Hybrid Benefit sat unapplied on 25 to 45 percent of eligible workloads, so the largest Azure saving available was an entitlement the buyer had already paid for
Rightsizing feels like the work. In roughly 22 of the 30 plus estates benchmarked it recovered single digits, while two licensing levers left 20 to 40 percent unclaimed.
Prepared by Redress Compliance · August 18, 2026 · Microsoft advisory. 30 to 40 Azure estates benchmarked, 2024 to 2025.
Executive summary
Azure Hybrid Benefit went unapplied on 25 to 45 percent of eligible Windows and SQL workloads. The licenses were owned and covered by Software Assurance. Nobody flagged the workloads, so they ran at pay as you go rates.
Reservation coverage sat below 50 percent on workloads that run every hour. A stable baseline should sit near 90 percent covered. Everything below that line is on demand pricing for demand that never varies.
Rightsizing recovered single digit percentages in roughly 22 of the 30 plus estates. It is where most cloud cost programs start, and it is the smallest of the three levers on the table.
Both large levers sit at the seam between procurement and engineering. Procurement owns the license entitlement, engineering deploys the workload, and neither team sees the other's view of the same virtual machine.
Which Azure lever actually moves the invoice?
Three levers move an Azure bill, and they are not equal. Licensing entitlement and commitment coverage move it in double digits. Infrastructure tuning moves it in single digits, and it is the one most programs run first.
The ranking matters because attention is finite. A quarter spent shutting down idle virtual machines is a quarter not spent reconciling license entitlement against running workloads.
| Lever | What it needs | Typical movement | Who owns it |
|---|---|---|---|
| Hybrid Benefit | A license entitlement view joined to running workloads | Up to 40 percent on eligible compute | Procurement, applied by engineering |
| Reservation and savings plan coverage | A stable baseline you are willing to commit | Substantial on the committed layer | Finance and engineering jointly |
| MACC drawdown | Monthly burn tracking against the term clock | 5 to 12 percent of committed value protected | Procurement |
| Rightsizing and idle cleanup | Utilization data per resource | Single digit percentages | Engineering |
Why the licensing levers get skipped
They are nobody's job. Cost dashboards report consumption, not entitlement, so the unapplied benefit never appears as a finding. The tool shows what you spent, not what you already own.
An Azure cost tool cannot see a license you bought in 2019 and never attached to a workload. It reports the rate you paid as though it were the rate available. That gap is the finding, and it stays invisible until somebody joins the two views by hand.
How much does Hybrid Benefit save, and why is it missed?
On an eligible workload the saving reaches up to 40 percent against pay as you go, and more than half on SQL Server once it is combined with a reservation. Microsoft publishes the arithmetic on its Azure Hybrid Benefit page.
The mechanism is simple. Existing Windows Server or SQL Server licenses with Software Assurance are applied to Azure compute, and the rate drops to the base compute charge.
| Workload | Eligible | Typical saving |
|---|---|---|
| Windows Server virtual machine with Software Assurance | Yes | Up to 40 percent |
| SQL Server with Software Assurance | Yes | More than half when paired with a reservation |
| Pay as you go, no Software Assurance | No | Not applicable |
Our Hybrid Benefit guide for Windows and SQL works the eligibility rules and the reservation stacking in detail.
Where the entitlement view breaks
License entitlement lives with procurement. Virtual machine deployment lives with engineering. Without a shared view, eligible workloads run at full rate for years and no report flags it.
That is why this is the fastest saving in Azure. It needs no architecture change, no migration, and no vendor conversation. It needs one reconciliation.
The Azure licensing and cost optimization brief
Hybrid Benefit eligibility, the reservation and savings plan split, MACC drawdown, and the levers consumption tools cannot see.
Get the brief →What 30 to 40 Azure estates showed
Across roughly 30 to 40 Microsoft Azure estates Fredrik Filipsson benchmarked between 2024 and 2025, the largest savings came from licensing levers rather than infrastructure tuning. The pattern held across estate sizes.
Hybrid Benefit was underapplied on 25 to 45 percent of eligible Windows and SQL workloads. In every case the licenses were owned. What was missing was the join between the entitlement list and the running estate.
Reservation coverage sat below 50 percent on workloads that should have been near 90. These were not bursty development environments. They were production systems running every hour of every day at on demand rates.
MACC commitments were underspent often enough to put 5 to 12 percent of committed value at risk of lapsing. That is money already promised, and losing it is the one Azure outcome with no upside at all.
In roughly 22 of those estates the rightsizing work recovered single digit percentages while the two licensing levers left 20 to 40 percent unclaimed. The sequence most programs run is therefore backwards.
