The Microsoft Enterprise Agreement, end to end, for 2026.

Two things happened to the Microsoft Enterprise Agreement between 2020 and 2026. The Level D discount that justified the vehicle fell from 22 to 25 percent to 10 to 12 percent, and Microsoft stopped renewing commercial enrollments below 2,400 licenses. The EA is now one option among three rather than the default for a large estate. This page works the arithmetic at Microsoft's published list prices, so the table you screenshot is one your CFO can rebuild.

An EA renewal arrives as one number on one piece of paper. It is eleven numbers. Microsoft prefers the single number because a single number can only be accepted or rejected, and rejecting it means rejecting the whole renewal. Split it into its eleven line items and each one has to survive on its own evidence. The table below is the split we use, with the document or console screen where each number actually lives.

Everything that follows is drawn from the 60 to 80 EA renewals described above, run out of our Microsoft advisory practice. Read alongside the EA negotiation strategies and the CIO playbook for evaluating Microsoft renewal proposals.

The eleven line items

Three of these eleven are settled before you sit down. Price level is set by license count. Eligibility is set by Microsoft. The renewal uplift is a published-nowhere percentage applied to your expiring rate. The other eight are yours to move.

What the Microsoft Enterprise Agreement is, and who it is for

The EA is a three year volume commitment. You commit a baseline of products and quantities at signing, price it at a level set by qualified user or device volume, and reconcile growth once a year through the true up. The program rules sit on the Microsoft Enterprise Agreement program page. The binding entitlements and use rights sit in the Microsoft Product Terms. Read both before you model a renewal.

The entry threshold is 500 users or devices. Below 500 the EA is not the route and CSP is. The structure rewards scale and punishes uncommitted change, which is why the baseline number deserves more scrutiny than the discount percentage.

Enrollments and co-termination

The EA is a master agreement. The commercial detail lives in enrollments, which are sub agreements for specific product families, such as the enrollment covering Microsoft 365 or the server and cloud enrollment. Each enrollment carries its own terms and its own dates.

Enrollments expiring on different dates hand the account team a rolling series of small negotiations, each one too small for you to fight over. Co-terminate the enrollment dates and you get one date, one number, and every product family on the table at once. That is the difference between three conversations you lose quietly and one you can win.

The annual true up in one screen

The true up is the annual count of seats and products added since the baseline. You report the net increase and pay for it, prorated for the remaining term. Three rules answer almost every question a buyer has about it.

• Annual true up. Seats can be added and are billed prorated. Seats cannot be reduced.
• Mid term. Additions run through the true up only. No reductions.
• Renewal anniversary. Seats can be added, and this is the only window in which seats can be reduced.

The result is a ratchet. True ups only go up, so the estate drifts above real headcount for as long as the term runs, and the renewal anniversary is the single point at which the baseline resets down. On a three year term that means you pay for every leaver until the anniversary comes round.

Three clauses that blunt the ratchet

• Unit price hold. Lock the unit price so growth bills at your rate rather than the prevailing level rate.
• Annual reporting basis. Report once a year, not at every change. Reporting at every change turns the true up into a rolling audit.
• Reduction right. Negotiate the ability to drop counts at the anniversary, and a midterm reduction right if the estate is likely to shrink. The base agreement gives you neither.

The full true up rules matrix, the qualified user and qualified device counting rules, and the counting mistakes that cost the most sit in the Microsoft EA true up complete guide. Read it before you submit a count.

Software Assurance: what is bundled and what goes unused

Software Assurance is bundled into the EA and priced into what you pay. Most buyers use a fraction of it. The benefit list sits in Software Assurance benefits. Inventory the rights before renewal and either use them or price them into the deal.

• New version rights. Access to upgrades released during the term.
• Use rights. Home use, roaming, and disaster recovery rights.
• Planning and training days. Bundled days that often expire unused.

Unused planning and training days are the easiest item here to trade. They cost Microsoft real delivery money, you have already paid for them, and an account team that will not move on unit price will often move on days.

The price level collapse

An EA prices seats at one of four volume bands, Level A through Level D, set by qualified license count and printed on the enrollment's Customer Price Sheet. Level D carried the deepest discount and was the commercial reason a 20,000 seat organization signed an EA instead of buying through a partner. Across the last three renewal cycles Microsoft has narrowed the gap between the bands.

