What are Committed Use Discounts in GCP?
Committed Use Discounts (CUDs) are a pricing model offered by Google Cloud Platform (GCP) that provide a discount on Compute Engine resources in exchange for committing to a minimum level of usage over one or three years. In return for that commitment, Google applies a reduced rate to qualifying resources for the duration of the term.
Unlike AWS Savings Plans, the commitment is to a specific resource type rather than a spend amount. Unlike AWS Reserved Instances, there is no upfront payment option. CUDs are billed monthly regardless of the term length chosen.
Types of CUDs
GCP offers two types of CUDs:
- Resource-based CUDs commit to a specific amount of vCPUs, memory, GPUs, or local SSDs in a particular region. The discount applies to Compute Engine VM instances that match the committed resource type. These offer discounts of up to 57% compared to on-demand pricing for most machine types.
- Spend-based CUDs commit to a minimum hourly spend on specific GCP services, including Cloud Run, Google Kubernetes Engine, Cloud SQL, and certain Compute Engine machine families. These work more like AWS Savings Plans in that the commitment is to a dollar amount rather than a specific resource configuration. Discounts are typically around 17% for a one-year term and 32% for three years.
How they work
When you purchase a CUD, Google applies the discounted rate automatically to eligible resources in the committed region. You are billed for the committed amount whether or not you use it fully, so accuracy in estimating usage is important.
Here are a few things that are specific to how GCP handles commitments:
- CUDs are regional. A commitment made in us-central1 does not cover resources running in europe-west1.
- Resource-based CUDs can be shared across projects within the same billing account, which helps organizations with multiple GCP projects avoid over-committing in each one individually.
- CUDs stack with sustained use discounts, which GCP applies automatically to resources that run for a significant portion of the billing month. The two discounts apply to different portions of usage and do not cancel each other out.
GCP CUDs in action
An engineering team runs a production workload on GCP using n2-standard-8 instances across three projects in us-central1. Their monthly on-demand compute spend for this workload averages $4,200.
After reviewing 60 days of utilization data, they identify a stable baseline and purchase a one-year resource-based CUD for the equivalent vCPU and memory:
| On-demand | With 1-year CUD | |
|---|---|---|
| Monthly compute cost | $4,200 | ~$2,450 |
| Annual saving | ~$20,900 | |
| Commitment risk | None | Billed whether used or not |
The discount applies automatically across all three projects since CUDs are shared at the billing account level.
What to get right before committing
CUDs are a financial commitment that cannot be cancelled once purchased. Here are a few things worth confirming before you buy:
- Analyze at least 30 to 60 days of usage data to establish a reliable baseline. GCP workloads can vary more than expected, particularly those tied to business cycles.
- Choose the right CUD type for your workload. Resource-based CUDs suit stable, VM-heavy workloads. Spend-based CUDs are better for mixed or service-heavy usage patterns
- Commit only to your consistent baseline. Usage above the commitment is covered by on-demand or sustained use discount rates.
- Account for the regional scope. If workloads are spread across multiple GCP regions, commitments need to be purchased per region.
GCP also applies Sustained Use Discounts automatically to Compute Engine resources that run for more than 25% of a billing month without any commitment required. CUDs and sustained-use discounts can apply simultaneously to the same workload, stacking the savings. Understanding how the two interact helps avoid double-counting expected savings when evaluating whether a CUD is worth purchasing.