The Most Consistent Savings Opportunity in Cloud
If there is a single cloud optimization action that produces the most predictable, highest-ROI result for enterprise organizations, it is increasing Reserved Instance (RI) and Savings Plan coverage of steady-state compute workloads. The price gap between on-demand and reserved rates is the largest, most defensible discount available in cloud — and yet the median enterprise captures only 55-65% of the potential savings.
This article is part of our Cloud Pricing Benchmarks: AWS vs Azure vs GCP Complete Guide. Here we focus on the RI vs on-demand price gap across all three major cloud providers — what the discounts actually look like, how coverage rates vary by enterprise maturity, and how to bridge the gap between where most organizations are and where the top quartile operates.
Based on analysis of 285 enterprise cloud contracts: the average enterprise with $10M annual compute spend leaves $1.8-2.4M on the table annually through suboptimal RI/Savings Plan coverage. The full savings potential is there every single year — it simply requires systematic commitment management to capture it.
AWS: Reserved Instances and Savings Plans
AWS offers two main commitment mechanisms that deliver discounts versus on-demand: Reserved Instances and Savings Plans. Understanding both is essential for maximizing coverage.
AWS Reserved Instance Discount Benchmarks
| RI Type | Term | Payment | Discount vs On-Demand | Best For |
|---|---|---|---|---|
| Standard RI | 1 year | No Upfront | 30–40% | Predictable compute, flexibility needed |
| Standard RI | 1 year | All Upfront | 38–45% | Cash available, cost certainty needed |
| Standard RI | 3 year | No Upfront | 45–55% | Stable architecture, 3+ year horizon |
| Standard RI | 3 year | All Upfront | 55–65% | Maximum savings, highest confidence workloads |
| Convertible RI | 1 year | No Upfront | 20–30% | Workloads that may change instance family |
| Convertible RI | 3 year | No Upfront | 35–45% | Long commitment with flexibility to convert |
AWS Savings Plans: More Flexible, Similar Discounts
AWS Savings Plans were introduced as a more flexible alternative to Reserved Instances. They deliver comparable discounts but can apply to any EC2 instance type across regions (Compute Savings Plans) or within a specific instance family and region (EC2 Instance Savings Plans).
| Savings Plan Type | Term | Discount Range | Coverage |
|---|---|---|---|
| Compute Savings Plan | 1 year | 20–37% | Any EC2, Fargate, Lambda |
| Compute Savings Plan | 3 year | 33–54% | Any EC2, Fargate, Lambda |
| EC2 Instance Savings Plan | 1 year | 30–45% | Specific instance family, any size in region |
| EC2 Instance Savings Plan | 3 year | 50–66% | Specific instance family, any size in region |
Azure: Reserved VM Instances and Azure Savings Plan
Azure mirrors AWS's two-track approach with Azure Reserved VM Instances (fixed to instance series) and Azure Savings Plan for Compute (flexible across instance types). Azure also benefits from Azure Hybrid Benefit layering on top of RI discounts for Windows Server workloads.
Azure Reserved Instance Discount Benchmarks
| Reservation Type | Term | Discount vs On-Demand (Linux) | With Hybrid Benefit (Windows) |
|---|---|---|---|
| Azure Reserved VM Instance | 1 year | 35–45% | 55–65% |
| Azure Reserved VM Instance | 3 year | 50–60% | 65–75% |
| Azure Savings Plan for Compute | 1 year | 20–35% | 40–55% |
| Azure Savings Plan for Compute | 3 year | 35–50% | 55–68% |
The Azure Hybrid Benefit effect is substantial. For enterprises with Windows Server workloads and Software Assurance, Azure Reserved Instances with Hybrid Benefit can deliver 65-75% discounts versus on-demand — outpacing equivalent AWS discounts for the same workload type. This is the primary reason Azure often wins on economics for Microsoft-ecosystem enterprises.
Benchmark Your RI Coverage
We analyze your cloud compute spend and identify your RI/Savings Plan coverage gaps — quantifying the annual cost of under-reserving.
GCP: Committed Use Discounts vs On-Demand
GCP's commitment mechanism (Committed Use Discounts) works differently from AWS and Azure in an important way: CUDs commit to resource amounts (vCPU/memory) rather than to specific instance types. This provides more flexibility post-commitment but requires resource-level forecasting.
| GCP Commitment Type | Term | Discount vs On-Demand | Flexibility |
|---|---|---|---|
| Resource CUD (Compute) | 1 year | 28–35% | Any N1/N2 machine type in region |
| Resource CUD (Compute) | 3 year | 46–57% | Any N1/N2 machine type in region |
| Spend CUD (Cloud SQL) | 1 year | 20% | Any Cloud SQL in region |
| Spend CUD (Cloud Run) | 1 year | 17% | Cloud Run minutes in region |
| Sustained Use Discount (automatic) | No commitment | Up to 30% on 100%-running instances | Automatic, no commitment required |
Note: GCP's Sustained Use Discounts (SUDs) provide automatic discounts even without any commitment — up to 30% for instances running 100% of a month. This means GCP's effective baseline rate for always-on workloads is already discounted relative to AWS's on-demand pricing before any CUDs are applied. The CUD discount stacks on top of SUD: a 3-year CUD provides 46-57% off on-demand, but the on-demand rate it's discounted against is already lower than AWS's equivalent due to SUD.
