.png)
Google Cloud Committed Use Discounts reduce your compute costs by up to 70% in exchange for a one-year or three-year commitment. Understanding which of the two CUD types — resource-based or Flex — applies to your workload is the difference between maximizing the discount and paying for committed capacity that does not match your actual usage pattern. Three program changes in 2025-2026 have materially expanded what Flex CUDs cover and changed how discounts are billed.
This guide covers both CUD types with exact discount rates from Google Cloud official pricing documentation, the September 2025 expansion, the billing model migration, and the decision framework for which instrument to use per workload.
| Still Paying On-Demand? GCP Doesn’t Mind! |
Resource-Based CUDs: How They Work
Resource-based Committed Use Discounts provide a discount in exchange for your commitment to use a minimum amount of specific Compute Engine resources in a specific region. You commit to vCPUs, memory, GPUs, Local SSD storage, or sole-tenant nodes for a 1-year or 3-year term. Google Cloud bills you monthly for the committed resources at the discounted rate — whether or not you actually use those resources.
This is the central trade-off of resource-based CUDs: the discount is tied to specific resource types in a specific region. If your workload changes machine family, moves to a different region, or reduces below the committed resource level, you continue paying for the commitment at the discounted rate for the full term. The commitment does not reserve capacity in a specific zone — for capacity guarantees, you need a separate reservation.
Exact discount rates
For all machine types (general-purpose, compute-optimized, accelerator-optimized): 37% discount on a 1-year commitment, 55% discount on a 3-year commitment, compared to on-demand prices. For memory-optimized machine types: 70% discount on a 3-year commitment. Source: Google Cloud VM instance pricing documentation.
These are flat percentage discounts calculated from the on-demand price of the committed resources on the day the commitment becomes active. The commitment fee is the sum of the discounted prices of all committed resources. If on-demand prices change after your commitment becomes active, your commitment fee does not change.
What can be committed
Hardware commitments: vCPUs and memory for specific machine series, GPUs attached to supported machine types, Local SSD storage, and sole-tenant nodes. Software license commitments: applicable premium operating system licenses, purchased separately from hardware commitments. Custom machine types incur a 5% premium over the standard CUD prices — this premium applies to the portion and duration of your commitment where you run these custom machine type VMs.
Resources not eligible for resource-based CUDs: preemptible VM instances (Spot VMs), N1 shared-core machine types (f1-micro, g1-small), and extended memory.
Billing and term mechanics
After purchase, you are billed monthly for your commitment and must pay the monthly commitment fee even if you do not use all of your committed resources. Commitments can be purchased for up to 6-year terms for hardware resources. Auto-renewal is available: if enabled, a new commitment term automatically begins when the existing term ends.
You cannot cancel a resource-based CUD. Google Cloud documentation states explicitly: you must pay the agreed-upon monthly amount for the duration of the commitment. If you accidentally purchased a commitment or made a mistake configuring it, contact Cloud Billing Support. There is no self-service cancellation path. Size resource-based CUDs conservatively based on your stable minimum usage, not your average or maximum. Source: Google Cloud official documentation.
CUD sharing: the June 2026 default change
By default, resource-based CUDs apply to the project for which they were purchased. CUD sharing, when enabled, allows CUDs to be shared across all projects in the same Cloud Billing account — the discount applies to eligible usage anywhere in the billing account.
As of June 16, 2026: all new Cloud Billing accounts created on or after that date have CUD sharing enabled by default. For billing accounts created before June 16, 2026, CUD sharing must be manually enabled.
Also read: Multi-Cloud Savings Plan Strategy: AWS, Azure, and GCP Compared
Flex CUDs: How They Work
Compute Flexible Committed Use Discounts are spend-based commitments. Instead of committing to specific resources in a specific region, you commit to a minimum hourly spend amount across all eligible services and VM families. Google Cloud applies the discount automatically to the highest-discount eligible usage within your billing account, regardless of which VM family, region, or eligible service that usage comes from.
Exact discount rates
Flex CUDs offer a flat-rate discount: 28% off on a 1-year commitment, 46% off on a 3-year commitment. These discounts apply to all eligible usage up to the committed spend amount. Usage above the committed amount is charged at on-demand rates. Source: Google Cloud Blog announcing Flex CUDs (November 2022).