- Your quote benchmarked against 500,000+ real closed deals, adjusted for size, region, and industry
- Commitment sized against trailing consumption rather than the growth forecast the vendor helped write
- Uplift and drawdown language flagged with the exact quote, the page, and the replacement text
Reservations or savings plans, and how much coverage is enough?
Cover the stable base with reservations and the variable layer with a savings plan, then aim for roughly 90 percent coverage of the baseline. Microsoft sets out both instruments in its cost management documentation.
- Reservations lock a specific resource for one or three years at a deep discount, and suit workloads that will not change shape.
- Savings plans commit to an hourly spend and flex across compute, which suits workloads that move between regions or virtual machine families.
- The blend is what most estates need, because most estates have both a fixed floor and a variable layer above it.
Coverage decisions belong inside a governance model rather than a spreadsheet. Our Azure cost governance framework sets out who signs off on what.
What coverage below 50 percent actually costs
It means paying the on demand rate for capacity that runs every hour. The workload is not flexible, so the premium buys no flexibility. It is simply an unexercised option.
How do you avoid underspending the MACC?
Track burn monthly against the term clock and route every eligible marketplace purchase against the commitment. The Microsoft Azure Consumption Commitment is a spend floor you agreed in exchange for discount, and unspent commitment is lost value.
Eligible third party software bought through the marketplace counts toward the commitment on the terms Microsoft sets out in its Azure pricing resources. Estates that never route marketplace spend leave the drawdown lever unused.
The sizing of the next commitment is the real decision. Our brief on negotiating the MACC prices the commit against trailing consumption rather than the forecast.
An underspent commit is also a negotiation asset
If you miss the floor consistently, the next commitment should be smaller. That is an argument backed by your own consumption history, which is the only benchmark the vendor cannot dispute.
What the estates measured, 2024 to 2025
Two cuts of the engagement file frame the size of the gap.
Windows and SQL workloads that qualified for Hybrid Benefit and ran at full rate anyway, across the estates benchmarked.
The gap that remained after rightsizing work had recovered its single digit percentages, in roughly 22 of the estates.
A reconciliation pass takes days rather than quarters. The blocker is not analysis, it is that two teams hold two halves of the same picture and rarely put them on one page.
Watch the briefing · 6:12The Microsoft EA Renewal PlaybookWhere the Azure lines sit inside the wider agreement, and when to open them.
Your first five moves
- Reconcile owned Windows and SQL licenses against running Azure workloads, which is the single join that surfaces the whole finding.
- Apply Hybrid Benefit to every eligible workload, since 25 to 45 percent of them were running without it across the estates benchmarked.
- Lift reservation coverage on the stable baseline toward 90 percent, and leave the variable layer to a savings plan.
- Track MACC burn monthly and route eligible marketplace spend against it, so that 5 to 12 percent of committed value does not lapse.
- Only then tune the infrastructure. The Microsoft practice runs the entitlement reconciliation before the consumption review.
Frequently asked questions
How much can Azure Hybrid Benefit save?
Up to 40 percent on eligible Windows Server compute, and more than half on SQL Server when it is combined with a reservation. The saving applies only where the underlying license carries active Software Assurance.
Why do so many estates miss Hybrid Benefit?
Because the entitlement and the deployment sit in different teams. Procurement holds the license record and engineering deploys the workloads, so eligible machines run at full rate until somebody joins the two lists.
How much of the estate was missing the benefit?
Between 25 and 45 percent of eligible Windows and SQL workloads, across the 30 to 40 estates benchmarked in 2024 and 2025. The licenses were owned in every case.
Is rightsizing a waste of time?
No, but it is the smallest lever. In roughly 22 of the estates benchmarked it recovered single digit percentages while unapplied Hybrid Benefit and weak reservation coverage left 20 to 40 percent unclaimed.
What reservation coverage should a stable workload have?
Near 90 percent. Below 50 percent you are paying on demand rates for capacity that runs every hour, which buys flexibility the workload will never use.
Should I use reservations or a savings plan?
Use both. Reservations suit stable resources that will not change shape, savings plans suit compute that moves across regions or virtual machine families, and most estates carry both patterns.
What happens to an unspent MACC?
The unspent portion is lost. Across the estates benchmarked, 5 to 12 percent of committed value was at risk of lapsing, which is the only Azure outcome that carries no upside.
Can marketplace purchases count toward the commitment?
Yes, eligible third party software bought through the Azure Marketplace draws down the commitment. Estates that never route marketplace spend leave that lever unused.
Does the agreement vehicle change the price?
Yes. The Enterprise Agreement, CSP, and Microsoft Customer Agreement carry different pricing, commitment rules, and flexibility, so the vehicle should match how your consumption actually behaves across the year.
What is the fastest saving available in Azure?
Applying Hybrid Benefit to workloads that already qualify. It needs no architecture change, no migration, and no vendor conversation, only a reconciliation between the entitlement list and the running estate.
