Reserved Instances vs Savings Plans Portfolio in 2026

Commitment portfolio is the highest-leverage FinOps decision. Get it wrong by 10% and you waste 5-figures monthly.

Why portfolio matters

Commitment portfolio is the highest-leverage single FinOps decision. Get it 10% wrong and the bill carries five-figure monthly waste.

• Wrong portfolio. Locked into capacity you do not use, or paying on-demand for capacity you would have committed to.
• Right portfolio. 30%+ savings vs all-on-demand without lock-in pain.
• 10% miss cost. Five-figure monthly waste at any meaningful scale; the math compounds.
• Hard to reverse. Bad commitments live for the term; the cost is locked in until expiry.

The 60/30/10 model

• 60% covered by 1-year Compute Savings Plans (flexibility).
• 30% covered by 1-year EC2 Savings Plans (deeper discount).
• 10% reserved for stable known workloads.

Quarterly rebalance

The portfolio drifts as workloads change. Quarterly rebalance keeps the math honest; most teams find 2-3 percentage points of additional saving from the discipline.

• Model actual usage. Pull last 90 days of usage; project the next 90 days; the data drives the rebalance.
• Expiring next quarter. Rebalance commitments expiring in the upcoming window; not the long tail.
• Trend break. Workload shifts (new product, traffic decline) change the optimal mix; quarterly is the right cadence.
• Discipline saving. 2-3 percentage points additional saving from disciplined quarterly review.

Hedging instrument selection

Each instrument fits a different workload shape. Mixing instrument to workload wrong leaves real money on the table; the match is mechanical once you have the data.

• Reserved Instances. Stable plus identical workload (e.g. always 4 m5.large in us-east-1); deepest discount.
• EC2 Savings Plans. Family-stable workload (any size of m5 in us-east-1); deeper discount than Compute SP.
• Compute Savings Plans. Flexibility needed (any instance family, any region); shallower discount, max flexibility.
• Mismatch cost. Wrong instrument to wrong workload leaves money on the table; the match decides the discount.

Antipatterns

• 100% on-demand. 30% savings unrealized.
• 100% RIs. Locked into yesterday.
• No quarterly rebalance. Drift.

What to do this week

Three moves. (1) Apply this lever to your highest-spend workload. (2) Measure the dollar impact for one month. (3) Roll the practice out to the next two services if the savings hold.