The September 2025 expansion
On September 5, 2025, Google Cloud expanded Flex CUD coverage significantly. Prior to this expansion, Flex CUDs covered most general-purpose VMs (N1, N2, E2, N2D) and compute-optimized VMs (C2, C2D). The expansion added:
- Memory-optimized VM families for memory-intensive workloads
- H3 and H4D high-performance computing VM families for scientific and HPC workloads
- Cloud Run request-based billing
- Cloud Functions serverless compute
This expansion means a single Flex CUD commitment can now cover a broader mix of VMs, containers, and serverless workloads simultaneously. Source: Google Cloud Blog, September 5, 2025.
The billing model migration
Flex CUDs originally operated on a credits-based model: usage was charged at the full list price, and CUD savings appeared as credit offsets. Google Cloud migrated Flex CUDs to a direct discount model in 2025-2026: the discounted rate is now charged directly for CUD-eligible usage rather than as a credit offset.
Timeline: new Cloud Billing accounts created on or after July 15, 2025 are automatically on the direct discount model. All existing customers were automatically migrated to the new model by January 21, 2026. Source: Google Cloud Blog, September 5, 2025.
The practical impact on cost reporting: billing data before the migration shows list price usage plus credit offsets. Billing data after the migration shows discounted rates directly. Any cost analysis that compares pre- and post-migration periods without normalizing for this change will produce misleading effective savings rate figures.
Cancellation impossibility
Like resource-based CUDs, Flex CUDs cannot be cancelled. You must pay the agreed monthly commitment amount for the full duration of the term. If you use less in a given hour than you committed to, you underutilize your commitment and do not realize your full discount for that hour. Overages above the committed amount are charged at on-demand rates.
The hourly check: Google Cloud checks your usage every hour to apply the discount. This means the stable floor of your hourly spend — not the average and not the peak — is the correct sizing basis for Flex CUD commitments.
Sustained Use Discounts: The Automatic Discount That Stacks
Sustained Use Discounts are automatic discounts that Google Cloud applies to eligible Compute Engine resources when they are used for a significant portion of the billing month. No commitment or purchase is required.
SUD-eligible resources: vCPUs and memory for N1, N2, N2D general-purpose machine types; C2 compute-optimized; and M1, M2 memory-optimized machine types. The maximum SUD percentage is either 20% or 30% depending on the resource and machine type. Source: Google Cloud Sustained Use Discounts documentation.
The stacking relationship: SUDs do not apply to resource usage already covered by CUDs. For resources on eligible machine families not covered by a CUD, SUDs apply automatically to the uncovered usage. This creates a natural layered discount structure: CUDs cover the committed baseline, and SUDs capture additional savings on consistent uncovered usage of eligible machine families.
CUDs Are Not Capacity Reservations
A common misunderstanding: purchasing a CUD does not reserve compute capacity. A resource-based CUD commitment for 100 N2 vCPUs in us-central1 guarantees a discounted price for those vCPUs — it does not guarantee those vCPUs will be available to launch new VMs during a surge or capacity-constrained event.
Google Cloud documentation states this explicitly: a commitment provides a discounted price agreement, but it does not reserve capacity in a specific zone. A reservation ensures that capacity is held in a specific zone even if the reserved VMs are not running. To get both discounted prices and capacity guarantees, you must both purchase a CUD commitment and create a reservation for those zonal resources.
For GPU and Local SSD resources specifically: when you commit to these resources, you must also create a reservation and attach it to your commitment.
For workloads where capacity availability is a hard operational requirement, CUDs alone are insufficient. Create a separate zonal reservation for the required capacity in addition to the CUD commitment. The CUD applies automatically to the reserved resources, giving you both the discount and the capacity guarantee. Source: Google Cloud official VM instance pricing documentation.
Also read: GCP Committed Use Discounts (CUDs): Complete Guide to Types, Pricing & Savings
Resource-Based vs Flex CUD: The Decision Framework
The correct choice depends on three factors: how predictable the specific machine family and region are, whether the workload spans multiple services or VM families, and how important discount depth is relative to commitment flexibility.
Choose resource-based CUDs when
The workload runs on a specific machine family in a specific region and that configuration is confirmed stable for the full commitment term. Core database servers, production API servers running the same instance type for 12 or more months, and batch processing fleets with confirmed machine family requirements all fit this profile.
