This article is part of our Cloud Contract Negotiation Guide — the complete resource for enterprise cloud contract strategy. For the full cluster including Azure, GCP, and EDPs, start there.
Why AWS Commitment Instruments Matter
Cloud cost optimisation has become a board-level priority for most enterprises. AWS compute costs — driven by EC2 instances, Lambda, Fargate, and similar services — represent 30–50% of total AWS spend for most organisations. On-demand pricing delivers maximum flexibility but costs significantly more than commitment-based pricing. For organisations with mature, predictable compute workloads, the choice between Savings Plans and Reserved Instances can represent $5M–$50M+ in cumulative savings or costs over a three-year period.
Both instruments are self-service offerings available directly through the AWS console — unlike Enterprise Discount Programmes (EDPs), which require direct AWS commercial negotiation. However, deploying either instrument carelessly can backfire: over-committing to workloads that then shrink, selecting the wrong plan type for evolving architectures, or failing to combine these instruments with private pricing agreements can transform an attractive discount into a financial liability.
AWS Savings Plans Explained
AWS Savings Plans are hourly dollar-spend commitments that apply broadly across multiple AWS services. Unlike Reserved Instances — which are tied to specific instance types, operating systems, and regions — Savings Plans work across families of services and provide meaningful flexibility. AWS offers two types of Savings Plans.
Compute Savings Plans
Compute Savings Plans apply your discount across EC2 instances, AWS Lambda, and Fargate. This broad applicability makes them ideal for organisations with diverse workload portfolios or those transitioning between compute delivery models. A Compute Savings Plan commits you to spending a specific dollar amount per hour (e.g., $10,000/hour) across all three services combined. If you spend $12,000/hour, you pay the $10,000 commitment rate plus $2,000 at on-demand prices. If you spend only $8,000/hour, you pay the full $10,000 commitment regardless.
Compute Savings Plans offer discounts up to 66% compared to on-demand pricing when you commit for three years with all-upfront payment. One-year commitments deliver 30–40% discounts. The all-upfront payment model offers better rates than partial upfront or no-upfront options, but requires sufficient budget certainty to commit cash at contract signature.
EC2 Instance Savings Plans
EC2 Instance Savings Plans apply specifically to EC2 instances and are locked to a specific instance family, operating system, and region. This narrower scope — compared to Compute Savings Plans — delivers higher discounts: up to 72% for three-year, all-upfront commitments compared to 66% for Compute plans. However, that additional 6 percentage points in discount comes with a constraint: you cannot shift your discount across instance families. A commitment to m6i instances in us-east-1 cannot be applied to m7i instances or to instances in us-west-2.
EC2 Instance Savings Plans are appropriate for organisations with stable, long-lived workload patterns — particularly those running dedicated database servers, long-running batch processes, or large-scale web application infrastructure where instance types do not change frequently.
Reserved Instances: When Higher Discounts Beat Flexibility
Reserved Instances (RIs) are AWS's original commitment instrument. Unlike Savings Plans, RIs are associated with specific instance types and come in two variants: Standard RIs and Convertible RIs.
Standard Reserved Instances
Standard RIs offer the deepest discounts: up to 72% off on-demand pricing for three-year, all-upfront commitments. They are inflexible — if you commit to m6i.2xlarge instances in us-east-1, your discount applies only to m6i.2xlarge instances in that region. You cannot convert to a different instance family, OS, or region without purchasing a new commitment and potentially absorbing a loss on your existing reservation.
The inflexibility of Standard RIs makes them valuable only when your workload pattern is genuinely fixed. Organisations that have run the same database server or application infrastructure for three years and expect to do so for the next three years can safely use Standard RIs. Organisations undergoing infrastructure transformation or cloud-first migrations should avoid Standard RIs.
Convertible Reserved Instances
Convertible RIs allow you to exchange your commitment for a different instance family, OS, or region — without losing your reservation. This flexibility makes them appropriate for evolving architectures. Convertible RIs offer discounts up to 54% for three-year commitments — 18 percentage points lower than Standard RIs, but meaningful nonetheless. The exchange mechanism is straightforward: if you can demonstrate that the new instance configuration has equal or greater upfront cost, AWS allows the conversion.
