Microsoft introduced Azure Savings Plans in October 2022, positioning them as a more flexible alternative to Reserved Instances. The messaging was straightforward: Savings Plans apply a discount to a broader basket of compute services without tying you to a specific VM series or region. What Microsoft didn't lead with is that Savings Plans typically deliver 5-15% less discount than Reserved Instances for equivalent commitments.
For procurement teams and CFOs managing Azure spend, the RI vs. Savings Plan decision is consequential. The wrong choice — particularly when compounded across a large Azure estate — can mean leaving $500,000 to $2M in savings on the table annually. This guide covers how both mechanisms work, when to use each, and how to negotiate both within your Microsoft Enterprise Agreement.
Azure Reserved Instances: How They Work
Overpaying for Enterprise Software? We handle software and cloud contract negotiation on a 25% gainshare basis — you keep 75% of every dollar saved. No retainer. No risk.
Get a free Enterprise Software savings estimate →Azure Reserved Instances are commitments to use a specific VM series in a specific Azure region for either 1 or 3 years. You pay upfront (full), monthly, or with no upfront cost (though no-upfront carries a lower discount). In exchange, Microsoft applies a reservation discount to matching compute usage, eliminating the on-demand hourly rate for that capacity.
The key characteristic of RIs is their specificity. A reservation for a D4s v3 instance in East US does not apply to a D4s v3 in West Europe, nor to a D8s v3 in East US. The discount only applies when your actual usage matches the reservation on the dimensions of region, VM family, and instance size flexibility group.
The exception is Instance Size Flexibility, which Microsoft enables by default for most reservation purchases. With size flexibility, a D4s v3 reservation can also cover D2s v3 or D8s v3 instances within the same VM series — applying normalisation ratios. A D2s v3 has a normalisation factor of 2, a D4s v3 has a factor of 4, and a D8s v3 has a factor of 8. So one D4s v3 reservation (factor 4) can cover two D2s v3 instances.
Reserved Instance Discount Ranges by VM Series (2026)
| VM Category | 1-Year RI Savings | 3-Year RI Savings | Notes |
|---|---|---|---|
| General Purpose (D-series) | ~40% | ~60% | Most commonly used; good instance size flexibility |
| Compute Optimised (F-series) | ~40% | ~62% | Ideal for high CPU/low memory workloads |
| Memory Optimised (E/M-series) | ~38% | ~60% | SAP HANA and in-memory databases |
| GPU (NC/ND-series) | ~30% | ~57% | AI/ML workloads; lower flexibility |
| Storage Optimised (L-series) | ~35% | ~58% | Big data and NoSQL databases |
| Azure Dedicated Host | ~30% | ~50% | For regulatory isolation; Oracle BYOL use cases |
Azure Savings Plans: How They Work
Azure Compute Savings Plans work differently. Instead of committing to specific VM types, you commit to a dollar-per-hour spend level on compute. Microsoft then applies a discount to your eligible compute consumption, regardless of VM series, size, or region. The commitment might be "$10/hour on compute for 3 years" — any compute spend up to that level receives the Savings Plan discount rate.
This flexibility is valuable when your workload mix is unpredictable or when you actively use multiple VM series across multiple regions. A financial services firm running development workloads in East US and production in West Europe with a mix of D-series and E-series VMs would struggle to allocate RIs optimally. A Compute Savings Plan covers the entire footprint without over-engineering the purchase.
The downside is the discount depth. Azure Compute Savings Plans typically deliver 65% maximum savings versus on-demand for a 3-year commitment, compared to up to 72% for RIs in the same VM series. For a stable, predictable workload, that 7-percentage-point difference is meaningful. On $3M annual Azure compute spend, it represents $210,000 in annual savings foregone for the sake of flexibility you don't actually need.
Azure Savings Plans and Reserved Instances can be used simultaneously. RIs apply first to matching usage, then Savings Plans apply to remaining eligible compute. The optimal strategy for most large Azure deployments is RIs for predictable, stable workloads plus a Savings Plan for the variable tail. Running only Savings Plans is the costlier choice.
The Enterprise Agreement Angle
What most Azure FinOps teams don't explore is that both RIs and Savings Plans are available as EA commitments, not just pay-as-you-go purchases. When you're negotiating your Microsoft EA renewal, reserved capacity and savings plan commitments can be integrated into the deal in ways that create additional discount leverage.
Microsoft's enterprise sales team has flexibility to offer enhanced RI discounts when they're packaged as part of a broader EA commitment. An enterprise committing to $10M in annual Azure spend across a 3-year EA, inclusive of $3M in RI commitments, is in a fundamentally stronger negotiating position than one purchasing $3M in RIs separately through the Azure portal. The total commitment value changes Microsoft's internal deal economics significantly.
Similarly, Microsoft Azure Consumption Commitment (MACC) arrangements — which provide incentives for hitting Azure spend thresholds — interact with RI and Savings Plan purchases. Understanding whether your RI purchases count toward MACC thresholds, and negotiating this explicitly, is a lever that our Microsoft negotiation team uses consistently to improve deal terms.
