AWS
Microsoft Azure
Google Cloud
Key takeaways
1
Resource-based CUDs suit stable, matching resources in one region. A GCP Flex CUD follows eligible spend across supported VM families, regions and services.
2
Check SKUs and terms. Flex GPU coverage is limited to eligible G2/G4 GPUs; M1–M4 CPU and memory receive no one-year Flex discount.
3
Compare commitment fees plus uncovered usage after planned changes. Unused commitments remain payable, so base purchases on demand you expect to sustain.
Choosing a GCP Flex CUD or a resource-based commitment starts with understanding what your fleet will need over time. Some workloads may stay in the same region and machine series; others may move between families or services. Those changes affect how much of your commitment you can use and whether the discount delivers the savings you expected.
We compare eligibility, flexibility and discount application, then follow a stable VM fleet and a migration to eligible Cloud Run usage. The examples show how commitment fees and uncovered usage affect total cost, helping you choose coverage that fits your expected demand.
Google’s resource-based commitment rules tie coverage to resource quantities, commitment type and region. A GCP Flex CUD instead commits you to hourly spend across an eligible pool, valued at discounted prices.
We compare eligibility, flexibility and discount application, then follow a stable VM fleet and a migration to eligible Cloud Run usage. The examples show how commitment fees and uncovered usage affect total cost, helping you choose coverage that fits your expected demand.
How a GCP Flex CUD Compares With Resource-Based Commitments
Compute Flexible committed use discounts and resource-based CUDs exchange a financial commitment for lower eligible usage prices. The difference is what must keep matching that commitment.Google’s resource-based commitment rules tie coverage to resource quantities, commitment type and region. A GCP Flex CUD instead commits you to hourly spend across an eligible pool, valued at discounted prices.
| Decision factor | Resource-based CUD | Compute Flexible CUD |
|---|---|---|
| Commitment unit | Matching vCPU, memory or other eligible resource quantities | Discounted dollars per hour |
| Location | Committed region | Eligible regions within one billing account |
| Projects | Purchasing project or shared eligible projects | Eligible projects paid by that billing account |
| Workload changes | Resources must still match the commitment type | Spend can move among supported resources and services |
| Standard plans | One or three years | One or three years |
| Unused commitment | Still payable | Still payable; no hourly rollover |
Resource-based coverage does not require an unchanged VM size. Matching resources can be consumed across eligible VM configurations. Hardware and applicable OS-license commitments are separate purchases; supported resource-based term extensions can also outlast the original plan.
Eligible memory-optimized hardware discounts can reach 70%, but compare actual SKU prices rather than assuming every resource receives that discount.
Check your resource-based sharing configuration before assuming project-only coverage. Billing accounts created from June 16, 2026 default to sharing. Older accounts without active resource-based commitments on that date were switched; those with active commitments retained their previous scope.
For the broader distinction between resource-based, flexible and service-specific products, our GCP committed use discounts guide provides the surrounding context.
Google’s current eligibility and discount table distinguishes categories that a blanket 28% or 46% claim would miss. Eligibility and published rates were checked against this guidance and Google’s resource-based CUD documentation on October 21, 2026. Compute Engine resource coverage
Eligible memory-optimized hardware discounts can reach 70%, but compare actual SKU prices rather than assuming every resource receives that discount.
Check your resource-based sharing configuration before assuming project-only coverage. Billing accounts created from June 16, 2026 default to sharing. Older accounts without active resource-based commitments on that date were switched; those with active commitments retained their previous scope.
For the broader distinction between resource-based, flexible and service-specific products, our GCP committed use discounts guide provides the surrounding context.
