Rightsizing saves a workload. The commit sets the price of every workload.
Every cloud team owns a dashboard that shows spend rising, sliced twelve ways. What most lack is the machinery that turns the picture into decisions: which workstreams return money this quarter, what the commitment should be next cycle, and what the same workload costs on the other three clouds. AI supplies that machinery, and the order matters, because a commit sized 20 percent too high converts every efficiency win beneath it into unused commitment at true up.
Prepared by Redress Compliance · August 10, 2026 · Cloud advisory. Based on 20 to 25 cloud commitment reviews, 2024 to 2025.
Executive summary
The commitment is the biggest lever because it multiplies everything else. Rightsizing saves the cost of one workload; the commit, an AWS EDP, an Azure MACC, Google Cloud committed use discounts, or OCI Universal Credits, sets the discount on the entire bill for years.
In our file four out of five commits were sized above trailing consumption growth, and the forecast that justified the sizing almost always traced back to a vendor built model.
Size instead from trailing twelve month consumption forward, with three scenarios priced in full: minimum, the floor you are confident of; medium, the trend; growth, the plan. Price the cost of unused commit and the cost of overage in each, then choose with open eyes.
Coverage is a vendor aligned metric, and unused commit at true up is the number that belongs on the dashboard. The standard advice measures FinOps maturity by the share of spend under commitment, pushed toward 100 percent.
Every point bought above the consumption floor is a prepaid bet that a growth forecast comes true, and in our file the median estate carried 12 to 18 percent unused or misapplied commitment at true up while reporting excellent coverage to the board.
Buy the floor, negotiate flexibility on the rest, drive unused commit toward zero, and let coverage land wherever honest scenarios put it.
Below the commit, four workstreams fund themselves inside the quarter, and the ranking rarely changes. AI reads the daily cost exports and sizes each workstream in dollars, which turns an infinite backlog into a ranked queue.
Typical first year yield as a share of cloud spend runs roughly 4 percent from commit resizing, 3 percent from right sizing, 2 percent from scheduling, and 1.5 percent from storage.
Scheduling recovers 60 to 70 percent of run time on eligible non-production workloads, and storage meets the least resistance, because nobody defends a snapshot from 2023.
Two prerequisites decide the pace: tagging hygiene, since workstream sizing is only as good as the ownership tags, and egress honesty, since any cross cloud comparison that ignores egress and licensing effects will be dismantled in one email.
Findings expire, so optimization is a monthly rhythm rather than an annual review. Estates change daily, and a rightsizing list from last quarter describes an estate that no longer exists.
The working loop runs on fresh exports: week one, sizing out, what changed, what regressed, what is newly recoverable; week two, owner review, where each workstream owner accepts or defers with dollar values attached to both choices; week three, execution tracked upload over upload.
Week four, the FinOps and procurement joint, consumption trend against commit trajectory so the next true up surprises nobody.
The estates in the best shape ran this loop. The estates that reviewed annually described a dead estate accurately.
The commitment vehicles and where buyers overpay
| Vehicle | Structure | Where buyers overpay | Key negotiable |
|---|---|---|---|
| AWS EDP | Annual spend commit, tiered discount | Commit sized to forecast, not trend | Tier breaks, carry terms, eligibility |
| Azure MACC | Multi year consumption commitment | True up shortfalls at term end | Term length, included services, flexibility |
| Google Cloud CUDs | Resource and spend based commitments | Overlapping commitments by project | Scope, duration mix, portfolio view |
| OCI Universal Credits | Prepaid credit pool | Credits expiring unused | Credit term, rollover, rate card |
| AI platform commits | Usage commitment on model calls | Forecasts built on pilot enthusiasm | Term length, model flexibility, true down rights |
The commit is a flagship contract that happens to be denominated in consumption, so treat it with flagship discipline.
Discount tiers, term length, carry forward of unused commitment, product eligibility, and renegotiation triggers are all trade space, and all benchmarkable against real deal cohorts.
Model usage agreements with the AI platforms now reach commit scale and behave the same way, consumption based, forecast justified, and oversized by default, with one addition: model efficiency improves fast.
So shorter terms and stronger flexibility clauses are worth more than an extra discount point.
The vehicle level detail sits in the AWS EDP playbook, the Azure MACC negotiation guide, and the Google Cloud CUD playbook.
Where the workstream savings come from
- Database and compute right sizing. Instances provisioned for a launch day three years ago, databases on performance tiers their query load never touches, and dev environments running at production scale. Per instance evidence makes each change defensible to the owner who has to approve it. Roughly 3 percent of spend in year one.
- Compute scheduling. Non-production estates running nights and weekends for nobody. Scheduling recovers 60 to 70 percent of run time on eligible workloads, and the eligibility list is exactly what AI extracts from tagging and usage patterns. Roughly 2 percent.
- Storage cleanup and tiering. Orphaned volumes, forgotten snapshots, and hot tier data nobody read this year. The workstream with the least political resistance. Roughly 1.5 percent.
- Cross cloud workload pricing. The same workload priced on AWS, Azure, Google Cloud, and OCI with egress and licensing effects included. Occasionally it justifies a move. More often it is leverage, and a credible comparison beats any speech about multi cloud strategy.
- Fix tagging first. Workstream sizing is only as good as the ownership tags on the resources, and a two week tagging sprint often unlocks more than a quarter of analysis. The commit side of the same estate is covered in the AWS Savings Plans guide and the OCI cost optimization guide.
The enterprise FinOps and cost optimization brief
Commit scenario sizing, the workstream savings queue, the monthly operating rhythm, and the cross cloud comparison that gives you leverage at renewal.