RI vs Savings Plan: The Decision Framework
Use Reserved Instances When:
Use Azure Savings Plans When:
The Full Comparison
| Dimension | Reserved Instances | Azure Savings Plans |
|---|---|---|
| Discount mechanism | Applied to specific VM type + region + OS | Applied to eligible compute up to hourly commitment |
| Max discount (3-year) | Up to 72% vs on-demand | Up to 65% vs on-demand |
| Scope | Specific VM series, region, size group | Any eligible compute across all regions and VM series |
| Flexibility | Low — but Instance Size Flexibility helps within VM family | High — applies wherever you use compute |
| Exchange/refund | Can exchange (once) or cancel with 12% early termination fee | Cannot cancel; commitment is fixed for term |
| Applies to | VMs, SQL Database, SQL Managed Instance, Cosmos DB, Synapse, Blob Storage (specific services) | VMs, Container Instances, Azure Dedicated Host, App Service Premium |
| Interaction with Savings Plans | Applied first; Savings Plan covers remaining usage | Applied after RIs; covers residual compute |
| EA negotiation lever | Strong — Microsoft offers enhanced RI pricing in EA context | Moderate — less negotiating room built into pricing |
| MACC eligibility | Eligible (negotiate explicitly) | Eligible (negotiate explicitly) |
Common Mistakes That Cost Enterprises Millions
Over-purchasing Savings Plans for stable workloads. This is the most expensive mistake. Microsoft's Savings Plan marketing emphasises simplicity, but simplicity costs. An enterprise running $5M/year in stable D-series compute that buys a Compute Savings Plan instead of RIs foregoes $350,000 annually in deeper discounts — over a 3-year term, that's over $1M.
Purchasing RIs without usage analysis. The opposite problem: buying RIs without confirming that the underlying workloads are actually stable. Unutilised RIs still carry their commitment cost. An RI for a D4s v3 in East US costs money whether or not you use D4s v3 instances in East US. RI utilisation rates below 80% typically indicate poorly matched purchases.
Not exchanging RIs after re-architecture. Microsoft allows RI exchanges — one exchange per reservation, with no penalty. Enterprises that re-architect their infrastructure often end up with RIs that no longer match their workload. They assume the RI is wasted and pay full on-demand rates on the new workload. The correct move is to exchange the RI for one that matches the new workload before the exchange window behaviour changes.
Buying RIs through the portal rather than negotiating through the EA. Self-service RI purchases in the Azure portal are priced at Microsoft's standard published rates. EA customers have leverage to negotiate improved RI pricing as part of their broader commitment. Our Microsoft EA negotiation service regularly secures an additional 10-15% on RI pricing for clients who fold their RI commitments into a structured EA conversation.
Our general recommendation for enterprises with stable Azure footprints: purchase RIs for your bottom 70-80% of compute that is truly predictable (production databases, core application tiers, SAP infrastructure). Use a Savings Plan for the remaining 20-30% of variable compute. This hybrid approach typically achieves 95%+ of the maximum possible discount while preserving flexibility for the tail of your estate.
The Negotiation Playbook: Getting Better Than Portal Pricing
Azure RI and Savings Plan pricing shown in the Azure portal and Cost Management is Microsoft's published rate. It is not the floor — it's the ceiling. Enterprise Agreement customers negotiating with Microsoft's enterprise sales team can access improved pricing on both instruments when they're structured correctly.
The negotiation variables include: total EA commitment value (larger commitments provide more room), payment terms (upfront payments receive additional discounts), mix of services committed (Microsoft values Azure PaaS commitments more than commodity compute), and strategic context (a customer that Microsoft wants to grow in Azure has more leverage than one already running at scale).
We negotiate Microsoft EA renewals that include structured RI and Savings Plan commitments as part of the deal architecture. A $15M 3-year EA with $4M in integrated RI commitments is negotiated very differently from a $15M EA with RIs purchased separately. The former allows Microsoft to count the full $15M toward their sales rep quota and gives us leverage to negotiate both the EA discount and the RI pricing simultaneously.
Our Microsoft negotiation team works on a 25% gainshare basis. We only get paid when you save. On a $10M Azure footprint where we improve your RI/Savings Plan mix and negotiate better EA terms, the savings typically range from $1.5M to $3M over the contract term. At 25% gainshare, you keep $1.1M to $2.25M of savings you wouldn't have captured otherwise.
Get a Free Azure Cost Optimisation Assessment
We analyse your current Azure spend, identify RI and Savings Plan optimisation opportunities, and negotiate better terms on your behalf. 25% gainshare — no savings, no fee.
Get Your Free Azure Assessment Microsoft Negotiation Service →Related Reading
- Microsoft EA Renewal 2026: 15 Tactics to Reduce Your Cost
- Microsoft Copilot ROI: Before You Add It to Your EA
- Enterprise FinOps: When Cloud Cost Tools Aren't Enough
- Microsoft EA Renewal Guide 2026 — Free White Paper
- Microsoft Negotiation Service
- Cloud Cost Negotiation Service
- Software Savings Estimator Tool
- How Our Gainshare Model Works