Microsoft Enterprise Agreement Guide 2026

You are registered. Your guide is ready. Read the full 2026 edition of the Microsoft Enterprise Agreement Guide below.

Prepared by Atonement Licensing · buyer-side advisory · last reviewed June 2026. Figures are list-level or clearly labelled indicative ranges. The 12,000-seat estate used below is a representative benchmark scenario for illustration, not a quote.

Executive summary

The Microsoft Enterprise Agreement is the most consequential software contract most enterprises sign, and Microsoft's proposal for it is built around Microsoft's sales targets, not your utilisation. The gap between the two is typically 15 to 30 percent of contract value. It hides in the edition mix, the commitment quantities, and the Azure structure, not in the headline discount, which is why buyers who negotiate only the percentage leave most of the money behind.

On a representative 12,000-seat estate, an opening proposal that standardises the base on E5 and deploys Copilot broadly models near $10.8M per year. The same headcount, right-sized across E5, E3, and F3 with Copilot held to the credible population, models near $6.4M per year — a Year 1 difference of roughly $4.4M on the same people. The decision in front of procurement is which populations sit on which plan, how far to commit on Copilot, how to size the Azure commitment, and which contract protections to place before signature.

This guide explains how the EA is constructed and where that gap lives: the enrollment architecture and the Microsoft Product Terms that govern use rights, the Azure commitment instruments and how a MACC is consumed and discounted, the Microsoft 365 right-sizing work that funds everything else, the Copilot decision now attached to every renewal, the 180 day renewal playbook, the compliance exposures that surface as renewal pressure, and the standardise-versus-diversify question that decides your long-term negotiating position. Read this before your next anniversary, not after it.

Microsoft EA architecture: agreement types, Product Terms, and the commercial framework

An Enterprise Agreement is a three-year commitment available to organisations above roughly 500 users or devices. It is built from enrollments: most commonly the Enterprise Enrollment for organisation-wide products, Microsoft 365, Windows, Office, and the Server and Cloud Enrollment, SCE, for server workloads and Azure. The defining feature is the organisation-wide commitment: certain products must be licensed across the entire qualified user or device base, which is exactly the mechanism that pushes spend upward when Microsoft proposes a higher edition for consistency.

Use rights do not live in the EA itself. They live in the Microsoft Product Terms, the consolidated successor to the old Product Use Rights document, published and updated monthly on Microsoft's licensing site. The EA's commercial paper sets quantities, prices, and custom amendments; the Product Terms decide what a licence actually permits, from virtualisation rights to outsourcing rules. Any negotiated protection that matters, price holds, transition rights, audit terms, must be written into the agreement or an amendment, because the Product Terms version can change underneath you during the term.

How the price is actually built

EA pricing is layered: a published list price, a programmatic price level set by size band, and negotiated discounts on top. The critical insight is that the price level is far less negotiable than the edition mix and the commitment quantities. Most overspend is not a failure to win another discount point; it is agreeing to license E5 organisation-wide when a segmented mix of E3 and E5 matches actual need. A buyer who wins 5 extra points of discount on the wrong mix still pays more than a buyer at standard discount on the right mix.

Custom amendments worth writing

The EA's real value to a sophisticated buyer is that it accepts custom amendments, and the ones that matter are predictable. A price hold covering the full SKU list for the term, not just the products in the initial order, so that mid-term additions price at the negotiated level. A cap on the renewal uplift, expressed as a percentage, agreed three years before anyone knows what list prices will do. Defined merger, acquisition, and divestiture treatment, so a corporate event does not strand licences or trigger a renegotiation from weakness. An audit standstill during renewal windows. And explicit transition terms if Microsoft proposes moving the relationship from EA paper to MCA-E, including carriage of negotiated discounts onto the new vehicle.

None of these appear in a default proposal, all of them have been signed for buyers who asked, and every one of them is cheaper to obtain at signature than to request mid-term. The negotiation habit that pays most is simple: every commercial understanding the account team offers verbally goes into the agreement or an amendment, or it does not exist.

Action. Before any EA event, pull the current Product Terms, decompose Microsoft's proposal by SKU, and build a usage-led counter-mix. The mix is the negotiation.

Azure cost optimisation: MACC, reservations, savings plans, and Hybrid Benefit

Azure inside an EA is typically committed through a Microsoft Azure Consumption Commitment, a MACC: a dollar amount you agree to consume over the term in exchange for discounting and access to commitment-decrement benefits. The bargaining power is real, Microsoft discounts committed Azure aggressively, but so is the risk: an over-sized MACC becomes a treadmill, pushing workloads to Azure to burn commitment rather than because the economics favour it.