Enterprise RI Coverage Benchmarks: Where Organizations Stand
Knowing the discounts available is only part of the picture. The other critical variable is coverage rate — what percentage of eligible compute spend is actually covered by reservations or commitment programs.
| Enterprise Maturity Level | AWS RI/SP Coverage | Azure Reservation Coverage | GCP CUD Coverage |
|---|---|---|---|
| Unoptimized (bottom quartile) | 25–45% | 20–40% | 20–40% |
| Median enterprise | 55–65% | 50–62% | 50–62% |
| Top quartile | 80–90% | 78–88% | 75–87% |
| Best-in-class | 88–95% | 85–93% | 82–90% |
The implication: bridging from median (60%) to top-quartile (85%) coverage on a $10M compute bill saves approximately $2.2M annually. This calculation assumes a blended RI discount of 40% (conservative) and applies the incremental savings from the additional 25% of compute moved from on-demand to reserved pricing.
The True Cost of Under-Reserving: A Worked Example
To make this concrete, let's model the cost of under-reserving for a representative enterprise.
Scenario: Company with $15M annual AWS compute spend
- Current RI/Savings Plan coverage: 58% of compute
- On-demand portion: 42% × $15M = $6.3M at full on-demand rates
- Steady-state workloads within that on-demand portion (reservable): ~70% of $6.3M = $4.41M
- Available discount if reserved (1-year Compute SP): 35% average
- Annual savings opportunity: $4.41M × 35% = $1.54M/year
That $1.54M is available every year. It doesn't require contract negotiation with AWS. It doesn't require a new vendor. It requires purchasing Savings Plans for workloads that are already running, already predictable, already steady-state.
Why don't organizations capture it? Our interviews with FinOps teams identify three recurring obstacles:
- Perceived risk of over-commitment. Teams are concerned about committing to capacity that might not be needed. The solution: commit at 80-85% of current steady-state, not 100%. This leaves buffer for reduction.
- Lack of visibility into steady-state vs variable workloads. Without proper tagging and cost allocation, it's difficult to know which workloads are truly steady-state. Investment in cloud cost management tooling solves this.
- Organizational disconnect between FinOps and engineering. Engineers provision on-demand; FinOps doesn't have authority to purchase commitments. This governance gap costs millions annually.
Quantify Your RI Coverage Gap
Submit your cloud billing data and we'll calculate your current coverage rate, the steady-state reservable compute, and the exact savings available from closing the gap.
RI and Savings Plan Strategy: Best Practices
Start With Savings Plans, Not Reserved Instances
For AWS, start new reservation programs with Compute Savings Plans rather than Standard Reserved Instances. Compute SPs cover all EC2 instance types and sizes in any region, eliminating the risk of stranded RIs when workloads resize or migrate. The discount is slightly lower than Standard RIs for identical commitments, but the operational simplicity justifies it for most enterprises.
Tier Your Commitment Strategy by Confidence Level
Not all workloads have equal predictability. A tiered approach:
- Tier 1 — Highest confidence (3-year commitments): Core infrastructure that will exist for 3+ years: production databases, authentication services, core application servers. Target 3-year All Upfront RI or 3-year Savings Plans.
- Tier 2 — High confidence (1-year commitments): Application workloads that have been stable for 12+ months. Target 1-year No Upfront RI or 1-year Savings Plans.
- Tier 3 — Variable (Spot + On-Demand): Batch processing, variable traffic workloads, development environments. Use Spot for interruption-tolerant workloads, on-demand for everything else.
Automate RI Management
Manual RI/Savings Plan management at enterprise scale is not sustainable. Tools like AWS Cost Explorer, Azure Advisor, and GCP Cost Intelligence generate automated recommendations for RI purchases. Third-party tools (Spot.io, Apptio, CloudHealth) provide portfolio-level optimization across all three clouds simultaneously.
Benchmark: enterprises using automated RI recommendation tools achieve 8-12% higher coverage rates than those managing reservations manually, per our data. The investment in tooling pays back within 1-2 months.
Conclusion: The Reservation Action Plan
Reserved Instance and Savings Plan optimization is the foundational layer of cloud cost management. No other single action produces more predictable, more sustained savings. Your action plan:
- Audit current coverage rates. Pull RI/SP coverage for AWS compute, Azure reservation coverage, and GCP CUD coverage. Compare to the benchmarks above.
- Identify the coverage gap value. Calculate the on-demand spend that could be reserved, multiply by the applicable discount rate, and quantify the annual savings opportunity.
- Segment workloads by commitment confidence. Identify Tier 1 (3-year), Tier 2 (1-year), and Tier 3 (spot/on-demand) workloads.
- Purchase Savings Plans for immediate coverage expansion. AWS Compute Savings Plans are the fastest path to coverage improvement with lowest operational risk.
- Set a coverage target. 85% of eligible compute covered by reservations is achievable for most enterprises within 90 days. Make it a team objective with clear ownership.
The companies that consistently optimize cloud costs don't have better tools or better vendors. They have better discipline about reserving what they know will run — and they validate that discipline against benchmark data to know when they're falling short. Start benchmarking your coverage today.