A discount you only receive if you agree to a second, uncapped bill is not a discount. It is a billing model swap with a coupon stapled to the front.
Microsoft told its sales force this week that it will authorize discounts of 30% to 50% on Copilot subscriptions for large enterprise customers, starting as early as October. That got reported almost everywhere as a price cut, which is how Microsoft would like it reported.
The condition attached to it did not travel nearly as far. The bigger discounts are available to customers who also commit to usage-based fees on certain features, on top of the seats they are already buying. Strip out the headline percentage and what Microsoft is actually selling is a migration: off a fixed per-seat line item, onto a metered one, with the coupon as the inducement to move.
That distinction is the whole story, and it decides whether your bill goes down or up.
What You Are Actually Being Offered
The seat discounts scale with volume, and the tiers are specific:
- Roughly 30% off for customers buying between 1,000 and 10,000 Copilot seats.
- Up to 50% off for customers committing to more than 10,000 seats.
- Both tiers conditioned on accepting usage-based charges for specific features inside the product.
- Timed to land alongside a rebuilt Copilot application that folds Microsoft’s scattered AI tools into [one surface](https://analyticsindiamag.com/ai-news/microsoft-offers-copilot-discounts-of-up-to-50-as-it-prepares-ai-super-app).
The metering half is not speculative, and it is not new. Microsoft has been building the rails in public for months. Copilot Credits are documented, priced and live: pay-as-you-go runs at a cent per credit, with a prepaid annual pool available at a discount and a reserved-capacity tier offering as much as 40% off for organizations that pre-commit to a minimum monthly compute spend.
Read that sentence again with a procurement brain on.
Microsoft now sells you a discount for committing to seats, a second discount for committing to consumption, and a third for pre-buying the compute in advance. Each one is real. Each one also moves another slice of your spend from a number you can forecast to a number that depends on how enthusiastically your staff use a product Microsoft is about to make much easier to use.
Run the Arithmetic Before You Sign
Here is the calculation nobody published this week, and it takes about ninety seconds.
Microsoft 365 Copilot lists at $30 per user per month on an annual commitment, sitting on top of a qualifying base license. Take a company with 10,000 seats, which is exactly the threshold where the 50% tier unlocks. At list, that is $300,000 a month, or $3.6 million a year.
Halve it and you save $1.8 million a year.
Now price the other side of the trade. At a cent per credit, $1.8 million buys 180 million credits. Spread across 10,000 seats over twelve months, that is 18,000 credits per seat per year, or about 1,500 credits per seat per month.
So the breakeven is roughly 1,500 credits per user per month. Below it, the deal is genuinely cheaper. Above it, the 50% discount has been fully consumed and you are paying more than list for the same software.
That is our arithmetic on published list prices, not Microsoft’s, and your actual credit consumption per task is the variable that decides it. Which is the point. Microsoft knows the consumption curve for these features across its installed base. You do not, and neither does your CFO, and the contract asks you to commit before you find out.
The behavior that determines your bill is also the behavior Microsoft is actively engineering upward. A consolidated application that puts every AI tool one click away is a good product decision and a consumption accelerant at the same time, and those are not in tension for the vendor.
There is precedent, and it is not ambiguous. GitHub Copilot moved to usage-based billing with a similar shape: a familiar seat price, a credit allowance, overage past it. Copilot Cowork went metered per compute-consuming task in June. Advisors who negotiate these contracts for a living have been telling clients to model seats and consumption as two separate budgets since well before this week’s announcement.
None of that is hidden. It is simply not what the discount headline says.
BusinessTech’s position is that this was reported the way the vendor framed it, and the framing does not survive contact with the terms. Microsoft is not cutting the price of Copilot. It is buying seat commitments with margin it expects to recover through metering, and it is doing so at precisely the moment it ships a product designed to raise the metered number. That is a legitimate commercial strategy. It is not a price cut, and calling it one does a disservice to every procurement team that reads the headline and assumes its renewal just got cheaper.
The responsibility for the framing sits with Microsoft’s go-to-market organization, which briefed a percentage and let the condition ride along as a subordinate clause. The responsibility for falling for it will sit with whoever signs.
What should happen next is dull and specific. Ask Microsoft for per-feature credit consumption data from comparable deployments before you accept the usage commitment. Insist on a cap, or a look-back, or a renegotiation trigger tied to actual burn. Treat the 50% as what it is, which is the price Microsoft is willing to pay for a multi-year consumption annuity.
If the consumption numbers were flattering to the customer, they would already be in the press release next to the discount.