The Pre-Purchase Plan dangles up to 20 percent off. It also expires. Here is when committing up front pays, when pay as you go wins, and how to size a commit you will actually consume.
Microsoft sells Copilot Credits three ways, but the real decision is binary: stay flexible on pay as you go, or commit up front for a discount that tops out at 20 percent and expires. For most buyers, at most volumes, flexibility wins. This guide shows the exceptions, where Capacity Packs fit, and how to size a commit you will actually burn.
Microsoft documents three purchase routes in its Copilot Credits overview. They differ on commitment and flexibility, not on what a credit does. Every credit is still one cent of metered agentic work, pooled at the tenant.
The choice between them is a choice about risk. Pay as you go pushes all the risk onto Microsoft, who only get paid for real usage. Prepay moves the risk to you, in exchange for a discount. Capacity Packs sit in between.
One cent per credit, billed monthly in arrears for exactly what you consumed. No up front purchase, no expiry. It decrements your Azure commitment as you go, and Azure meters the spend so you can attribute it to departments. The cost is fully variable, which is its strength and its only weakness.
A fixed pack of 25,000 credits for 200 dollars per tenant per month, bought through the admin center. Credits reset every month rather than rolling over. Packs suit steady, predictable monthly volume that you would rather see as a fixed line than a pure variable.
An annual pool bought up front at a tiered discount, documented in Microsoft's Pre-Purchase Plan announcement. Exceed the pool and you fall back to pay as you go. Unused credits expire at term end, which is the whole risk in one sentence.
The three models at a glance
| Model | Commitment | Discount | Expiry risk |
|---|---|---|---|
| Pay as you go | None | None, list one cent | None |
| Capacity Packs | Monthly | Fixed pack price | Resets monthly |
| Pre-Purchase Plan | Annual, up front | 5 to 20 percent | Unused expires |
White Paper · Microsoft
What Microsoft Copilot Cowork Really Costs
What a task really costs in dollars, the prepay floor, and the same work on Claude direct. Read it free.
Prepay pays in a narrow band: high, proven, steady volume where even a single digit discount on a large base is real money, and where you are confident you will burn the pool. The discount only reaches double digits at very large commitments, so the case has to rest on scale, not on the headline rate.
Notice what is missing from that list: a vendor forecast. A forecast is not proof. The account team is paid to book committed revenue early, and a prepay does exactly that. Your interest is the opposite, to keep the dollars flexible until consumption is certain.
A prepay discount you do not fully consume is not a discount. It is a deposit you forfeit.
For most buyers, most of the time. Pay as you go wins whenever your volume is uncertain, seasonal, early stage, or modest, because it bills only for real usage and carries no expiry. It is also the only model that lets you optimize freely, because you are never holding a pool you have to drain.
Our buyer side default is pay as you go first, prepay only on proven volume. That is the opposite of the account team default, which is to lock revenue early. The pillar covers the broader meter in the Copilot Credits pillar, and the commitment interplay in the MACC guide.
Capacity Packs are the quiet middle option, and they suit a specific shape of demand: steady month to month, predictable, but not large enough to justify an annual lockup. Because they reset monthly, the worst case is a single month of partial waste, not a year of forfeited budget.
The trap with packs is buying too many. Each unused pack is 200 dollars of monthly waste, and packs do not roll over. Buy to the monthly floor and let pay as you go absorb the spikes.
If you do prepay, size to the floor of your demonstrated range, not the vendor projection. A slightly undersized pool that falls back to pay as you go beats an oversized pool that expires. The asymmetry is the whole point: overage is cheap, expiry is total loss.
Work a simple case. Say you have proven a steady 2.5 million credits a year. Tier 3 of the Pre-Purchase Plan buys 3 million credits at 7 percent off. The discount looks like 2,100 dollars saved. But you only need 2.5 million, so 500,000 credits expire unused, a 5,000 dollar write off that more than erases the saving.
The oversized commit that loses money
| Item | Amount | Effect |
|---|---|---|
| Proven annual volume | 2,500,000 credits | What you will actually use |
| Tier 3 pool, 7% off | 3,000,000 credits | Discount $2,100 |
| Unused at term end | 500,000 credits | Forfeited $5,000 |
| Net result | −$2,900 | Worse than pay as you go |
Size instead to a tier that sits at or below your proven floor, take the overage at pay as you go, and the math flips back in your favor. The discipline is the same one the whole cluster repeats: commit to what you have proven, never to what the vendor projects.
Run this before you sign any prepay.
Yes. The Pre-Purchase Plan tops out at 20 percent off list, and only at the deepest commitment tier of 300 million credits. Most buyers land at 5 to 10 percent. Below the top tier the discount is modest against the flexibility you give up.
No. Pre-Purchase Plan credits expire at the end of the one year term. Any pool you buy and do not consume is forfeited, which is why sizing on a vendor forecast is risky. Capacity Pack credits reset monthly rather than rolling over.
At list, pay as you go is one cent per credit and prepay shaves 5 to 20 percent off that. But pay as you go bills only for real usage, so it is usually cheaper in total than a prepay pool that overshoots and expires.
Only after pay as you go has demonstrated steady, predictable volume that clears the tier you are sizing to, with margin. Commit to proven consumption, never to a forecast. Size to the floor of your demonstrated range, not the ceiling.
You fall back to pay as you go at one cent per credit for the overage. There is no interruption, but you lose the prepay discount on everything above the pool. That makes a slightly undersized commit safer than an oversized one.
They are a middle path. A Capacity Pack buys 25,000 credits for 200 dollars per tenant per month and resets monthly, so it carries a monthly commitment rather than an annual one. It suits steady volume that you do not want billed as a pure variable, without the annual expiry risk of the Pre-Purchase Plan.
The task mix model in dollars, the credit to dollar conversion across light, medium, and heavy work, the build versus buy math against Claude direct, and the governance controls to set before you provision.
Used across more than five hundred enterprise engagements. Independent. Buyer side. Built for procurement leaders sizing Copilot inside the next EA renewal.
500+ enterprise clients. 11 vendor practices. We model your task mix in dollars, weigh direct Claude against buying through Microsoft, and set the governance and commitment terms before Microsoft sizes them for you.
One email a month on Copilot Credits, EA renewal levers, Azure commitment, and audit defense. Buyer side only.