Three structural levers sit beneath the MACC and work regardless of your discount. Reservations commit specific resource families for one or three years at reductions that can reach a large share of on-demand cost. Azure savings plans commit an hourly spend across compute with more flexibility and somewhat shallower discounts. Azure Hybrid Benefit applies existing Windows Server and SQL Server licences with Software Assurance to Azure workloads, removing the licence component from the meter. Most enterprises can cut 30 to 45 percent from steady-state Azure spend by structuring these three correctly, before asking Microsoft for anything.

Negotiating the commitment itself

Size the MACC to a conservative forecast, not to Microsoft's growth model. Then negotiate the failure modes: carry-forward or term extension for unconsumed commitment, a mid-term checkpoint with the right to rebase downward, and clarity on what counts as consumption, including Marketplace and Azure OpenAI usage. A slightly worse discount on a commitment you will actually consume beats a deeper discount on one you will not. Consumption risk is the quiet cost of every cloud commitment.

Consumption risk and the commitment treadmill

Unconsumed commitment is the quiet failure mode of every MACC. The pattern repeats: a three-year commitment sized to an optimistic migration plan, a delayed programme, and then a final year spent moving workloads to Azure not because the economics favour it but because the dollars are already promised. At that point the commitment is steering the architecture, which is precisely backwards.

The defences are contractual and operational. Contractually: a mid-term checkpoint with a documented right to rebase the commitment downward, carry-forward or extension language for unconsumed amounts, and clarity that Marketplace and Azure OpenAI consumption decrement the total. Operationally: a monthly consumption review against the burn-down curve, owned by FinOps, with escalation the moment the curve diverges. A commitment that is tracked monthly gets renegotiated from evidence; one that is discovered in month 30 gets absorbed as loss.

Action. Build the reservation, savings plan, and Hybrid Benefit structure first, then size the MACC on the optimised run rate. Committing to unoptimised spend locks in your own waste.

Microsoft 365 licence rationalisation: seat audits, SKU benchmarking, and true-up preparation

The single largest recurring saving in most EAs is matching the M365 edition to the role. Microsoft markets E5 as the default, but its premium value, advanced security, compliance, analytics, and voice, is genuinely consumed by a minority of users in most organisations. Knowledge workers are placed on E5 when E3 plus a targeted security add-on delivers the same protection at lower cost, and frontline populations sit on E3 when an F-series licence fits the role.

Right-sizing is a data exercise, not an opinion. Pull actual feature usage from the Microsoft 365 admin center and the security and compliance portals, segment users by role, and license to the role rather than to the organisation-wide default. The savings are durable because they reduce the committed baseline, not just a one-year order. The bar chart below shows where over-licensing concentrates, expressed as indicative per-seat savings ranges.

True-ups and the annual order

The EA's signature mechanic is the annual true-up: deploy what you need during the year, report the net increase annually, and pay for it. There is no true-down within the term. Quantities only ratchet upward, which is why the initial baseline and the edition mix matter so much. Over-commit at signing and you carry the cost for three years; under-baseline and you simply true-up to reality without penalty. Treat that asymmetry as a planning tool: baseline conservatively to genuine current need, deploy growth as it occurs, and reconcile annually.

Run the true-up on your calendar, not Microsoft's. Ninety days before each anniversary, audit seat assignment against activity, reclaim and reassign inactive seats, and restructure the SKU mix before the order is placed. A true-up prepared in the final week accepts whatever the account team reports; one prepared a quarter out reports your own numbers. Done annually, this routinely saves 15 to 25 percent on the M365 component alone.

Action. Stand up a quarterly licence governance review: usage telemetry, seat reclaim, SKU mix. The renewal then inherits a defensible baseline instead of three years of drift.

Copilot and AI licensing inside the EA

Every EA renewal since 2024 has carried a Copilot conversation, and the commercial mechanics deserve their own discipline. Microsoft 365 Copilot lists at 30 dollars per user per month on an annual commitment and requires a qualifying base licence, so the loaded cost per enabled seat depends on your existing mix; populations below E3 carry a base uplift on top of the add-on. Seats added mid-term co-term to the EA end date, and quantities follow the same one-way ratchet as every other online service in the agreement.

