Committed Use Discounts (CUDs) on Google Cloud, Explained
Google Cloud's version of commitment pricing comes in two shapes. Both reward steady usage, and resource-based CUDs bill whether you use them or not.
A committed use discount (CUD) is the discount Google Cloud gives you in exchange for a one-year or three-year commitment. In Google’s words, “you commit to either using a minimum level of resources or spending a minimum amount, for a term duration of one or three years.” Those two options are the two families of CUD: resource-based and spend-based. Choosing between them is a choice between committing to hardware and committing to dollars.
Resource-based CUDs: commit to vCPUs and memory
Resource-based CUDs are available only for Compute Engine. You commit to a minimum amount of specific resources, such as vCPUs, memory, GPUs, Local SSD disks or sole-tenant nodes, in a specific region. Premium operating system licenses have their own separate commitments; one commitment cannot cover both hardware and licenses.
Google publishes the discount ceilings for hardware commitments: up to 70% for memory-optimized machine series and up to 55% for all other machine series.
The billing rule is the part to remember. Google states: “You are billed monthly for your committed resources until the end of your commitment term, regardless of whether or not you use those resources. You can’t cancel a commitment after its purchase.” A commitment becomes active at 12 AM Pacific Time on the day after purchase.
Spend-based CUDs: commit to an hourly amount
Spend-based CUDs discount eligible services in exchange for committing to a minimum spend, again for one or three years. With standard spend-based CUDs you get discounted rates until your hourly spend on eligible resources reaches your hourly committed amount; anything above that is charged at the on-demand rate. They apply across every project the Cloud Billing account pays for. Google sells them three ways:
- Compute flexible commitments: one commitment covering eligible spend across Compute Engine, GKE and Cloud Run, regardless of project or region.
- Service-specific commitments: bought per service (for example Cloud SQL or Memorystore) and applied only to that service.
- Flexible Savings Plans: cross-product spend commitments measured over a monthly entitlement window rather than an hourly one, aimed at variable or spiky usage.
If you know the AWS vocabulary, spend-based CUDs play the role of Savings Plans and resource-based CUDs the role of Reserved Instances.
How CUDs interact with sustained use discounts
Compute Engine also grants sustained use discounts (SUDs) automatically on eligible machine series used for more than 25% of a billing month, stepping up at the 50%, 75% and 100% marks. A full month earns up to 30% on N1, M1 and M2 machines (and GPUs attached to N1) and up to 20% on N2, N2D and C2. Two limits matter here: SUDs do not apply to usage already covered by CUDs, and only Self-serve (Online) Cloud Billing accounts receive them. So a CUD replaces, rather than stacks on, the SUD for the usage it covers. E2 and many newer machine series are not on the SUD eligibility list at all, so for them a commitment is the only discount mechanism of the two.
CUDs and scheduled workloads
Because a resource-based commitment bills monthly whether or not the resources run, it should cover only the vCPUs and memory you run around the clock. Stopping a VM that consumes committed capacity does not reduce the commitment charge; it just leaves committed capacity idle. The right order is:
- Put development, test and other part-time VMs on an off-hours schedule.
- Measure the vCPU and memory floor that remains running every hour in each region.
- Commit to that floor. Scheduled capacity above it keeps paying on-demand rates only while it runs.
Spend-based CUDs follow the same logic with dollars: size the hourly commitment to the lowest steady spend, not the business-hours average.
Where ZopNight fits
ZopNight’s commitment recommendations cover GCP CUDs alongside AWS and Azure reservations. They are advisory only, priced from Google’s published 1-year and 3-year CUD rates, based on the VM’s measured uptime (no measured uptime means the saving is shown as unknown), and suppressed when the projected run cost does not clear the commitment with a margin. ZopNight’s GCP billing view is net of credits, so committed-use and sustained-use discounts appear as the invoice bills them. For more GCP levers, see the GCP cost optimization guide.
Questions we get a lot.
If yours isn't here, email us and we'll answer directly.
Can I cancel a Compute Engine commitment?
No. Google states that you can't cancel a commitment after its purchase, and resource-based commitments are billed monthly until the end of the term whether or not you use the resources.
Do sustained use discounts stack with committed use discounts?
No. Google's documentation states that sustained use discounts don't apply to resource usage that is already covered by committed use discounts.