Memory-optimized workloads (M1, M2, M3) in particular benefit from resource-based CUDs: the 70% 3-year discount is significantly higher than the Flex CUD 46% 3-year rate. For SAP HANA workloads, large in-memory databases, and memory-intensive analytics, resource-based 3-year CUDs deliver the highest available discount on any major cloud provider.
Still choosing CUDs with spreadsheets?See how Usage.ai recommends the right mix of Resource-Based and Flex CUDs—without the guesswork. |
Choose Flex CUDs when
The compute footprint spans multiple machine families, multiple regions, or a mix of VM and serverless compute. Flex CUDs follow eligible spend automatically — if a workload migrates from N2 to C3, shifts regions, or moves compute to Cloud Run, the Flex CUD continues applying to the new usage without requiring a new commitment or creating stranded spend.
The September 2025 expansion makes Flex CUDs the correct primary instrument for most mixed compute environments. An organization running production VMs alongside GKE workloads, Cloud Run services, and Cloud Functions can cover all eligible compute spend with a single Flex CUD.
The memory-optimized trade-off: the 46% Flex 3-year rate is 24 percentage points lower than the 70% resource-based 3-year rate for memory-optimized machines. For memory-optimized workloads with stable machine family and region, resource-based CUDs capture significantly more savings. For those with uncertain configurations, the Flex flexibility may justify the discount trade-off.
The layered approach
The highest-performing GCP commitment portfolios use both types simultaneously: resource-based CUDs on stable, confirmed machine family workloads to capture the deeper rate, and Flex CUDs on dynamic, mixed-service, or cross-region spend. SUDs then stack automatically on top for any eligible uncovered usage of eligible machine families.
Sizing CUD Commitments: The Floor Rule
Both CUD types punish over-commitment: you pay the full committed amount regardless of actual usage. The correct sizing basis is the stable minimum, not the average or peak.
For resource-based CUDs: identify the minimum number of vCPUs and memory consistently running in each machine family and region over the past 60-90 days, including overnight and weekends. Commit to that floor. Any capacity above that minimum fluctuates enough that a commitment creates over-commitment risk.
For Flex CUDs: identify the minimum hourly spend level across all hours in the past 60-90 days, including low-utilization periods. Commit to that floor. Google Cloud provides built-in CUD recommendations in the Cloud Billing console based on historical usage data — a good automated starting point available directly in the billing interface.
GCP CUDs for Cloud SQL, Memorystore, and Other Services
Beyond Compute Engine, Google Cloud offers spend-based CUDs for several additional services. These are separate commitments from the Compute Flex CUD.
Cloud SQL CUDs: spend-based commitments for Cloud SQL managed database instances covering compute costs in the committed region.
Memorystore CUDs: apply to Memorystore for Valkey, Redis Cluster, Redis, and Memcached usage. A Memorystore CUD provides the flexibility to use Valkey, Redis Cluster, Redis, or Memcached spending toward a single commitment on one billing account. Memorystore CUDs do not apply to Cloud Storage for backups, persistence, network egress, or Memorystore for Redis M1 capacity tier instances under 5 GB.
Region specificity: most spend-based CUDs are restricted to the region selected during purchase. Eligible usage in other regions does not count toward the commitment and is charged at on-demand rates. Multi-region deployments need separate commitments per region.
Billing account scope: spend-based CUDs are tied to the specific billing account used for purchase and cannot be shared across different billing accounts, even within the same organization.
Monitoring CUD Utilization
Google Cloud provides CUD analysis reports in the Cloud Billing console. These reports show utilization rates for each active commitment — the percentage of committed resources or committed spend actually consumed.
Target utilization for resource-based CUDs: 90% or above. Utilization consistently below 80% indicates over-sizing — you are paying for committed resources that are not running. The response is to evaluate whether planned workloads were deployed and to purchase more conservatively at the next commitment.
Target utilization for Flex CUDs: 90% or above at hourly granularity. Monthly averages mask overnight gaps. A monthly average of 85% can hide overnight hours where spend falls well below the committed floor, wasting commitment value during those periods.
