An overview of the tools, discounts, and practices for controlling spending on Google Cloud. Covers the main cost levers and when to use each one.
Why Cost Optimization Matters
Google Cloud is usage-based, and Compute Engine VM vCPU, GPU, and memory resources are billed with a 1-minute minimum and then per-second increments. That precision is a good foundation, but the on-demand price is usually the highest rate you will pay for steady workloads. Most organizations reduce their bill significantly through discounts, rightsizing, and visibility tools.
The key is knowing which lever to pull for your situation.
The Main Cost Levers
| Lever | What It Does | When to Use |
|---|---|---|
| Committed-Use Discounts (CUDs) | Commit to 1 or 3 years of usage for a discount (up to 55% for most hardware resources, up to 70% memory-optimized) | Steady, predictable workloads you will run for months or years |
| Sustained-Use Discounts (SUDs) | Automatic discount for VMs that run a large portion of the month (up to 30% for N1) | Any long-running VM on eligible machine families; no action required |
| Spot VMs | Use spare capacity at up to 91% off | Interruptible workloads: batch jobs, CI/CD, distributed processing |
| Resource Labels | Key-value pairs for cost attribution | Any environment where you need to track spending by team, project, or component |
| Billing Budgets and Alerts | Set spending thresholds and get notified when you approach them | All accounts, especially projects with spending limits |
| Recommender | Automated suggestions to reduce waste (idle VMs, oversized machines, etc.) | Existing infrastructure that has been running for a while |
| Pricing Calculator | Estimate costs before deploying resources | Planning a new deployment or comparing architecture options |
Decision Guide
| Your Situation | What to Use |
|---|---|
| Steady, predictable workloads (production servers, databases) | CUDs for the strongest commitment discount. SUDs apply automatically on eligible families. |
| Interruptible batch jobs, CI/CD, rendering | Spot VMs for the biggest discount. Design for preemption. |
| Unknown or unpredictable costs | Billing Budgets + Alerts to get notified. Enable anomaly detection for automatic deviation alerts. |
| Need to attribute costs to teams or departments | Resource Labels on every resource. Export to BigQuery for reporting. |
| Existing infrastructure, want to cut waste | Recommender to find idle VMs, oversized machines, and unused resources. |
| Planning a new deployment | Pricing Calculator to estimate before you build. Compare regions and machine types. |
| Data residency requirements, egress-heavy workloads | Model network transfer and storage lifecycle costs before choosing regions and storage classes. |
How Discounts Stack
Discounts are mutually exclusive per resource. A single VM cannot receive both a CUD and a Spot VM discount. Google applies them in this priority order:
- Resource-based CUDs (resource-specific commitment discount)
- Compute Flexible CUDs (spend-based, covers eligible Compute Engine, GKE, and Cloud Run usage)
- Sustained-Use Discounts (automatic, applied to remaining on-demand usage)
- Spot VM pricing (separate from the above; Spot VMs are excluded from CUDs and SUDs)
In practice: If an N2 VM hour is covered by a resource-based CUD, that usage gets the CUD discount. Remaining eligible on-demand N2 usage can receive SUDs if it runs long enough. Spot VMs are a separate category with their own pricing.
TL;DR
- CUDs give strong discounts for committed, steady workloads. 1-year or 3-year terms.
- SUDs are automatic (up to 30%) for long-running VMs on eligible families. No purchase needed.
- Spot VMs offer up to 91% off for interruptible workloads. Design for failure.
- Resource Labels enable cost attribution by team, environment, or component.
- Billing Budgets send alerts when spending crosses thresholds. They are not hard caps.
- Recommender finds idle resources and rightsizing opportunities automatically.
- Pricing Calculator helps estimate costs before deployment. Account for egress and free tier separately.
- Discounts are mutually exclusive per resource. CUDs apply first, then SUDs on remaining on-demand usage.
Resources
Google Cloud Cost Management Overview of all cost management tools and best practices.
Cloud Billing Documentation Official documentation for billing accounts, budgets, and exports.
Committed-Use Discounts Resource-based and flexible CUDs, discount rates, and when to buy them.
Sustained-Use Discounts Automatic discounts for long-running VMs on eligible machine families.
Spot VMs Up to 91% off for interruptible workloads, and how to design for preemption.
Billing Budgets and Alerts Set spending thresholds, get notified, and detect cost anomalies.
Resource Labels Cost attribution with key-value pairs and how they differ from tags.
Recommender Automated insights for cost, security, performance, and reliability.
Pricing Calculator Estimate costs before deploying and compare architecture options.