Microsoft has never published a price level discount schedule and does not publish one now. The two columns below are what we have observed on Customer Price Sheets and first renewal quotes across our own 2024 to 2025 file, expressed as the bands they are. If your reseller shows you a "standard Level D discount", ask which document it is printed in. There is not one.

Compare the Level C and Level D rows rather than reading down a column. In 2020 Level D sat seven points below Level C at both ends of the band, 22 to 25 against 15 to 18. In 2026 that premium is two points at both ends, 10 to 12 against 8 to 10. Crossing 15,000 licenses used to be worth planning around. It is now worth about as much as a competent procurement lead can win by asking twice.

EA price level discounts, 2020 versus 2026 — observed, not published by Microsoft

Tier	User threshold	Headline discount 2020	Headline discount 2026	2026 delta
Level A	500 to 2,399	Baseline	Baseline	—
Level B	2,400 to 5,999	8 to 10%	4 to 6%	Narrowed
Level C	6,000 to 14,999	15 to 18%	8 to 10%	Narrowed
Level D	15,000+	22 to 25%	10 to 12%	Narrowed sharply

Buyer side response

• Test the band before you take the discount. Crossing a price level band at renewal can beat a negotiated discount outright. Confirm the true qualified user count against the band edges — 500, 2,400, 6,000, 15,000 — against Microsoft 365 plan pricing before you accept a discount instead.
• Re-run the vehicle comparison. EA versus MCA versus CSP at the renewal cycle, with the actual deployment economics rather than the publisher's narrative. The criteria are in EA vs MCA vs CSP below.
• Anchor the discount tier protections to the EA term. Frame the tier protections as non-commercial buyer side protection, not a concession.
• Audit the Level D customers' pricing benchmarks against MCA and CSP equivalents before accepting the publisher's renewal package.

The 2026 price reset

Here is the awkward part, and the reason this section is shorter than the headline suggests. Microsoft has not published a percentage increase on Microsoft 365 E3 or E5 for 2026 EA renewals. There is no rate card, no effective date, no blog post. What arrives is a renewal quote with a per seat number on it and no derivation. Anyone who prints a specific 2026 reset percentage is either quoting one customer's quote or making it up.

What Microsoft did publish, and what you can therefore hold it to, is three things.

One: the 5 percent monthly billing premium. Announced in January 2025 and effective 1 April 2025, annual term subscriptions billed monthly carry a 5 percent uplift over the same subscription billed annually. On 20,000 E3 seats that is 20,000 x $36.00 x 0.05 = $36,000 a month, or $432,000 a year, for identical software delivered on identical terms. It is the cleanest money on the whole renewal and the easiest to give back: pay annually.

Two: the last real increase, so you have a reference point. Microsoft raised enterprise list prices on 1 March 2022, announced the previous August and described at the time as the first substantive update since Office 365 launched in 2011. Those percentages are on the record:

That is the shape of a published Microsoft increase: named SKUs, a stated old and new price, one effective date. If a 2026 renewal uplift cannot be described that way, it is a quote, not a price change, and quotes are negotiable.

Three: current list. Every net price in an EA is one of these figures less your price level and less your negotiated discount. Neither of the two reductions is published; the starting figures are.

Published list prices

One line in that table repays reading twice. Microsoft unbundled Teams in April 2024 and priced the pieces: E3 without Teams at $33.75 plus Teams Enterprise at $5.25 comes to $39.00, against $36.00 for E3 with Teams included. E5 behaves the same way, $54.75 plus $5.25 = $60.00 against $57.00. Splitting Teams out costs $3.00 per user per month, $720,000 a year on 20,000 seats. The unbundling is real and the pricing makes taking it up irrational for almost everyone. That is a deliberate design, and it tells you how Microsoft prices optionality when it is forced to offer it.

Against a renewal uplift you cannot see a rate card for, three moves work. Ask for the uplift as a percentage of your expiring per seat rate, SKU by SKU, in writing, before discussing anything else; an account team that will not write it down does not have it. Change the mix, which is worth more than the uplift, as the arithmetic further down shows. And buy the price hold: a written per seat protection for the enrollment term, framed as standard buyer side language rather than as a concession you are asking for. Across our 2024 to 2025 file, first renewal quotes came in 8 to 15 percent above the expiring contracted per seat rate before any discount conversation, and the spread inside that band tracked estate size more than it tracked anything else. That is an observed range from our own engagements, not a Microsoft rate.

