Microsoft removed programmatic volume discounts from Online Services while leaving on-premises pricing untouched, which flips the math for a specific set of workloads. This is the buyer-side comparison of which eligible estates should re-examine the on-prem and Software Assurance path they wrote off years ago.
Microsoft removed programmatic volume discounts from Online Services while leaving on-premises pricing untouched, which flips the math for a specific set of workloads. This is the buyer-side comparison of which eligible estates should re-examine the on-prem and Software Assurance path they wrote off years ago.
Two moves in 2026 reset the comparison. First, effective November 1, 2025, Microsoft eliminated programmatic volume discount Levels B, C, and D on Online Services purchased through volume licensing. All eligible Online Services (Microsoft 365, Azure, Dynamics 365, Power, Defender, GitHub) move to flat Level A list pricing, applied at each customer's next renewal regardless of seat count. A 10,000-seat organization renewing an EA in 2026 hits this too. Second, on July 1, 2026 Microsoft raised Microsoft 365 suite list prices: E3 from $36 to $39 per user per month and E5 from $57 to $60, with 57 SKUs repriced and an average buy-price increase of roughly 12 to 13 percent.
On-premises software pricing is not affected by the discount removal. That is the pivot. For years, on-prem plus Software Assurance lost to cloud subscriptions on nearly every commercial comparison because the cloud carried a negotiated volume discount that on-prem could not match. Strip out the discount and the effective per-user or per-core cost of the cloud path rises, while the on-prem list price sits still. Our buyer guide to the end of EA volume discounts covers the mechanics of the shift; this page addresses the narrower question of which workloads should now be re-priced against on-prem before you sign a cloud renewal.
The discount removal did not make cloud more expensive in isolation. It removed the structural advantage that made on-prem uncompetitive by default.
Be precise about scope. This is not an argument that on-prem beats cloud broadly. It is an argument that a set of eligible, well-understood workloads (stable core counts, predictable growth, no dependency on cloud-only features) deserve a fresh three-year total-cost model now that the cloud side lost its discount cushion. For most estates the answer will still be cloud. For a minority, the on-prem plus Software Assurance path that failed to compete in 2022 now clears the bar.
Buyers who model the discount removal in isolation understate the impact. Level A list pricing and the July 1, 2026 list increase compound. A customer previously buying E5 at a Level D discount off the old $57 list is now paying the new $60 list with no discount. Depending on your prior level, the effective increase per seat lands well into the twenties as a percentage, not the single-digit figure a naive read of the list increase suggests. We break the arithmetic down in why your Microsoft increase is 20 percent, not 8.
Large enterprise customers under the EA program could see price increases of nearly 13 percent from the discount removal alone (CIO, October 27, 2025), running into the millions for some organizations. Layer the SKU repricing on top and the compounded three-year exposure is the number that should drive the on-prem re-examination, not the headline list change. The full mechanics of losing Levels A through D are documented in what losing the price levels actually costs.
Not every workload has an on-prem equivalent, and several that do carry offsetting cost traps described later. The workloads worth pricing against on-prem are those with a genuine on-prem SKU, stable or predictable capacity, and no hard dependency on cloud-only capability. In our experience advising on renewals, the recurring candidates are these.
Workloads to exclude from this exercise: anything tied to Copilot, Teams Phone, cloud-native security in Defender and Sentinel, or Power Platform, where no on-prem substitute exists. For those, the only lever is negotiation, addressed in our EA negotiation strategy after the discount.
On-prem plus Software Assurance is a real subscription, not a one-time purchase. Software Assurance renews every three years and typically costs about a third of the license's original price. The payoff is version protection: you can upgrade to the current release (for example Windows Server 2025) without repurchasing. Model SA as a recurring cost, not a bonus, or your comparison will flatter the on-prem side.
| Item | List price (approx.) | SA cost basis | Notes |
|---|---|---|---|
| SQL Server 2025 Enterprise, per 2-core pack | $15,123 | ~1/3 of license, per 3-yr cycle | Unchanged from 2022 list |
| SQL Server 2025 Standard, per 2-core pack | $3,945 | ~1/3 of license, per 3-yr cycle | Unchanged from 2022 list |
| Windows Server 2025 Standard, per 16-core pack | $1,176 | ~1/3 of license, per 3-yr cycle | Up ~10% vs 2022; CALs still required |
| Windows Server 2025 Datacenter, per 16-core pack | $6,771 | ~1/3 of license, per 3-yr cycle | Up ~20% vs 2022; ~5.8x Standard |
| Microsoft 365 E5 (cloud comparator) | $60 pupm list | N/A (subscription) | Level A only, no volume discount |
Two structural notes. Windows Server is licensed per physical core with a minimum of eight cores per processor and sixteen per physical server, and both editions still require CALs. Core CAL Suite rose 15 percent and Enterprise CAL Suite rose 20 percent, and (critically) Windows Server 2022 CALs do not grant access to Windows Server 2025. If your model reuses existing CALs, verify version alignment before you commit.
Price Software Assurance as a three-year subscription costing a third of the license. Treat it as free version insurance and your comparison lies to you.