For organisations with multi-year cloud commitments but uncertain instance-level workload patterns, Convertible RIs represent a middle ground: you secure a material discount while retaining the ability to respond to architectural changes.
AWS Compute Discounts: Comparison Table
| Instrument | 1-Year Discount | 3-Year Discount | Flexibility | Best For |
|---|---|---|---|---|
| Compute Savings Plan (All-Upfront) | 30–40% | 50–66% | Highest (EC2, Lambda, Fargate) | Diverse compute portfolios |
| EC2 Instance Savings Plan (All-Upfront) | 35–45% | 55–72% | Moderate (same family, OS, region) | Stable EC2-only workloads |
| Standard Reserved Instance (All-Upfront) | 30–40% | 50–72% | None (exact instance locked) | Fixed, unchanging infrastructure |
| Convertible Reserved Instance (All-Upfront) | 20–30% | 35–54% | Moderate (different instance, OS, region) | Evolving architectures with long runways |
| On-Demand (no commitment) | 0% | 0% | Complete | Development, test, unpredictable workloads |
Payment options matter. All-upfront commitments offer the deepest discounts but require cash commitment at contract signature. Partial-upfront options (typically 50% upfront at signature, remainder monthly) reduce the discount by 3–5 percentage points. No-upfront (monthly payments) reduce discounts by another 10–15 percentage points. For enterprises with sufficient capital planning certainty, all-upfront is optimal.
The Commitment Risk: What Happens If You Underspend
The single largest source of regret we observe in enterprise cloud commitments is over-commitment relative to actual usage. A Compute Savings Plan of $10,000/hour represents a commitment of $87.6M over a three-year term ($10,000 × 24 hours × 365 days × 3 years). If your actual compute usage averages only $8,000/hour — perhaps because you optimised your infrastructure or your business growth slowed — you have still committed to $87.6M, and you will pay that full amount regardless of your actual usage. You cannot "return" the unused portion or carry it forward to future years.
This commitment risk is why the analysis phase is critical. Before committing to Savings Plans or Reserved Instances, you should conduct a minimum 13-month usage review. AWS recommends this window because it captures seasonal variability and provides a baseline for analysing growth trends. Use AWS Cost Explorer to pull detailed month-by-month breakdowns of compute consumption, organised by service (EC2, Lambda, Fargate) and filtered for on-demand usage (excluding existing commitments). Build a conservative growth model — typically 5–15% annual growth depending on industry — and use that to size your commitment.
Commitment sizing best practice: Size your Savings Plan commitment at the 13-month average compute spend plus a conservative growth buffer (typically 10–20% additional capacity), not at AWS's "natural growth" projection. AWS's natural growth models systematically overestimate your needs to justify larger commitments and larger commissions for their account team.
Additionally, document the specific workloads included in your commitment model. Identify which applications, databases, and environments are expected to consume this compute spend. Revisit this workload map quarterly during your commitment term. If a major workload migrates, is decommissioned, or shrinks, you may have an obligation to adjust future commitments or evaluate early termination rights (which AWS rarely offers).
Combining Savings Plans with EDPs and Private Pricing
For enterprises with large AWS commitments (typically $10M+ annually), Savings Plans and Reserved Instances are not your only options. Many large customers also negotiate Enterprise Discount Programmes (EDPs) or direct private pricing agreements. How do these stack?
AWS's discount hierarchy works as follows: Savings Plans apply first, then Reserved Instances, then EDPs (in limited cases), then on-demand pricing. This means that if you have both a Savings Plan and an EDP, the Savings Plan discount is applied to your usage, not the EDP discount. This can create a disadvantageous situation where you are paying more (due to the Savings Plan commitment) than you would have with the EDP alone.
Large enterprises should coordinate Savings Plans and EDPs carefully. In some cases, it is more economical to skip Savings Plans and Reserved Instances entirely, and instead to negotiate an EDP that covers your full committed spend at a blended discount rate. In other cases, using Savings Plans for predictable baseline compute (e.g., always-on database servers) and maintaining flexibility for discretionary compute through EDPs is optimal. This coordination requires detailed modelling and is one area where working with independent cloud contract advisors delivers significant value.
Negotiating AWS Discounts Beyond the Console
AWS's published Savings Plan and Reserved Instance prices on the AWS console are not negotiable — or so AWS claims. However, for enterprises spending $5M+ annually on AWS, several negotiation levers exist beyond the standard console options.