Which Resources and Services Qualify for Each CUD
Before choosing a GCP Flex CUD, identify eligible bill items, forecast the hourly demand remaining after existing commitment coverage, and model commitment fees plus uncovered usage charges through planned workload changes.Google’s current eligibility and discount table distinguishes categories that a blanket 28% or 46% claim would miss. Eligibility and published rates were checked against this guidance and Google’s resource-based CUD documentation on October 21, 2026. Compute Engine resource coverage
| Resource or service | Resource-based treatment | Flex discount for one year / three years |
|---|---|---|
| Listed general-purpose series and C2/C2D CPU and memory | Matching hardware commitment; N1 shared-core excluded | 28% / 46% |
| H3/H4D CPU and memory | Matching hardware commitment | 17% / 38% |
| M1–M4 CPU and memory | Matching memory-optimized commitment type | No discount / 63% |
| C4N/M4N CPU and memory | Matching hardware commitment type | 28% / 54% |
| G2 CPU, memory and NVIDIA L4 GPU | Matching hardware commitment | 21% / 43% |
| G4 CPU, memory and NVIDIA RTX PRO 6000 GPU | Matching hardware commitment | 16% / 42% |
| Other supported GPU devices | Device-specific resource-based rules | No Flex GPU coverage |
| Tau T2D and X4 resources | Matching commitment type; T2D standard types only | Not on the current Flex eligibility list |
| Z3/Z4D/Z4M supported resources | Matching hardware commitment | Eligible; check SKU prices |
| Supported Local SSD resources | Series-specific commitment rules | 28% / 46% |
| Eligible sole-tenancy premium | Not covered by resource-based CUDs | 28% / 46% |
GKE and Cloud Run coverage
| Resource or service | Resource-based treatment | Flex discount for one year / three years |
|---|---|---|
| GKE Standard / Autopilot eligible usage | Standard node resources can use Compute Engine commitments | Autopilot eligible SKUs 28% / 46%; Standard nodes use applicable Compute Engine rates |
| Cloud Run instance-based services, jobs and worker pools | No Compute Engine resource-based coverage | 28% / 46% |
| Cloud Run request-based services and functions | No Compute Engine resource-based coverage | 17% / 17% |
The listed general-purpose series are C3, C3D, C4, C4A, C4D, E2, N1, N2, N2D, N4, N4D and N4A. Verify the exact billed resource in Google’s Compute Flexible eligible-SKU register, rather than assuming every charge from a service qualifies. Published discounts can differ with other account pricing arrangements.
For a one-year GCP Flex CUD, M1–M4 spend can consume otherwise-unused commitment without receiving a discount. M4N is different: its specific machine-family documentation explicitly lists 28% for one year and 54% for three years.
G2/G4 eligibility does not extend to every GPU. Cloud Run GPU charges are not eligible for Compute Flexible CUDs. Google’s hardware commitment matrix identifies resources available under resource-based commitments. Resource-based GPU commitments require attached reservations; Flex does not. A reservation separately provides capacity assurance.
Google’s GKE exclusions include Spot Pods, cluster management fees and Autopilot Scale-Out ComputeClass. Spot and preemptible VMs are outside Compute Flex coverage. Ordinary disk and network charges are not covered simply because the VM qualifies.
BigQuery and Cloud SQL use separate commitment products. Our BigQuery CUD guide explains the analytics-specific choice.
The fleet contains ten c4n-standard-4 VMs in Iowa (us-central1). Google’s C4N pricing table lists these hourly prices per VM:
For a one-year GCP Flex CUD, M1–M4 spend can consume otherwise-unused commitment without receiving a discount. M4N is different: its specific machine-family documentation explicitly lists 28% for one year and 54% for three years.
G2/G4 eligibility does not extend to every GPU. Cloud Run GPU charges are not eligible for Compute Flexible CUDs. Google’s hardware commitment matrix identifies resources available under resource-based commitments. Resource-based GPU commitments require attached reservations; Flex does not. A reservation separately provides capacity assurance.
Google’s GKE exclusions include Spot Pods, cluster management fees and Autopilot Scale-Out ComputeClass. Spot and preemptible VMs are outside Compute Flex coverage. Ordinary disk and network charges are not covered simply because the VM qualifies.
BigQuery and Cloud SQL use separate commitment products. Our BigQuery CUD guide explains the analytics-specific choice.
Compare a Stable Fleet With an Eligible Migration
Compare a flexible commitment with a resource-based purchase covering the same starting fleet. These are synthetic estimates using published USD prices checked October 21, 2026, not measurements or customer results.The fleet contains ten c4n-standard-4 VMs in Iowa (us-central1). Google’s C4N pricing table lists these hourly prices per VM:
| Pricing basis | Hourly price |
|---|---|
| On-demand | $0.316275 |
| One-year resource-based | $0.199252 |
| One-year Compute Flexible | $0.227718 |
Assume 8,760 hours, constant prices and no existing commitments. Exclude negotiated discounts, free-tier offsets, promotional credits, disks, networking, licenses, taxes, vendor fees and Cashback.
