GCP for DevOps — Discounts & Pricing

4 minute read

Discounts & Pricing

GCP offers several discount mechanisms to reduce costs for predictable workloads. Understanding committed use discounts (CUDs), sustained use discounts (SUDs), Spot VMs, and custom machine types helps DevOps teams optimize infrastructure spend significantly.

GCP Discount Types Overview

GCP Discount Landscape:
┌──────────────────────────────────────────────────────────┐
│                                                           │
│  Automatic Discounts          Commitment Discounts        │
│  ┌─────────────────────┐     ┌────────────────────────┐  │
│  │  Sustained Use       │     │  Committed Use (CUD)   │  │
│  │  • Auto-applied      │     │  • 1-year: ~37% off    │  │
│  │  • Up to 30% off     │     │  • 3-year: ~55% off    │  │
│  │  • Compute Engine    │     │  • Spend or resource    │  │
│  │  • After 25% month   │     │  • Non-cancellable     │  │
│  └─────────────────────┘     └────────────────────────┘  │
│                                                           │
│  Spot/Preemptible VMs         Free Tier / Always Free     │
│  ┌─────────────────────┐     ┌────────────────────────┐  │
│  │  Spot VMs            │     │  Free Tier              │  │
│  │  • Up to 91% off     │     │  • e2-micro (1/month)  │  │
│  │  • Can be preempted  │     │  • 5 GB Cloud Storage  │  │
│  │  • Batch, CI/CD      │     │  • 1 GB App Engine     │  │
│  │  • Max 24h runtime   │     │  • Many services       │  │
│  └─────────────────────┘     └────────────────────────┘  │
└──────────────────────────────────────────────────────────┘

Sustained Use Discounts (SUDs)

SUDs are automatic discounts applied when you run Compute Engine resources for a significant portion of the month. No commitment required.

Sustained Use Discount Tiers:
┌────────────────────────────────────────────────┐
│  Monthly Usage    │  Effective Discount         │
├────────────────────────────────────────────────┤
│  0% - 25%        │  Full price                 │
│  25% - 50%       │  20% discount on this tier  │
│  50% - 75%       │  40% discount on this tier  │
│  75% - 100%      │  60% discount on this tier  │
├────────────────────────────────────────────────┤
│  Full month       │  ~30% net discount          │
└────────────────────────────────────────────────┘

Note: SUDs apply to N1 and N2 machine types.
E2 and Tau machines do NOT get SUDs.

Committed Use Discounts (CUDs)

CUDs provide deeper discounts in exchange for 1-year or 3-year commitments.

Resource-based CUDs — Commit to specific vCPU and memory amounts:

# Create a 1-year resource-based CUD
$ gcloud compute commitments create my-cud-1year \
    --region=us-central1 \
    --resources=vcpu=16,memory=64GB \
    --plan=12-month \
    --type=COMPUTE_OPTIMIZED
Created [my-cud-1year].

# List existing commitments
$ gcloud compute commitments list --region=us-central1
NAME            REGION         STATUS   START_TIME            END_TIME              PLAN
my-cud-1year    us-central1    ACTIVE   2026-03-30T00:00:00   2027-03-30T00:00:00   12-month

Spend-based CUDs — Commit to a minimum spend per hour on specific services:

CUD Type	1-Year Discount	3-Year Discount	Applicable To
Resource-based	~37%	~55%	Compute Engine (vCPU, memory)
Spend-based	~25%	~52%	Cloud SQL, Cloud Run, GKE

CUD vs SUD Comparison

Feature	SUD	CUD (Resource)	CUD (Spend)
Commitment	None	1 or 3 years	1 or 3 years
Max discount	~30%	~55%	~52%
Auto-applied	Yes	Yes (after purchase)	Yes (after purchase)
Cancellable	N/A	No	No
Flexibility	Any VM	Same region/type	Same service
Scope	Per project	Per region	Per region

Spot VMs (Preemptible)

Spot VMs are up to 60-91% cheaper but can be reclaimed by GCP with 30 seconds notice. Ideal for fault-tolerant, batch, and CI/CD workloads.

# Create a Spot VM
$ gcloud compute instances create batch-worker-1 \
    --zone=us-central1-a \
    --machine-type=e2-standard-4 \
    --provisioning-model=SPOT \
    --instance-termination-action=STOP \
    --image-family=debian-12 \
    --image-project=debian-cloud
Created [batch-worker-1].