Read the 2026 Microsoft price increase analysis for the SKU by SKU detail.

NCE conversion

The New Commerce Experience began in the CSP channel in 2022 and its rules have been migrating into EA renewal paper ever since. Two of those rules cost money and one costs flexibility.

The cancellation window is seven calendar days. Cancel an annual term subscription inside seven days of the term start and you get a prorated credit. On day eight there is no cancellation and no reduction for the remainder of the term, whatever the deployment does. That single rule is why a seat count agreed in a rush at renewal is a seat count you own for a year. The 5 percent monthly billing premium is the second rule. The third is that per seat pricing is fixed at the term start rather than flexed across the term, which is the flexibility EA customers used to have without asking.

Three things to write back in

• A per seat price hold for the enrollment term. Named SKUs, named rate, named end date. Not "pricing consistent with the current agreement".
• A reduction right at each anniversary. A stated percentage of committed seats you may drop without penalty. Ten percent is a normal ask and is granted more often than it is requested.
• SKU substitution rights. The right to move a seat from E5 to E3, or E3 to F3, at the agreed rates. This is the operational version of price flexibility and it survives audits better than a verbal understanding.

The Copilot bundling pattern

Copilot is the load bearing commercial conversation across the 2026 EA renewal cycle. The publisher's preferred trajectory is broad population coverage across the M365 user base, with Copilot as the default M365 add on.

Copilot is priced at the enterprise tier with no discount on the SKU itself. The commercial concession framework comes from Copilot specific Azure consumption credits and bundled deployment services.

Four counter moves

1. Deployment trajectory framework. Anchor the Copilot population to the actual productivity deployment plan, not broad coverage.
2. Discount tiers at population thresholds. Tied to staged adoption rather than full enterprise rollout.
3. Unbundle the package. Separate the Copilot SKU from Azure credits and deployment services. Negotiate each independently.
4. Copilot terms. Deployment flexibility, SKU substitution rights, and cancellation framework.

Copilot and Defender attach is where the account team recovers margin. A concession given against core seats comes back through attach, so an E3 discount paid for with a broad Copilot commitment is not a discount. Price Copilot on measured adoption, in its own negotiation, against the current scope on the Microsoft 365 Copilot page. The deeper analysis sits in the Microsoft Copilot licensing guide for 2026.

The Azure MACC framework

The Azure consumption framework is restructured at every renewal cycle as the Microsoft Azure Consumption Commitment, or MACC. MACC is the prepaid commitment against Azure consumption across the EA term.

MACC has historically been a five year horizon. The publisher pushes the six or seven year framework across the 2026 cycle. MACC is constructed as a unified element of the EA renewal package, with M365 and Azure discount structure tied to MACC acceptance.

Three buyer side moves

• Unbundle the MACC commitment from the EA renewal package. Negotiate as a standalone commercial conversation with the Azure account team.
• Three or four year horizon rather than six or seven year. Preserves optionality across the multi cloud context.
• Consumption baseline plus measured growth, not the publisher's preferred aggressive growth assumption.

60 to 80

Microsoft EA renewals benchmarked, 2024 to 2025

8 to 14%

E5 right size return on the M365 envelope

22 to 38%

Azure MACC over run against trailing twelve month consumption

Source: Redress Compliance advisory engagement file, 2024 to 2025.

Our position, and who it irritates

Ask any Microsoft partner to model a clean 2026 EA and the same design comes back: everyone on E5, Copilot broadly deployed, a long MACC sized on the strategic plan, all of it on one signature. It is coherent. It is also, on the arithmetic below, the most expensive design available. On 20,000 seats, all E5 at list is $13,680,000 a year against $10,116,000 for the tiered mix. No price level discount recovers $3,564,000, because 10 to 12 percent of $13,680,000 is $1,368,000 to $1,641,600 and the mix gap is more than twice that.

