The negotiation of the MACC – the multi-year Azure spending commitment that Microsoft asks organizations to sign – is becoming one of the most delicate moments of contract renewal, just as the rules of the game are changing. Who sits at the table without their own data risks signing a commitment inflated on incorrect projections: this article explains why and how to avoid it.
In the last eighteen months, the world of Microsoft licensing has evolved more rapidly than in the last decade. Who manages EA renewals perceives it clearly: conversations at the bargaining table are changing compared to ten years ago.
Microsoft sales teams are incentivized on AI and cloud revenues and during commercial negotiations we increasingly find topics such as Copilot and Azure spend which until recently were separate discussions. They are no longer. There is a thread that connects everything, and it is important to understand it before sitting down to negotiate.
From May 1st, 2026, Microsoft introduces M365 E7 "The Frontier Suite" at 99 dollars per user per month. It is a significant jump compared to the current E3/E5 architecture and it is not a simple functional update.
The bundle includes Copilot – which alone has not had the hoped-for market penetration – but above all it includes Agent 365, the module that allows creating, orchestrating, and governing AI agents within business processes. And it is here that the conversation on prices changes nature. To use agents, the most advanced and strategically relevant part of the suite, credits are required, some of which are based on the underlying Azure infrastructure with consumption-based billing.
Ogni agente attivato, ogni workflow automatizzato, ogni task delegato all’AI genera consumo cloud misurabile. AI e cloud non sono più due conversazioni separate: sono la stessa conversazione.
Agents are the consumption engine most difficult to estimate, but they are not the only one.
Qualsiasi workload Azure – database, macchine virtuali, storage, analytics – genera spesa che si accumula nel tempo. È su questa spesa complessiva che Microsoft chiede alle organizzazioni di prendere impegni formali. Si chiamano MACC – Microsoft Azure Consumption Commitment – e funzionano così: ci si impegna su un volume di consumo Azure annuale, in cambio di uno sconto sulle tariffe standard. Quello sconto – tecnicamente chiamato Azure Commitment Discount – viene applicato mese per mese sui servizi effettivamente consumati durante il periodo di validità dell’accordo.
È un beneficio reale, ma con una caratteristica importante: Microsoft non ha alcun obbligo contrattuale di replicarlo nei cicli successivi. Ogni sconto è specifico all’accordo che lo ha generato.
In un contesto in cui il consumo cloud è sempre più legato all’adozione di tecnologie AI – agenti, automazioni, workflow – la cui diffusione reale all’interno delle organizzazioni è difficile da prevedere con precisione, la dimensione di quell’impegno di spesa diventa una decisione strategica, non solo commerciale. Un commitment Azure sovradimensionato non è solo un costo finanziario: è un segnale che l’organizzazione sta pagando oggi per un futuro che non sa ancora se avrà la forza, in termini di risorse e volontà organizzativa, di costruire.
All this fits into a contractual framework that is itself evolving and the direction is clear. Microsoft is progressively moving organizations towards the Microsoft Customer Agreement (MCA), with the declared objective of having a single contractual framework across all industries and sizes.
It is no longer a question of if, but of when. In Italy, a direct market, the process is already underway: in 2025, EA contracts mainly composed of cloud services and the Level A segment were involved, i.e., organizations up to 2400 license points (therefore roughly 2400 users), as read in the Microsoft Partner blog. This allows hypothesizing that in the coming years the perimeter will also extend to larger companies.
The difference compared to the EA is not only administrative, it is structural. The EA guaranteed a price lock for the entire three-year duration and discounts negotiated upstream, regardless of consumption growth. A position conquered in negotiation remained valid for three years. In MCA this logic is completely different: the discount is not an acquired right, but must be earned by demonstrating year-on-year consumption growth. A discount obtained in a previous cycle is not carried over, it must be re-justified with a credible growth narrative and concrete data.
Microsoft has already removed structural discounts in the EA in November 2025, with cost increases of up to 30% on some segments: in MCA, that protection does not exist by definition. According to some unofficial market sources close to Microsoft, starting from March 2nd, 2026, all new MACC contracts and renewed ones – both in direct and indirect markets – must be signed within the MCA-E. If it is confirmed that the MACC will no longer be anchored to the EA, it is a signal that the transition process is not gradual: it is already underway, and concerns anyone who has a renewal on the agenda in the coming months.