Get the white paper →Running optimization as a monthly rhythm
Findings expire because estates change daily, so the operating answer is a monthly loop from fresh exports rather than an annual review from stale ones.
Week one takes fresh cost exports in and puts workstream sizing out: what changed, what regressed, and what is newly recoverable since the last pass.
Week two is the owner review, where each workstream owner accepts or defers this month's list with dollar values attached to both choices, which is what turns an analysis into a decision with a name on it.
Week three is execution and tracking against the phased plan, upload over upload, so progress is measured rather than remembered. Week four is the FinOps and procurement joint session: consumption trend against commit trajectory, so the next true up never surprises anyone in either function.
The discipline matters more than the tooling, because the same analysis that is worth a percentage of the bill in week one is worth nothing by the following quarter.
The wider estate playbook sits with the cost optimization practice, and the vehicle by vehicle detail in the AWS practice and the Google Cloud practice.
- Percentile standing for your exact deal size and industry, from real closed transactions
- Scenario simulation before the call: test alternative terms and see the financial impact of each
- A negotiation playbook, talking points, and a two page executive brief on day one
What we saw across cloud commitment reviews, 2024 to 2025
Across the 20 to 25 cloud commitment reviews we ran in 2024 and 2025, the commit decided the outcome more than any operational efficiency programme beneath it. The common advice measures FinOps maturity by coverage, pushed as close to 100 percent as possible.
We disagree, because coverage rewards exactly the behaviour that costs buyers money:
Share of commitments sized above trailing consumption growth, on a forecast that almost always traced back to a vendor built model.
Median unused or misapplied commitment carried into true up, in estates reporting excellent coverage numbers to the board.
Three patterns recurred. Buyers who arrived at a commit renewal with a cross cloud workload comparison in hand negotiated measurably better discount tiers than buyers who arrived with last year's contract.
Workstream level savings, database right sizing, storage cleanup, and compute scheduling, funded themselves inside the quarter in most estates.
And the estates in the best shape ran optimization as a monthly rhythm from fresh exports, while annual cloud reviews reliably described an estate that had already changed.
The buyer side move is to optimize in order: size the commit to consumption rather than ambition, negotiate it like the flagship contract it is, then bank the workstreams beneath it and keep the loop running.
Your first five moves
- Export daily costs for the trailing six months from every cloud account, then map commit trajectories against consumption trend and date the next true ups.
- Build minimum, medium, and growth scenarios for the next commit, priced with the cost of unused commitment and the cost of overage in each, because four out of five commits in our file were sized above trend.
- Put unused commit at true up on the dashboard and drive it toward zero, replacing coverage, which rewards the overcommitment that left a median 12 to 18 percent dead at true up.
- Bank storage, scheduling, and right sizing this quarter, roughly 6.5 percent of spend between them, after a two week tagging sprint so the sizing is defensible to workstream owners.
- Price your top workloads across the other three clouds before the commit conversation, with egress and licensing included, because the comparison is leverage even when you never move. The cost optimization practice runs the sizing and the negotiation with you.
Frequently asked questions
Why is the cloud commitment the biggest cost lever?
Because it multiplies everything else. Rightsizing saves the cost of one workload; the commit sets the discount on the entire bill for years.
A commit sized 20 percent too high converts every efficiency win beneath it into unused commitment at true up, which is why the order matters: size and negotiate the commit first, then bank the workstream savings underneath it.
How should we size a cloud commitment?
Start from trailing twelve month consumption and work forward with three scenarios: minimum, the floor you are confident of; medium, the trend; and growth, the plan. Price all three including the cost of unused commitment and the cost of overage, then pick with open eyes.
The vendor's sizing model starts from your growth story and works up, which is why four out of five commits in our file were sized above trailing consumption growth.
Is high commitment coverage a sign of FinOps maturity?
No. Coverage is a vendor aligned metric, and every point bought above the consumption floor is a prepaid bet that a growth forecast comes true. In our file the median estate carried 12 to 18 percent unused or misapplied commitment at true up while reporting excellent coverage.
Put unused commit at true up on the dashboard instead, drive it toward zero, and let coverage land where honest scenarios put it.
Which workstreams return the most money in year one?
In our review file the typical first year yield as a share of cloud spend runs roughly 4 percent from commit resizing, 3 percent from database and compute right sizing, 2 percent from compute scheduling, and 1.5 percent from storage cleanup and tiering.
Estates differ in magnitude; the ranking rarely does. Scheduling alone recovers 60 to 70 percent of run time on eligible non-production workloads.
Do we need to move clouds to benefit from a cross cloud comparison?
Rarely. Pricing the same workload on AWS, Azure, Google Cloud, and OCI occasionally justifies a move, but more often it is leverage: buyers who arrived at a commit renewal with a credible comparison in hand negotiated measurably better discount tiers.
The one condition is egress honesty, because a comparison that ignores egress and licensing effects will be dismantled by the vendor in a single email.
How do AI platform commitments differ from hyperscaler commitments?
Structurally they behave the same way: consumption based, forecast justified, and oversized by default, usually on a forecast built from pilot enthusiasm. The same three scenario discipline applies, with one addition.
Model efficiency improves fast, so a shorter term and stronger flexibility and true down rights are worth more than an extra discount point on a long commitment.
How often should cloud optimization run?
Monthly, from fresh exports. Findings expire because estates change daily, so a rightsizing list from last quarter describes an estate that no longer exists.
The working loop is sizing in week one, owner review with dollar values in week two, tracked execution in week three, and a FinOps and procurement session in week four comparing consumption trend against commit trajectory so the next true up surprises nobody.