The deeper objection is structural. One bundled paper means one price, and one price can only be accepted or refused whole. Eleven line items can be refused one at a time, which is the only form of refusal a buyer with a live Microsoft estate can actually execute. Running it that way makes the renewal longer, colder and less collegial. Account teams say so, sometimes to the CIO directly. That reaction is not evidence the approach is wrong.

The tidiest EA on paper is usually the most expensive one signed

Microsoft's partner channel sells a clean end state: everyone on E5, Copilot rolled out enterprise wide, and an Azure MACC sized on the strategic forecast. It is easy to approve and easy to defend internally, which is why it gets signed. In roughly three out of four enterprise renewals we have benchmarked, that posture costs the buyer 18 to 32 percent more than a tiered E3 base with selective E5 add ons, a quarantined Copilot cohort, and a MACC sized on trailing twelve month run rate. The tidy version is not simpler to run. It is pre-negotiated in Microsoft's favor.

Microsoft 365 mix and run rate

This is the section that pays for the rest. The M365 line is the largest recurring number in almost every EA, and it is the one number a buyer can move without Microsoft's permission, because it is a function of which users hold which SKU rather than of what discount was agreed.

M365 SKU layers and the population bands we see

SKU layer	Use case	Typical population share	Buyer side framing
M365 E3	Productivity baseline	60 to 70%	Default for the broad knowledge worker population
M365 E5	Productivity + full E5 stack	10 to 20%	Anchor to users that deploy all four E5 layers
M365 E5 Security	E3 + security add on	15 to 25%	Where security suite is deployed but compliance is not
M365 E5 Compliance	E3 + compliance add on	5 to 10%	Regulated user populations only
M365 F1 or F3	Frontline workers	Variable	Operations, retail, manufacturing populations

The same estate, two ways

Take 20,000 seats and put a published list price against each population. The modeled shares below each sit inside the observed bands in the table above and sum to 100 percent. No frontline population is assumed; adding F3 seats at $8.00 only pushes the total down further.

Each row is 12 months of the per user rate: 12,000 x $36.00 x 12 = $5,184,000, 3,000 x $57.00 x 12 = $2,052,000, 4,000 x $48.00 x 12 = $2,304,000, 1,000 x $48.00 x 12 = $576,000. They sum to $10,116,000. Everyone on E5 is 20,000 x $57.00 x 12 = $13,680,000. The gap is $3,564,000 a year, 26 percent of the all E5 line. Substitute your own headcount and your own four population splits; the arithmetic does not change shape.

There is a break even inside those numbers worth knowing before the account team reaches it. E3 at $36.00 plus E5 Security at $12.00 plus E5 Compliance at $12.00 is $60.00. E5 is $57.00. For a user who genuinely needs both add ons, E5 is $3.00 a month cheaper and the upsell is honest. For a user who needs one, $48.00 beats $57.00 by $9.00 a month, which on 4,000 users is $432,000 a year. The E5 pitch lives or dies on how many of your users need the second add on, and that is a question your own tenant answers: Microsoft 365 admin center, Reports, Usage, Microsoft 365 Apps for the 90 day feature picture, and the Microsoft Purview compliance portal for whether the compliance surface is switched on at all.

Security suite

The security stack is where an E5 conversation restarts after it has been lost on productivity. Seven SKUs get presented as one stack. Four of them, Defender for Endpoint, Defender for Identity, Defender for Office and Defender for Cloud Apps, sit inside the $12.00 E5 Security add on alongside Entra ID P2. The other three do not. Defender for Cloud and Microsoft Sentinel are Azure consumption products billed on usage, and the Intune Suite is a separate add on at $10.00 per user per month. Buying E5 Security does not buy any of the three, and the pitch rarely makes that distinction.

• Defender for Endpoint. Endpoint detection and response.
• Defender for Identity. Identity detection and response.
• Defender for Office. Email and collaboration security.
• Defender for Cloud Apps. SaaS security.
• Defender for Cloud. Azure security posture management.
• Sentinel. SIEM platform.
• Intune Suite. Unified endpoint management.