For workloads heading to Azure anyway, Azure Hybrid Benefit is often the strongest option because it applies on-prem-style licensing with active SA to cloud compute. For Windows Server, AHB can reduce Azure compute cost by up to 80 percent versus pay-as-you-go. For SQL Server it delivers an average 29 percent reduction. The conversion ratios matter for modeling: one SQL Server Enterprise core with SA converts to four vCores of Azure SQL Database or Managed Instance at General Purpose or Hyperscale tier (or one vCore at Business Critical), and one Standard core converts to one General Purpose vCore.
Read the dual-use rules carefully. SQL Server AHB has no dual-use right beyond a 180-day migration overlap, after which on-premises use must stop. Windows Server Datacenter keeps indefinite dual use for VM licensing, but Windows Server Standard is migrate-only with the same 180-day overlap. If your plan assumes you keep running the license on-prem while also claiming AHB in Azure, that assumption is wrong for SQL and Windows Server Standard. Our broader guidance on this sits in adapting your cloud licensing strategy.
On-prem is not a clean escape hatch. Several 2025 and 2026 changes add cost or constraint that a superficial model misses. Price these into the comparison or the on-prem case will look better than it is.
Your ability to hold on-prem licensing depends on the contract vehicle available to you. Microsoft is phasing out the EA for smaller and mid-market organizations: below 2,400 users, Microsoft will not offer an EA renewal after your current term ends. The default replacement, MCA-E, requires a $500,000 minimum annual Microsoft spend commitment and uses standardized pricing with no Software Assurance, no built-in support, and no partner advocate. If you do nothing, MCA-E is where you land.
This matters for the on-prem question because Software Assurance and perpetual on-prem licensing are traditionally acquired and maintained through the EA. Above 2,400 users the EA still typically wins on negotiated discount and the three-year lock. Below 2,400 users, the per-seat gap between EA Level A and a competent CSP is often within negotiating distance, and CSP now offers three-year subscription terms for Microsoft 365 E3, E5, and Teams Enterprise since June 1, 2025. Map your vehicle first: an on-prem strategy that assumes continued EA access will collapse if you are pushed to MCA-E. See our CIO navigation guide for the vehicle decision tree.
Run this as a workload-by-workload exercise, not an estate-wide one. For each candidate workload, build a three-year total cost model with the cloud comparator at Level A list (plus the July 1, 2026 increase), against on-prem list plus Software Assurance at a third of license cost per cycle, plus CALs, plus any hotpatching, Arc, or virtualization SA cost. Keep the migration and operational cost of running on-prem infrastructure in the model, because that is where cloud usually reclaims ground.
The honest conclusion: for the majority of seats, cloud remains the right answer even after the discount removal, because operational cost, feature dependency, and Copilot adoption pull that way. But the discount change is real leverage on a defined minority of workloads, and Microsoft's account teams will not surface the on-prem comparison for you. Build the model, name the eligible workloads, and use the credible on-prem alternative as both a cost saver where it wins and a negotiation anchor where it does not. The manufactured-discount and benchmarking tactics that support that anchor are covered in the discount removal cost guide.
No. The removal of programmatic volume discount Levels B, C, and D applies to Online Services (Microsoft 365, Azure, Dynamics 365, Power, Defender, GitHub). On-premises software pricing is not affected. That gap is precisely why some eligible on-prem workloads are now worth re-pricing against cloud.
Workloads with a genuine on-prem SKU, stable or predictable capacity, and no dependency on cloud-only features. The recurring candidates are steady-state SQL Server, Windows Server estates with predictable VM density, Exchange and SharePoint deployments, and Azure-bound workloads that can use Azure Hybrid Benefit. Exclude anything tied to Copilot, Teams Phone, Power Platform, or cloud-native security.
Software Assurance renews every three years and typically costs about a third of the original license price. Model it as a recurring subscription, not a one-time or free add-on. For SharePoint Server Subscription Edition and virtualized SQL Server, active SA is mandatory, so it is not optional cost you can strip out.
It depends on your replacement vehicle. Below 2,400 users Microsoft will not renew an EA, and the default MCA-E carries a $500,000 annual minimum with no Software Assurance and no partner advocate. Confirm you can still acquire and maintain SA before building an on-prem strategy, because an assumption of continued EA access will collapse under MCA-E.
AHB lets you apply on-prem Windows Server and SQL Server licenses with active SA to Azure compute, cutting Windows Server compute cost by up to 80 percent versus pay-as-you-go and SQL Server by an average 29 percent. It usually wins for workloads heading to Azure anyway. Note the dual-use limits: SQL and Windows Server Standard allow only a 180-day migration overlap, while Windows Server Datacenter keeps indefinite dual use for VM licensing.
Yes. Hotpatching Windows Server 2025 outside Azure requires Azure Arc at $1.50 per core per month (a 64-core host is $1,152 a year). Windows Server 2022 CALs do not cover 2025, Core and Enterprise CAL Suites rose 15 to 20 percent, and virtualized SQL Server requires SA. Price all of these in or the on-prem case will look better than it is.
Every material Microsoft licensing change in 2026: EA terms, Copilot pricing, Defender packaging, and Azure rules, plus what each one costs you.
Gated with a work email on the download page. No sales follow up you did not ask for.
Get the White Paper →
500+ enterprise clients. 11 vendor practices. Industry recognized. One conversation can change what you pay for the next three years.
One buyer side briefing a week. Renewal signals, audit moves, and the levers that work. No vendor spin.