Portfolio A keeps the fleet unchanged
Both commitments remain fully utilized. Annual cost equals ten VMs × hourly price × 8,760 hours.| Full-year cost | No commitment | Resource-based | Compute Flexible |
|---|---|---|---|
| Stable fleet | $27,705.69 | $17,454.48 | $19,948.10 |
Resource-based saves approximately 37%; Flex saves 28%. With these assumptions, portability adds cost without improving utilization.
Portfolio B moves half the spend to Cloud Run
After 4,380 hours, five VMs are replaced by eligible Cloud Run instance-based CPU and memory usage worth $1.581375/hour at on-demand prices. This assumes equal spend, not equivalent performance or migration cost. Google’s Cloud Run commitment rates give that eligible usage the same 28% one-year discount used here.| After migration | Resource-based | Compute Flexible |
|---|---|---|
| Commitment utilization | 50% | 100% |
| Hourly commitment fee | $1.992520 | $2.277180 |
| Uncovered usage per hour | $1.581375 | $0 |
| Total hourly cost | $3.573895 | $2.277180 |
| Full-year cost including both halves | $24,380.90 | $19,948.10 |
The resource-based fee continues while Cloud Run bills separately. The GCP Flex CUD covers both eligible spend pools. Full-year savings are approximately 12% and 28%, respectively, against the same $27,705.69 baseline.
For overlapping spend-based coverage, service-specific commitments take precedence. Older Flex commitments are utilized before newer ones; within a commitment, higher-discount usage comes first, followed by proportional allocation where applicable. Size a GCP Flex CUD against what remains after existing coverage. Google’s sustained use discount (SUD) rules prevent SUDs from applying to CUD-covered usage. They do not give every uncovered charge an automatic discount.
Direct-discount billing also retains fee-utilization offsets in exports. Account for those offsets when reconciling discounted usage and fees so covered costs are not counted twice.
Use this comparison over the same period:
How Resource-Based and Flex CUDs Apply Together
You can combine instruments across different portions of usage. Google’s commitment application rules apply resource-based coverage first, then flexible coverage to remaining eligible usage. The same portion of usage cannot receive both discounts.For overlapping spend-based coverage, service-specific commitments take precedence. Older Flex commitments are utilized before newer ones; within a commitment, higher-discount usage comes first, followed by proportional allocation where applicable. Size a GCP Flex CUD against what remains after existing coverage. Google’s sustained use discount (SUD) rules prevent SUDs from applying to CUD-covered usage. They do not give every uncovered charge an automatic discount.
Direct-discount billing also retains fee-utilization offsets in exports. Account for those offsets when reconciling discounted usage and fees so covered costs are not counted twice.
How to Size Commitments and Compare Effective Savings
Size a GCP Flex CUD from eligible hourly spend valued at its applicable discounted prices. Size resource-based purchases from matching resource quantities. Monthly averages can hide quiet hours that leave either commitment unused.Use this comparison over the same period:
Effective savings = 1 − (commitment fees + uncovered usage charges) ÷ comparable no-commitment cost
Keep the baseline consistent. Google’s Flex savings examples demonstrate why commitment amounts and on-demand-equivalent usage are different. For an actual purchase decision, also compare against existing discounts and coverage, not only published list prices.
Now change the migration example: five VMs disappear with no replacement spend. Hourly demand falls to $1.581375, but fees stay $1.992520 or $2.277180. Both purchases then cost more than the remaining usage would on demand. Portability cannot replace vanished demand.
Before purchasing, check overnight and weekend usage, planned migrations, shutdowns and commitment expirations. Use Google’s CUD recommendations as a starting point, then test expected changes. Buy less or defer when expected savings do not justify the cost of unused commitments.
Monitor utilization and realized savings separately. Flex utilization measures covered discounted spend against the fee; resource-based utilization measures matching consumed quantities against committed quantities.
Our guide to when automated GCP commitment management makes sense helps assess the recurring work after you decide which commitments fit.