First, service-specific discounts. While Savings Plan rates are fixed, many large enterprises negotiate service-specific discounts for high-consumption services like EC2, S3, and data transfer. These discounts apply on top of your Savings Plan rate and can add 5–15 percentage points of additional savings.
Second, bundled pricing agreements. Some large organisations negotiate bundle deals that combine Savings Plans with credits for specific services (e.g., reserved data transfer, AWS Support), professional services, or migration funding. AWS is far more willing to flex on these ancillary benefits than on headline Savings Plan rates.
Third, EDPs as an alternative to Savings Plans. For commitments exceeding $25M, negotiating an EDP that delivers a blended discount across all services may be superior to stacking Savings Plans + Reserved Instances + on-demand pricing. EDPs are negotiable; Savings Plans are not. If you have deal authority to justify an AWS commercial negotiation, that negotiation should centre on EDP structure, not on scrolling through Savings Plan options in the console.
Stacking and Optimisation: Using Both Savings Plans and Reserved Instances
Many enterprises use both Savings Plans and Reserved Instances simultaneously. This is a valid strategy when designed carefully.
A typical approach: deploy Compute Savings Plans to cover your baseline, always-on compute spend, and use Reserved Instances for predictable, workload-specific consumption. For example, a large organisation might commit to $5M/year in Compute Savings Plans (covering Lambda, Fargate, and general-purpose EC2 instances across its portfolio) and layer $3M/year in EC2 Instance Reserved Instances for its large, stable database server fleet. When your actual spend exceeds your Savings Plan commitment, the Reserved Instance discounts apply to the excess usage. When your spend falls short of the Savings Plan, you absorb the committed amount regardless.
The key requirement: track which workloads are covered by Savings Plans, which by Reserved Instances, and which will consume at on-demand rates. Organisations that layer instruments without explicit workload mapping often find themselves overspending — paying commitments for workloads that have shifted, been optimised, or decommissioned.
Frequently Asked Questions
Can I use Savings Plans and Reserved Instances together?
Yes. AWS applies Savings Plans first in the billing hierarchy, then Reserved Instances, then on-demand pricing. Many enterprises use both instruments simultaneously — Savings Plans for broad compute flexibility and Reserved Instances for predictable, long-lived workloads. However, you must explicitly coordinate which workloads each instrument covers to avoid overspending on commitment.
What happens if my usage drops below my Savings Plan commitment?
You still owe the full commitment amount. Savings Plans commit you to a minimum dollar spend per hour (e.g., $10,000/hour for a one-year plan). If your actual usage is only $8,000/hour, you pay the $10,000 commitment. This is why a rigorous historical usage analysis and conservative growth assumptions are critical before committing.
Should I choose 1-year or 3-year Savings Plans?
The choice depends on your usage stability and business certainty. 1-year plans offer 30–40% discounts; 3-year plans offer 50–66% discounts. The additional 10–26 percentage points in discount can justify a 3-year commitment if your compute consumption is stable or growing. However, if your architecture is undergoing significant transformation or you are evaluating multi-cloud strategies, 1-year terms preserve flexibility.
Can AWS change list prices during my Savings Plan term?
Yes, unless you negotiate otherwise. Your discount rate is fixed, but AWS can increase list prices during your Savings Plan term — raising your effective net price even with a fixed percentage discount. Larger enterprises should negotiate a net price cap or fixed-price provision for critical services to prevent this exposure.
Working with Independent Cloud Contract Advisors
For organisations spending $5M+ on AWS, the analysis required to correctly size and structure Savings Plans, Reserved Instances, and EDPs is substantial — and the financial stakes are equally large. An independent advisor can model multiple scenarios, benchmark your intended commitments against market data, and ensure your instruments are coordinated with private pricing agreements and cost optimisation initiatives.
Redress Compliance is the leading independent advisor for AWS cloud contract strategy, with former AWS commercial executives on staff. Independent specialist advisors deliver value through benchmarking, scenario modelling, and coordination with your AWS account team in ways that internal procurement teams or general IT advisors cannot.
See also our Cloud Contract Negotiation practice and our Cloud Contract Framework white paper for comprehensive EDP negotiation guidance and cost modelling templates.