Your CUDs might be committed. Your savings aren’t.Find uncovered spend, underutilized commitments, and your biggest GCP savings opportunities in minutes. |
How Usage.ai Optimizes GCP Committed Use Discounts
Usage.ai manages GCP CUD coverage alongside AWS Savings Plans, Reserved Instances, and Azure commitment instruments in a single multi-cloud platform.
Commitment sizing: Usage.ai analyzes GCP billing export data to identify the stable hourly spend floor per Flex CUD and the stable resource floor per machine family and region for resource-based CUDs. The 24-hour recommendation refresh cycle means sizing reflects current actual usage rather than stale snapshots.
Coverage identification: Usage.ai identifies eligible compute spend running at on-demand rates where CUD commitments would apply, surfaces each coverage gap, and quantifies the exact monthly saving at current rates.
The over-commitment problem solved with insurance: the inability to cancel GCP CUDs makes over-commitment the primary reason teams size conservatively and under-commit. Usage.ai Insured Flex Commitments include a buyback guarantee — if a committed workload is decommissioned or the usage pattern shifts below the committed level, the unused commitment is bought back and the value returned as cashback in real money. Teams can commit at the accurate level rather than a conservative hedge.
$91M+ in savings delivered to 300+ enterprise customers across AWS, Azure, and GCP. Fee: percentage of realized savings only. $0 if Usage.ai saves nothing. 30-minute setup, billing-layer access only.
Frequently Asked Questions
1. What are Google Cloud Committed Use Discounts?
CUDs provide discounted prices for eligible Google Cloud resources in exchange for 1-year or 3-year commitments. Two types: resource-based CUDs commit to specific Compute Engine resources (vCPUs, memory, GPUs, Local SSDs) in a specific region — 37% (1-year), 55% (3-year) for all machine types, 70% (3-year) for memory-optimized. Flex CUDs commit to a minimum hourly spend amount across eligible services and VM families — 28% (1-year), 46% (3-year). Source: Google Cloud VM instance pricing documentation.
2. Can you cancel a Google Cloud CUD?
No. Google Cloud documentation is explicit: you cannot cancel the commitments you have purchased. You must pay the agreed-upon monthly amount for the full duration of the commitment. This applies to both resource-based and Flex CUDs. If you accidentally purchased a commitment or made a mistake configuring it, contact Cloud Billing Support. Size commitments conservatively from your stable minimum usage floor to minimize the risk of paying for underutilized commitments.
3. What is the difference between resource-based and Flex CUDs?
Resource-based CUDs commit to specific Compute Engine resources in a specific region. Deeper discounts (up to 70% for memory-optimized 3-year). Best for stable, known machine family and region. Flex CUDs commit to a minimum hourly spend amount that automatically applies across eligible VM families, regions, Cloud Run, and Cloud Functions (since September 2025). Lower discounts (28%/46%) but coverage follows your compute wherever it runs. Source: Google Cloud official documentation.
4. What expanded in the September 2025 Flex CUD update?
On September 5, 2025, Google Cloud expanded Flex CUD coverage to include memory-optimized VM families, H3 and H4D HPC VM families, Cloud Run request-based billing, and Cloud Functions. Previously, Flex CUDs covered only general-purpose (N1, N2, E2, N2D) and compute-optimized (C2, C2D) VMs. Google also migrated the Flex CUD billing model from credits to direct discounts, with full migration completed by January 21, 2026. Source: Google Cloud Blog, September 5, 2025.
5. Do GCP CUDs reserve compute capacity?
No. A CUD provides a discounted price agreement only — it does not reserve capacity in a specific zone. For capacity guarantees alongside discounted pricing, you need both a CUD commitment and a separate zonal reservation. For GPU and Local SSD commitments, you must create and attach a reservation when purchasing those resource-based CUDs. Source: Google Cloud VM instance pricing documentation.
6. How do Sustained Use Discounts interact with CUDs?
SUDs do not apply to resource usage already covered by CUDs. For eligible machine families (N1, N2, N2D, C2, M1, M2) not covered by CUDs, SUDs apply automatically at up to 20-30% as resources are used for larger fractions of the billing month. This creates a layered discount structure: CUDs cover the stable committed baseline, and SUDs capture additional savings on consistent uncovered usage of eligible families without any additional commitment. Source: Google Cloud Sustained Use Discounts documentation.
