Why Commitment Discount Structures Matter More Than List Prices
All three major cloud hyperscalers offer significant discounts in exchange for usage commitments — but the mechanics of those commitments, and the flexibility provisions that govern them, differ materially. Choosing the wrong commitment structure for your workload profile can lock you into a model that generates stranded credits rather than real savings. This guide covers each model in technical and commercial detail. For organisations already working with our Google Cloud advisory team, this analysis will help you model the optimal commitment blend across GCP, AWS, and Azure.
Google Cloud Committed Use Discounts (CUDs)
Google Cloud CUDs are available in one-year and three-year terms, covering Compute Engine, Cloud SQL, Cloud Spanner, and a growing range of services. Resource-based CUDs commit to specific machine types (vCPU and memory) and offer discounts of approximately 37% (1-year) and 55% (3-year) off on-demand pricing. Spend-based CUDs commit to a minimum dollar spend across eligible services and offer around 25% (1-year) and 52% (3-year) discounts.
Google's CUD model has one distinctive advantage: resource-based CUDs are automatically applied across the committed resource type regardless of which project or region consumes them (within the same billing account). This means unutilised CUD capacity in one project can be consumed by another — reducing stranded commitment risk compared to AWS and Azure equivalents. For organisations managing complex multi-project GCP estates, this cross-project applicability is a significant operational advantage.
The trade-off is reduced flexibility compared to AWS Savings Plans. CUDs require more granular planning of your resource inventory upfront, but they reward that planning with materialised savings across fragmented projects and regions.
AWS Reserved Instances and Savings Plans
AWS offers two overlapping commitment models. Reserved Instances (RIs) commit to specific EC2 instance families in specific regions, offering up to 72% discount (3-year, all-upfront) over on-demand. Savings Plans — introduced to address RI inflexibility — commit to a minimum hourly spend (compute or EC2-specific) and apply automatically across EC2, Fargate, and Lambda usage. Compute Savings Plans offer around 66% maximum discount (3-year) with full flexibility across instance families, sizes, and regions.
The key limitation of AWS's model is that RIs and Savings Plans are account-scoped by default (though Organisation-level sharing is available). In multi-account AWS environments, commitment utilisation tracking requires deliberate governance architecture — something that is often absent in enterprises that have grown their AWS estate organically.
For organisations with centralised cloud financial management, AWS Savings Plans represent the most flexible commitment option across any hyperscaler, particularly if your workload mix is heterogeneous or rapidly evolving.
Azure Reserved VM Instances
Azure Reserved VM Instances (RIs) offer up to 72% savings over pay-as-you-go on a 3-year commit, with the option to pay monthly. Azure's scope flexibility is a strength: reservations can be scoped to a single subscription or shared across all subscriptions within an Enrolment — making cross-subscription utilisation straightforward for EA customers. Azure Hybrid Benefit can be stacked on top of Reserved Instances for organisations with existing Windows Server or SQL Server licences, creating combined discounts that are often overlooked.
The Azure RI model integrates tightly with the EA commitment structure, and organisations negotiating Microsoft Enterprise Agreements should model RI requirements as part of the EA negotiation — not as a separate post-EA decision. Our BYOL vs Azure licensing calculator covers the Hybrid Benefit stacking in detail.
Multi-Cloud Portfolio Optimisation: How to Combine All Three
Enterprises running workloads across all three hyperscalers — as the majority of Fortune 500 companies do — need a commitment portfolio strategy, not three siloed RI reviews. The core principle is to commit most aggressively on your most stable, predictable workloads and retain on-demand flexibility for variable or experimental workloads.
Across hyperscalers, this typically means:
- GCP CUDs for stable data and analytics workloads where cross-project applicability reduces utilisation risk
- AWS Compute Savings Plans for flexible, multi-region compute that may shift across instance families
- Azure Reserved Instances + Hybrid Benefit for predictable Windows/SQL Server workloads, stacked within your EA framework
For organisations evaluating where to run AI and machine learning workloads specifically, our guide on Vertex AI and Gemini enterprise pricing covers Google's AI commitment options. And for the Google Workspace vs Microsoft 365 software-layer decision that often accompanies cloud platform choices, see our GW vs M365 TCO comparison. To discuss your specific multi-cloud commitment portfolio with our team, book a confidential advisory call.
Building Your Commitment Portfolio
Running workloads on GCP, AWS, and Azure? Model the commitment portfolio that maximises discount realisation. Use our enterprise software assessment tools to model your cloud commitment discount exposure across all three hyperscalers.