In this context, the trend of Microsoft sales teams could be to propose aggressive growth curves to justify contract discounts for high commitments. Ambitious AI adoption scenarios, agent usage estimates based on optimistic industry benchmarks, cloud consumption projections could assume an organizational maturity that often does not correspond to the reality of the first 18 months of deployment.
The commitment must be a floor, not a ceiling. The starting point is the real historical consumption, with a documented and conservative growth margin, typically 10-15% per year on projects already started and budgeted. Everything that goes beyond that threshold should be structured as an option, not an obligation.
The problem is that almost every EA environment accumulates over time inefficiencies difficult to see from the outside: unused resources, forgotten subscriptions, oversized workloads. But there is a further level of complexity that rarely emerges spontaneously at the bargaining table: not all historical Azure spend counts in the same way.
Support plans, some types of credits, and Marketplace offerings not explicitly designated in the agreement are typically excluded from the count of Eligible Services. If the commitment is built on a raw reading of past consumption without verifying the composition of that spend, one risks committing to an inflated base, paying for years for something that would never have contributed to reaching the threshold.
To this is added a concrete consequence that is worth knowing in detail. If at the end of the period the consumption of Eligible Services does not reach the commitment, Microsoft issues a Shortfall Invoice (a balance) for the difference. Those funds are not lost in a strict sense: in most cases they are converted into credit spendable on Azure in the following months.
The logic is that the customer had already made a financial commitment – if they do not reach it with real consumption, they complete it with a payment that remains however within the Microsoft ecosystem. It is better than a flat penalty, but only if the organization actually intends to continue consuming Azure at those levels. If the cloud footprint is reducing, or if the spending was inflated by inefficiencies meanwhile resolved, that prepayment becomes a constraint, money already paid that one is forced to spend on a platform regardless of real needs.
The most critical case remains that of non-renewal: if the agreement expires without a new Enrollment, the credit is not applied and the customer pays the delta without receiving anything in return.
Finally, the share of projected consumption that depends on the adoption of the Frontier Suite must be explicitly quantified. The license cost is known – 99 dollars per user per month – therefore plannable. What is not are the Azure consumption scenarios generated by the use of Agent 365: how many agents will actually be activated, with what frequency, on which processes. It is this component that must be separated and treated as a variable – not as a base for the commitment.
The Azure services that come into play in the execution of agents are those that generate measurable consumption and potentially eligible for the MACC. But their real incidence depends entirely on how many agents are deployed and with what intensity of use: it is exactly the type of variable that Microsoft sales teams tend to estimate optimistically and that instead must be treated with an explicit margin of uncertainty.
To this is added a second level of uncertainty: it is not taken for granted that all these services actually fall among the Eligible Services of the MACC. The list of what counts towards the commitment is defined contract by contract, and asking your Microsoft counterpart to put it in writing is one of the steps that is rarely taken and that instead can significantly change the mathematics of the negotiation.
Knowing in advance on which fronts the discussion will develop is already an advantage. The topics on the table are essentially two: the estimation of consumption scenarios related to AI agents and the analysis of historical Azure consumption data. Different fronts, but both require the same premise: having your own data, independent of those that Microsoft will present.
On the historical front, the starting point is the native Microsoft tools – Azure Cost Management, EA or MCA billing reports – which allow reconstructing consumption by service, subscription, and period. A complete analysis should also include SaaS spending related to M365: not because it contributes directly to the MACC, but because it determines the adoption context on which the growth of Azure consumption depends.
Knowing how many M365 licenses are active, which features are actually used, and at what point the Copilot adoption is, is indispensable to credibly estimate how much cloud consumption the Frontier Suite will generate in the next twelve-to-twenty-four months. The same exercise serves to identify E3 or E5 plans oversized compared to real use: optimizing them before negotiation is not just a saving operation, but a way to face negotiations with a clean baseline – and with recovered budget that can be reallocated more consciously on the Azure commitment.