We offer a read-only Savings Test option before purchasing is authorized. Once enabled, we manage commitments in your account through Autopilot or approval-based CoPilot, adjusting future purchases without resizing or modifying production workloads.
With our Flex Insured Commitments, teams can access a 46% discount on qualifying general-purpose Compute Engine CPU and memory usage through a three-year Compute Flexible CUD with none of the commitment risk.
If an eligible Flex Commitment costs more than equivalent pay-as-you-go usage, we provide Cashback protection for the qualifying difference. Our protection does not cancel the underlying commitment, and we do not automatically protect existing customer-owned commitments as Flex Commitments.
We charge an agreed percentage of realized savings, so the fee applies when verified savings are generated.
Once you have mapped eligible usage and expected migrations, a billing-based evaluation can help assess whether additional commitment coverage fits your fleet.
Now change the migration example: five VMs disappear with no replacement spend. Hourly demand falls to $1.581375, but fees stay $1.992520 or $2.277180. Both purchases then cost more than the remaining usage would on demand. Portability cannot replace vanished demand.
Before purchasing, check overnight and weekend usage, planned migrations, shutdowns and commitment expirations. Use Google’s CUD recommendations as a starting point, then test expected changes. Buy less or defer when expected savings do not justify the cost of unused commitments.
Monitor utilization and realized savings separately. Flex utilization measures covered discounted spend against the fee; resource-based utilization measures matching consumed quantities against committed quantities.
Our guide to when automated GCP commitment management makes sense helps assess the recurring work after you decide which commitments fit.
How We Help Evaluate GCP Commitments
Choosing the commitment is one decision; keeping future purchases aligned with changing demand is ongoing work. Our GCP commitment optimization approach helps teams evaluate existing coverage and the mix of resource-based and spend-based CUDs.We offer a read-only Savings Test option before purchasing is authorized. Once enabled, we manage commitments in your account through Autopilot or approval-based CoPilot, adjusting future purchases without resizing or modifying production workloads.
With our Flex Insured Commitments, teams can access a 46% discount on qualifying general-purpose Compute Engine CPU and memory usage through a three-year Compute Flexible CUD with none of the commitment risk.
If an eligible Flex Commitment costs more than equivalent pay-as-you-go usage, we provide Cashback protection for the qualifying difference. Our protection does not cancel the underlying commitment, and we do not automatically protect existing customer-owned commitments as Flex Commitments.
We charge an agreed percentage of realized savings, so the fee applies when verified savings are generated.
Choose Commitments for Your Expected Workload Changes
Choose resource-based coverage where matching resources are likely to persist and the applicable rate justifies the obligation. Choose a GCP Flex CUD where eligible spend is steadier than its underlying resource mix. Combine them only after accounting for allocation order. If the baseline itself is uncertain, a smaller purchase or waiting can be the stronger financial choice.Once you have mapped eligible usage and expected migrations, a billing-based evaluation can help assess whether additional commitment coverage fits your fleet.
Normally, no. Stopping workloads or disconnecting a management platform does not cancel the provider obligation. Google’s resource-based support guidance allows contacting Billing Support within 14 calendar days of the commitment’s start date for purchase mistakes, but approval of changes is discretionary.
Frequently asked questions
Can you cancel a GCP commitment after purchase?
Normally, no. Stopping workloads or disconnecting a management platform does not cancel the provider obligation. Google’s resource-based support guidance allows contacting Billing Support within 14 calendar days of the commitment’s start date for purchase mistakes, but approval of changes is discretionary.
Can resource-based CUDs cover a different VM size?
Yes, when the resources still match the commitment type, region and sharing scope. A change in VM size alone does not require a flexible commitment. Check CPU and memory quantities separately after resizing.
Can a GCP Flex CUD work with capacity reservations?
Yes, eligible usage can consume standalone reservations and receive Flex coverage. Flex does not support attached reservations. Resource-based GPU commitments require attached reservations; the discount and capacity mechanisms must be evaluated separately.
Will a GCP Flex CUD cover a move to BigQuery or Cloud SQL?
No. Those services have separate commitment products. A migration can preserve total cloud spend while removing spend eligible for Compute Flex, leaving the original commitment underutilized.
Share
Facebook
X
LinkedIn
Reddit