Put a published price against the biggest component and the bundle test becomes arithmetic. Defender for Endpoint P2 is $5.20 per user per month standalone, which is 43 percent of the $12.00 add on for one of five included products. A population running three or four of the five is buying well. A population running Defender for Endpoint and nothing else is paying $12.00 for a $5.20 product, a premium of about 131 percent, and should buy the standalone SKU. The population split matters more than the price: E5 Security across all 20,000 seats is 20,000 x $12.00 x 12 = $2,880,000 a year, against $576,000 on the 4,000 seats modeled above. Read the Microsoft security licensing unbundled analysis for the SKU by SKU version.

Power Platform

Power BI carried a published increase that most EA renewals in 2025 and 2026 absorbed without comment. Effective 1 April 2025, Power BI Pro moved from $10.00 to $14.00 per user per month and Power BI Premium Per User from $20.00 to $24.00. Pro had not moved since 2015. A 40 percent increase on the Pro line is a real number with a real effective date, which is more than can be said for most of what gets called the 2026 reset.

Two analyst populations

Premium Per User buys paginated reports, larger models, deployment pipelines and more frequent refresh. Across the estates we have measured, 10 to 20 percent of the analyst base uses any of it. The rest publish and consume, which is Pro.

$204,000 against $576,000 is 35 percent of the Power BI line, taken out before anyone mentions a discount. Power Apps and Power Automate deserve the same treatment: per app licensing for the populations that touch one application, per user only where the usage log shows more.

EA terms

Eight protections carry the risk in an EA, and none of them appears on the price sheet. Renewals concentrate on price because price is what the account team is compensated on, so terms drift one cycle at a time. The drift column below is what we find when we read the expiring enrollment against the one before it.

The eight EA terms protections, and how each one drifts

Element	What it controls	Common drift
Audit posture	Publisher audit cadence across the term	Audit cadence widens to annual
Audit data scope	Data collection during audits	Publisher dictated data set
Renewal data sharing	Data sharing at renewal	Publisher gets full estate view, customer gets nothing
Price protection	Per user pricing flexibility	Removed under NCE drift
SKU substitution	Rights to swap SKUs across the term	Restricted or removed
Cancellation	Early termination rights	Penalties expand under NCE
Data localization	Residency against regulatory requirements	Weak or generic language
Exit and data extraction	Data extraction and operational handover	Often absent altogether

Put all eight in as standard buyer side language in the first paper you send, not as asks in the last week. An amendment requested in week one is drafting. The same amendment requested in the final fortnight is a concession, and it gets priced as one.

EA vs MCA vs CSP: when the EA stops being the cheapest route

The EA is not automatically the cheapest route once you pass 500 seats, and the day one discount is the wrong number to decide on. Model the real twelve month trajectory under each program, including true ups and minimums, then decide on total cost across the term and on flexibility. The Microsoft Customer Agreement and the Cloud Solution Provider program both carry monthly flexibility the EA does not have.

The EA suits estates that are stable, growing, and on premises heavy. It is a poor fit for organizations that are consolidating, divesting, or moving fast to consumption based cloud, where the absence of a true down quietly costs real money for three years. Below 500 users or devices the question does not arise: CSP is the route.

Price at least one alternative at every renewal whether or not you intend to move. A costed MCA or CSP scenario is the only thing that makes the discount tier conversation move, so the modeling pays for itself even when you stay on the EA.

The renewal timeline and the levers that move the bill

Start twelve to eighteen months before the renewal date. Ninety days is not preparation, it is acceptance: too late to reconcile committed seats to real headcount, too late to test a price level band change, and far too late to stand up a credible alternative before the first quote lands. Microsoft's preferred sixty day intensive negotiation exists because it produces Microsoft's preferred outcome.

The renewal anniversary is the only window in which seats come down. A late start does not cost you a few points on price. It locks in three more years of ratcheted baseline.

Seven levers move the number and they compound, so work them together rather than one at a time: baseline reset back to true headcount, the price level band test, a written price hold on core seats for the full term, mix discipline that keeps Copilot and Defender in their own negotiations, competitive tension from a costed alternative, timing against Microsoft quarter and fiscal year end pressure, and term structure weighing one year flexibility against three year price. Verbal assurances on future SKUs do not survive a renewal, so each of these has to be written into the agreement to hold.