The problem emerges when trying to estimate the future commitment having the historical series built on the EA.
In the EA, spending is aggregated at the enrollment level, with an annual true-up cycle and prices that already incorporate programmed discounts by level. In MCA this structure no longer exists: in its place there are billing accounts, billing profiles, and invoice sections, with monthly billing and prices that start from the list price. The reports of the two frameworks have different columns, different aggregation logics, different pricing logics. The risk is to arrive at the negotiation with a historical series that, besides already having some waste incorporated, is not readable in the logic of the new contract and to build a commitment on data that do not reflect how the spending will be measured in MCA.
Native Microsoft tools are not sufficient for this work. Azure Cost Management only sees the Microsoft perimeter, does not offer a normalization layer between EA and MCA – historical data remains in one silo, those of the new framework in another. The allocation of costs depends entirely on the quality of tags on resources, which in complex environments is often incomplete. And on the commitment front, native tools show the current consumption compared to the threshold, but do not offer structured forecasting to estimate if and when that threshold will be reached.
FOCUS, the open standard for cloud cost normalization developed by the FinOps Foundation, fills this gap. It translates EA historical data into a common taxonomy with MCA, returning a clean baseline structured in the logic of the new framework.
The Cloud Cost Optimization technology of our partner Flexera supports FOCUS natively, combining data normalization with analysis, allocation, and forecasting capabilities that native Microsoft tools do not offer. It is the prerequisite for any reliable commitment estimate: it allows isolating the accumulated waste, understanding which services contribute to the commitment and which do not, and building consumption scenarios that reflect how the spending will actually be measured in MCA.
The initial commitment will still be an estimate – and it is normal for it to be so. The objective is not absolute precision at the time of signing, but building a governance that over time improves it. A structured FinOps taxonomy, aligned where appropriate with the TBM logic – the standard framework for translating IT costs into business language – putting IT, Procurement and Finance at the same table with a shared vision of spending – allows correctly allocating costs and monitoring deviations from the commitment.
With time, as MCA consumption data accumulate and governance consolidates, projections become more precise and the commitment can be refined with greater confidence, arriving at every renewal with own, verified data, difficult to challenge.
On the agent front, the estimate must be treated explicitly as a variable, not as a fact. What can be done is to define a reasoned range: a "floor" based on workloads already started, and a separate AI component, structured as a growth option with explicit thresholds not to be incorporated into the commitment base until real adoption justifies it.
The MACC negotiation tends to be treated as the last step of a commercial process already decided – a signature at the bottom of a conversation in which the numbers had already been written by someone else. It must not be like this.
Who arrives at the table with their own analysis of historical spending, an independent estimate of AI scenarios, and a clear position on what is Eligible and what is not, is bringing a counter-narrative based on data. And it is the only position from which it is possible to truly negotiate the structure of the commitment: the thresholds, the variable component, the discount renewal conditions.
The context is not favorable to those who arrive unprepared. The transition towards MCA removes the last structural protections that the EA guaranteed, and the uncertainty on AI adoption adds a variable that no industry benchmark can solve for you.
But the complexity is predictable and therefore manageable. The fronts of the discussion are known, the tools to analyze them exist, and the logic to build a defensible commitment is the same that applies to any serious financial decision: starting from real data, isolating variables, not signing commitments that depend on a future that you do not control.
In WEGG we are following this evolution closely, as ITAM and Finops consultants. We have already developed some operational reflections on how to face the MACC negotiation in this new context and we will talk about it in the webinar scheduled for April 15th, entitled "From EA to MCA: that underestimated work before and after the signature that changes the governance of cloud costs".
If the topic concerns you, it is the right time to compare notes: here is the link to register for the webinar.
Insights
OUR OFFICES
OUR OFFICES
PADUA
Via Arnaldo Fusinato 42, 35137
MILAN
Viale Enrico Forlanini 23, 20134
ROME
Viale Giorgio Ribotta 11, 00144
Copyright © 2025 WEGG S.r.l. • P.I 03447430285 • C.F. 02371140233 • REA 311023
Azienda Certificata ISO 9001:2015 – ITA / ISO 9001:2015 – EN