Negotiate Copilot as structure, not as a price: a wave-based ramp tied to adoption gates, a written price hold across the term, reduction rights at anniversary, and usage minimums rather than whole-workforce commitments. Track the usage data Microsoft collects on your tenant, because adoption telemetry will be the seller's renewal argument, and it should be yours instead. Where AI workloads are process-specific rather than broad productivity, price Azure OpenAI consumption against per-seat licensing; Azure OpenAI decrements a MACC while Copilot seats do not, which can decide the question by itself. The full analysis, including agent metering through Copilot Studio and the governance prerequisites, is in our Microsoft Copilot Licensing Guide.

Budget the loaded cost honestly. The add-on is 360 dollars per seat per year at list before any base uplift, agent consumption through Copilot Studio meters separately, and adoption telemetry shows most estates settle well below full active use. A Copilot line item built on the headline price and the full workforce is wrong in both directions at once: it overstates the population that should be enabled and understates the cost of each seat that is. Wave-based budgets, reviewed quarterly against active-use data, keep the spend tied to demonstrated value, and they give the renewal negotiation a factual record instead of a vendor narrative.

Action. Bring the Copilot volume decision to the renewal table as one negotiation with the base estate and Azure, and paper the ramp before enabling wave one.

The EA renewal is no longer a discount negotiation. It is a population and lever negotiation, and that is a contest the prepared buyer wins.

EA renewal negotiation: the 180 day playbook

Negotiating power at an EA renewal is built in the six months before the date, not at the table. By the time Microsoft presents a renewal proposal, the prepared buyer already holds an independent usage baseline, a benchmarked target price, and at least one credible alternative for a visible slice of the estate. This is the timeline we run.

Use creation: where renewal advantage comes from

Microsoft's renewal position rests on switching cost. Yours rests on demonstrated options. The alternatives that move a Microsoft number are specific and verifiable: a costed Google Workspace evaluation for a defined population, workload portability to AWS or Google Cloud for the next platform decision, third-party support or extended security update strategies for ageing server estates, and a competing CRM or ITSM platform kept genuinely in play. None of these need to be deployed to work; they need to be credible, costed, and visible.

Reading the discount

Renewal proposals are packaged to make the discount look larger than it is. The standard moves: quoting the percentage against a list price that was just raised, bundling one-time transition funds into the per-seat comparison, deepening the discount on products you barely use while holding firm on the ones that dominate your spend, and presenting an E5 step-up as a discount because the per-feature maths looks favourable against products you never intended to buy separately. None of this is improper; it is packaging, and it is defeated by one habit.

Normalise everything to annual cost per user for your actual mix, over the full term, with one-time funds stripped out and shown separately. Compare that number, and only that number, against your current spend and your benchmark. When the account team says a concession is exceptional, ask for it expressed in those terms. The conversations get shorter and the numbers get better.

Non-price terms are where renewals pay for decades: price holds on the full SKU list for the term, written caps on renewal uplift, true-down rights at renewal explicitly preserved, MACC rebase checkpoints, an audit standstill during negotiation windows, and transition assistance if Microsoft moves you from EA paper to MCA-E. The escalation path matters too. The account team owns the relationship, but exception pricing and custom terms are approved at the deal desk and above; ask in writing for that review when the field stalls. Quarter ends in September, December, March, and the June 30 fiscal year end remain the moments of maximum flexibility.

Action. Start at day 180, anchor with a written term sheet, and time the close to a Microsoft quarter end. A renewal that starts at day 60 is negotiated on Microsoft's information.

Microsoft compliance and true-up risk: LTSC, server licensing, and audit-ready estates

Microsoft compliance risk rarely arrives as a formal audit letter. It arrives as a friendly software asset management engagement, a partner-led review, or a set of pointed questions during renewal, and its findings convert directly into renewal pressure. The defensible posture is the same either way: know your own estate better than Microsoft does, and remediate on your schedule rather than theirs.

The recurring exposure areas are specific. Windows Server and SQL Server core licensing, where virtualisation, cluster mobility, and licence mobility rules are routinely misapplied. Client access licences for on-premises workloads that survived the cloud migration. Long-Term Servicing Channel builds deployed beyond their intended device classes. Azure Hybrid Benefit claimed on licence counts without Software Assurance coverage to support them. And premium feature enablement, E5 security or compliance features switched on for populations licensed at E3, which telemetry surfaces to Microsoft with precision.

