasteful items
• Discount plan evaluation (optimal CUD, SUD combination)
• Architecture optimization recommendations to reduce unnecessary spending
• Budget alert and monitoring system setup

We've helped over 50 enterprises save an average of 28% on cloud spending.

Conclusion: GCP Cost Management Best Practices

GCP cost management isn't hard, but requires ongoing attention.

Five Must-Do Things:

1. Set budget alerts on day one: Don't wait until month-end to discover overspending

2. Use automatic discounts: SUD is automatic, but confirm your machine types qualify

3. Buy CUD for stable loads: After observing 3-6 months, purchase committed discounts for stable usage

4. Clean up monthly: Idle resources, snapshots, unused IPs—clean regularly

5. Track costs with labels: Know where money goes to optimize effectively

Remember this principle: The key to cloud savings isn't "using the cheapest" but "only using what's needed."

Further Reading

• To understand GCP basics, refer to GCP Complete Guide
• To learn detailed service operations, see GCP Core Services Hands-on Tutorial
• To compare GCP and AWS costs, see GCP vs AWS Cloud Platform Complete Comparison
• To learn about AI service costs, see GCP AI/ML and Vertex AI Complete Guide

Image Descriptions

References

1. Google Cloud, "Google Cloud Pricing" (2024)
2. Google Cloud, "Sustained Use Discounts" (2024)
3. Google Cloud, "Committed Use Discounts" (2024)
4. Google Cloud, "Free Tier" (2024)
5. Google Cloud, "Pricing Calculator" (2024)

Need Professional Cloud Advice?

Whether you're evaluating cloud platforms, optimizing existing architecture, or looking for cost-saving solutions, we can help

Book Free Consultation