What to do next

1. Start twelve to eighteen months out, and time the close. Run the renewal as a structured buyer side program, and land the decision against Microsoft quarter and fiscal year end pressure rather than against your own deadline.
2. Reset the baseline before you price anything. Reconcile committed seats to real current headcount. The renewal anniversary is the only window in which the count comes down.
3. Test the price level band, then lock the tier protections. Confirm the true qualified user count against the band edges. Crossing a band can beat the discount you were about to accept. Hold the tier protections across the full term.
4. Unbundle the publisher's preferred unified package across the eleven underlying commercial elements, and put a costed MCA or CSP scenario on the table while you do it. Competitive tension is the only thing that moves the tier conversation.
5. Run the 2026 pricing reset against SKU rationalization. Compensate for the reset with mix improvements, and anchor the M365 population to the tiered model: defined E5, E5 Security or Compliance, and E3 populations.
6. Unbundle the NCE conversion against per user pricing flexibility, SKU substitution, and cancellation terms. Get the price hold on core seats in writing so true up growth bills at your rate.
7. Hold mix discipline on Copilot and Defender. Anchor the Copilot population to the actual productivity deployment plan, with discount tiers at population thresholds. Attach is how the account team recovers a concession made on core seats.
8. Assess each security suite SKU as a standalone proposition against deployment readiness.
9. Unbundle the Azure MACC from the EA renewal package. Three or four year horizon at the consumption baseline plus measured growth.
10. Hold the terms, and the term structure. Take the eight protections as standard buyer side language, add the true up reduction right, and weigh one year flexibility against three year price.
11. Engage independent buyer side support. Contact our Microsoft advisory practice for the renewal cycle scoping.

Frequently asked questions

How early should EA renewal preparation start?

Twelve to eighteen months before the renewal date. The publisher's preferred sixty day intensive negotiation is structured to deliver the publisher's preferred outcome. The structured cadence reverses that dynamic.

What is the typical EA renewal save?

Twenty to thirty percent run rate improvement against the publisher's first renewal package at the upper customer scale. The improvement compounds across SKU rationalization, discount tier protections, MACC unbundling, and the terms framework.

Should Copilot be included in the EA?

Depends on the population segmentation. For an eligible population of twenty to forty percent of the enterprise, EA inclusion is typically the right answer. For broader populations, the unbundled standalone Copilot subscription or a third party AI alternative is typically more cost effective.

What is the right MACC size?

The actual consumption baseline plus a measured growth assumption of fifteen to twenty five percent. The publisher's preferred MACC is typically materially above the realized consumption.

Can the discount tiers be re-negotiated?

Yes, but only with a credible alternative scenario. The MCA and CSP comparison is the load bearing leverage on the discount tier conversation.

Does the 2026 reset apply to existing EA terms?

No. The reset applies at the next renewal anniversary. Existing EA terms remain at the contracted rates until the renewal date, which is why the renewal cycle is the moment to unbundle the package.

How many seats do you need for a Microsoft Enterprise Agreement?

An Enterprise Agreement generally requires at least 500 users or devices. Below that threshold, CSP is the appropriate route.

What is a Microsoft EA price level?

A price level is a volume band, A through D, that sets the per seat rate. Level A starts at 500 seats and Level D begins at 15,000. Crossing a band at renewal can beat a negotiated discount.

Can I reduce Microsoft EA seats mid term?

No. The EA only allows reductions at the renewal anniversary. Seats added during the term cannot be removed until then, which is why baselines drift above headcount.

What is Software Assurance in an EA?

Software Assurance bundles new version rights, home use, roaming and disaster recovery use rights, and planning and training days into the EA. Most buyers use a fraction of it, so inventory the rights before renewal and either use them or price them into the deal.

Is the Enterprise Agreement always the cheapest option?

No. For shrinking, variable or cloud heavy estates a Microsoft Customer Agreement or a Cloud Solution Provider arrangement can beat the EA, so price at least one alternative at every renewal.

How do I protect against mid term price increases?

Negotiate a written price hold on core seats for the full term. Verbal assurances on future SKUs do not survive a renewal, so the protection has to sit in the agreement.

The detail sits in the Microsoft EA Renewal Playbook, and the engagement model in the Microsoft advisory practice. Read the related US professional services EA renewal case study and the Brazilian bank EA renewal case study.