The audit-ready estate is a renewal asset, not just a defence. A buyer who opens the renewal with an independently verified deployment baseline removes the other side's best pressure tactic before it is used, and converts compliance from a liability conversation into proof of operational discipline. The same baseline doubles as the true-up record and the right-sizing evidence, so the work pays three times. Assign it a permanent owner in software asset management, refresh it quarterly, and require that every remediation is documented against the specific Product Terms rule it answers, so the file stands on its own in front of any reviewer.

Action. Run an internal compliance baseline at day 180 of the renewal timeline, fix what you find, and walk into the negotiation with nothing for the other side to discover.

Microsoft versus third-party SaaS: standardise, diversify, and negotiate both

The suite argument is real: bundle economics, integration, and one renewal event. So is its cost: every workload that standardises on Microsoft removes a competitor from your next negotiation, and the E5 features you adopt mid-term quietly displace the incumbent tools whose contracts were your negotiating tension. Standardisation is a procurement decision with a decade-long price tag, and it should be priced as one.

The buyers who hold the strongest Microsoft positions run a deliberate portfolio. They standardise where the suite advantage is overwhelming and the workload is undifferentiated, and they keep genuine alternatives alive where the market is competitive: collaboration versus Zoom and Slack, security versus the dedicated vendors, CRM versus Salesforce, ITSM versus ServiceNow, analytics versus the independents. The point is not to fragment the estate; it is to make sure that at every renewal at least some visible, costed share of spend is contestable.

Negotiate both sides simultaneously. The quarter your Microsoft EA renews is precisely the quarter to be in live commercial conversations with the third-party alternatives, because that is when their pricing is sharpest and Microsoft's attention is highest. A buyer who signs the Salesforce renewal in March and the EA in June has told both vendors they were never really in competition.

Action. Before the renewal, classify each major workload as standardise or contest. Fund a real evaluation for the contested set, and let Microsoft see the work.

Key takeaways

• The EA gap between proposal and need is typically 15 to 30 percent of contract value, and it lives in edition mix and quantities, not the headline discount.
• Use rights live in the Microsoft Product Terms. Anything that matters must be written into your agreement, not assumed from a sales conversation.
• Structure Azure first, reservations, savings plans, Hybrid Benefit, then size the MACC on the optimised run rate, with rebase rights in writing.
• Right-size M365 to the role with usage telemetry. The baseline you sign is the cost you carry for three years; there is no true-down mid-term.
• Bring Copilot to the renewal table as structure: ramp, price hold, reduction rights, usage minimums.
• Run the 180 day playbook: baseline, benchmark, alternatives, term sheet, escalation, quarter-end close.
• Find your compliance exposures before Microsoft does, and treat SAM engagements with audit discipline.

Frequently asked questions

What is the minimum size for a Microsoft Enterprise Agreement?

The EA is generally available to organisations with roughly 500 or more users or devices. Below that threshold, CSP and the Microsoft Customer Agreement are the usual routes. Above it, the EA's organisation-wide commitment and negotiated custom terms become available, and the choice between vehicles is itself negotiable.

Can we reduce Microsoft 365 licences during the EA term?

Not in the standard mechanics. Quantities ratchet upward through the annual true-up and there is no true-down within the term. Reductions happen at renewal, which is why the initial baseline and edition mix matter more than any single discount point, and why true-down rights preserved in writing at renewal are worth real money.

What discount should we expect at an EA renewal?

There is no standard number, and chasing one misreads the structure. The gap between Microsoft's proposal and a usage-led counterproposal is typically 15 to 30 percent of contract value, driven mostly by edition mix and commitment quantities. Benchmark your estate band and vehicle before responding to any proposal.

Do Marketplace purchases count toward an Azure MACC?

Eligible Azure Marketplace purchases decrement a MACC. Routing qualifying third-party software through the Marketplace is a legitimate way to consume a commitment you have already made, provided the offers meet Microsoft's published eligibility rules. Confirm eligibility per offer before counting on the decrement.

When should EA renewal preparation start?

One hundred eighty days before expiry at the latest. The first ninety days build the usage baseline, the benchmark, and the alternatives; the second ninety run the commercial negotiation and escalation. Renewals that start 60 days out are negotiated on Microsoft's information instead of yours.

This guide accompanies the Microsoft Enterprise Agreement Guide overview page. Related research: the Microsoft Copilot Licensing Guide 2026, the Microsoft EA Negotiation Playbook 2026, and the Microsoft Audit Defense Playbook 2026.