# Create a Spot VM MIG for batch processing
$ gcloud compute instance-templates create spot-template \
    --machine-type=e2-standard-4 \
    --provisioning-model=SPOT \
    --instance-termination-action=DELETE \
    --image-family=debian-12 \
    --image-project=debian-cloud
Created [spot-template].

$ gcloud compute instance-groups managed create batch-pool \
    --zone=us-central1-a \
    --template=spot-template \
    --size=5
Created [batch-pool].

Spot VM Use Cases

Use Case	Why Spot Works	Savings
CI/CD build agents	Short-lived, restartable	60-80%
Batch data processing	Fault-tolerant, checkpointable	70-91%
Dev/test environments	Non-critical, can restart	60-80%
Machine learning training	Checkpoint + resume	60-91%
Rendering / transcoding	Parallelizable, restartable	70-91%

Custom Machine Types

Instead of over-provisioning with predefined sizes, create VMs with exactly the CPU and memory you need:

# Create a custom machine type (4 vCPUs, 10 GB RAM)
$ gcloud compute instances create custom-vm \
    --zone=us-central1-a \
    --custom-cpu=4 \
    --custom-memory=10GB \
    --image-family=debian-12 \
    --image-project=debian-cloud
Created [custom-vm].

# Extended memory (more than 6.5 GB per vCPU)
$ gcloud compute instances create high-mem-vm \
    --zone=us-central1-a \
    --custom-cpu=2 \
    --custom-memory=24GB \
    --custom-extensions \
    --image-family=debian-12 \
    --image-project=debian-cloud
Created [high-mem-vm].

Cost Comparison: Standard vs Custom vs Spot

Configuration	Monthly Cost	Savings
e2-standard-4 (4 vCPU, 16 GB)	~$97/month	Baseline
Custom (4 vCPU, 10 GB)	~$78/month	~20%
e2-standard-4 + CUD 1yr	~$61/month	~37%
e2-standard-4 + CUD 3yr	~$44/month	~55%
Spot e2-standard-4	~$29/month	~70%

BigQuery Pricing Optimization

# Switch from on-demand to flat-rate for predictable BQ workloads
# On-demand: $6.25/TB scanned
# Flat-rate editions: Predictable monthly cost

# Check query costs before running
$ bq query --dry_run \
    'SELECT * FROM my_dataset.large_table WHERE date > "2026-01-01"'
Query successfully validated. Estimated bytes processed: 5368709120

# That's ~5 GB = ~$0.03 on-demand
# For heavy BQ usage, consider BigQuery editions

Cloud Storage Cost Optimization

Storage Class	Use Case	Monthly Cost/GB	Retrieval Cost
Standard	Frequent access	$0.020	Free
Nearline	Monthly access	$0.010	$0.01/GB
Coldline	Quarterly access	$0.004	$0.02/GB
Archive	Yearly access	$0.0012	$0.05/GB

# Set lifecycle rules to move old data to cheaper classes
$ cat lifecycle.json
{
  "lifecycle": {
    "rule": [
      {
        "action": {"type": "SetStorageClass", "storageClass": "NEARLINE"},
        "condition": {"age": 30}
      },
      {
        "action": {"type": "SetStorageClass", "storageClass": "COLDLINE"},
        "condition": {"age": 90}
      },
      {
        "action": {"type": "Delete"},
        "condition": {"age": 365}
      }
    ]
  }
}

$ gcloud storage buckets update gs://my-data-bucket \
    --lifecycle-file=lifecycle.json
Updating gs://my-data-bucket/...

Common Mistakes

Mistake	Impact	Fix
Over-committing with CUDs	Paying for unused capacity	Start small, expand based on usage
Using Spot for stateful workloads	Data loss on preemption	Use Spot only for stateless/batch
Ignoring sustained use discounts	Missing automatic savings	Consolidate usage on eligible types
Standard storage for cold data	16x cost vs Archive class	Apply lifecycle policies

Quick Reference

Task	Command
Create CUD	`gcloud compute commitments create NAME --resources=vcpu=N,memory=MGB --plan=12-month`
Create Spot VM	`gcloud compute instances create NAME --provisioning-model=SPOT`
Custom machine type	`gcloud compute instances create NAME --custom-cpu=N --custom-memory=MGB`
List commitments	`gcloud compute commitments list --region=REGION`
Set storage lifecycle	`gcloud storage buckets update gs://BUCKET --lifecycle-file=FILE`

Next → Cost Analysis & Automation

Share on

Twitter Facebook LinkedIn