Under committing costs you a rate. Over committing costs you the whole gap.
Universal Credits are a prepaid annual commitment, which makes the number you sign a payment obligation rather than a budget. Almost all the negotiating attention goes to the discount and almost none to the commitment size, which is the wrong way round, because the two failure directions are not symmetric and the asymmetry points somewhere most buyers never look.
Prepared by Redress Compliance · August 10, 2026 · Oracle advisory. Based on 30 to 40 Universal Credits commitments modelled, 2023 to 2025.
Executive summary
The two failure directions cost completely different amounts, and that should drive the sizing.
Consume more than you committed and Oracle invoices the excess monthly in arrears at the rate card in your order, so the cost of under committing is the difference between your committed rate and your overage rate on the excess only.
Consume less and the unused credits are forfeited at the end of the credit period, so the cost of over committing is 100 percent of the gap. One error costs a margin, the other costs the whole amount.
It follows that the rational commitment sits below expected consumption rather than at it. That is the opposite of how these deals are usually sized, and it is not a risk preference argument, it is arithmetic.
If your forecast is uncertain in both directions, the expected cost of being 20 percent light is a fraction of the expected cost of being 20 percent heavy.
Signed commitments in our file ran 20 to 40 percent above the customer's own trailing twelve month run rate, which is precisely the wrong side of the asymmetry.
One clause decides whether the argument holds, so check it before relying on any of this. Some orders carry a separate overage rate card priced above the committed rate card.
Where that exists, the cost of under committing stops being a bounded margin and becomes an open ended premium, which inverts the whole calculation. Ask for a single rate card across committed and overage consumption: it is the term that converts under commitment from a risk into a managed position.
A three year order is three credit periods, not one pool, and roughly 1 in 4 pools expired with material value unspent. The term is how long you are locked in; the credit period is how long you have to spend, normally twelve months, with its own expiry.
Buyers hear three year commitment, plan a three year burn curve, and discover the year one shortfall did not carry forward. Negotiate the years together or size each one separately, because those are the only two coherent positions.
The two directions, priced
| If you | What happens | What it costs | Bounded |
|---|---|---|---|
| Consume more than committed | Excess invoiced monthly in arrears at the order rate card | The rate difference on the excess only | Yes, if one rate card applies |
| Consume less than committed | Unused credits forfeited at credit period end | 100 percent of the unused gap | No |
| Sign a separate overage rate card | Excess prices above the committed rate | An open ended premium on the excess | No |
| Move to pay as you go | No commitment, no forfeiture | Loss of the support offset eligibility | Yes, but it removes a benefit |
The overage mechanic is materially gentler than most buyers assume, and that single fact should reshape the sizing conversation.
Buyers size defensively because they imagine a penalty for exceeding the commitment, and there is not one in the standard shape: the excess is billed monthly in arrears against the rate card already in the order. Meanwhile the forfeiture on the other side is total.
Once both mechanics are on the table the sizing logic inverts, and the correct instinct is to commit the floor you are confident of and let genuine growth arrive as overage rather than as prepayment. Two caveats keep it honest. Check for a separate overage rate card, which breaks the argument.
And note that abandoning the commitment entirely for pay as you go forfeits eligibility for the support offset, so the flexible option is not free either. The negotiation frame sits in the cloud commitment negotiation guide.
Sizing to the defensible floor
- Commit what twelve months of telemetry supports, and put the growth case into years two and three instead of year one, which is where the ramp does its work.
- Ask for the ramp before the first draft order. In every commitment we modelled where a ramp was requested at that point, some version of it was granted. It is a free ask that most buyers never make.
- Ask for one rate card across committed and overage consumption, because that single term is what keeps the cost of under committing bounded and the sizing logic intact.
- Treat each credit period as its own sizing problem, since a three year order normally contains three twelve month windows each with its own expiry and no carry forward by default.
- Model the step, not the curve, because database infrastructure on a hyperscaler consumes on allocation rather than on use, so one provisioning decision can move a quarter of the annual commitment in a week.
The Oracle CIO complete playbook
The commitment arithmetic, the ramp shapes, the terms that outlive the discount, and the five year plan to control Oracle spend.
Get the white paper →Why the burn curve is a staircase
Most sizing models assume consumption ramps smoothly, and for compute and storage it broadly does: instances come up, volumes fill, the line climbs at a rate a finance team can extrapolate.
Database infrastructure on a hyperscaler behaves differently, because it consumes on allocation rather than on use.
The moment the infrastructure is provisioned it begins drawing the full rate whether or not a workload has migrated onto it.
So a single architectural decision taken by an engineering team on a Tuesday can move a quarter of the annual commitment inside a week, with no corresponding change in anything a business user would recognise as activity.
That has two practical consequences for the number in the order form.
The first is that the burn curve has to be modelled as a staircase with dated steps tied to provisioning decisions, rather than as a percentage growth line, and the steps have to be owned by whoever schedules the migrations rather than by whoever signs the commitment.
The second is that the risk profile changes direction over the period: early in a credit period the exposure is under consumption and forfeiture, and immediately after a large allocation the exposure flips to overage. A commitment sized on an annual average is wrong in both halves of the year.
Size to the floor you are confident of, schedule the steps deliberately, and let the ramp carry the growth into the periods where it will actually land.
Where AI consumption is being routed through the same balance, the sizing discipline is identical and the forecast risk is higher, which is covered in the AI credits analysis.
- Percentile standing for your exact deal size and industry, from real closed transactions
- Scenario simulation before the call: test alternative terms and see the financial impact of each
- A negotiation playbook, talking points, and a two page executive brief on day one
What we saw across commitment sizing work, 2023 to 2025
Across roughly 30 to 40 Universal Credits commitments modelled between 2023 and 2025, the headline discount absorbed almost all of the negotiating attention and the commitment size absorbed almost none, which is the wrong allocation of effort given how the two error directions price:
How far signed commitments exceeded the customer's own twelve month consumption history, on the expensive side of the asymmetry.
Annual credit periods that reached their anniversary with material value still on the balance, forfeited in full.
Three patterns recurred: signed commitments running 20 to 40 percent above the customer's own trailing twelve month run rate, roughly one annual credit pool in four reaching the end of its period with material value unspent.
And every ramp requested before the first draft order being granted in some form.
The buyer side move follows from the mechanics rather than from negotiating posture. Size to the defensible floor, ask for the ramp early, secure one rate card across committed and overage consumption, treat each credit period separately, and model provisioning steps rather than a smooth curve.
The wider library sits in the Oracle practice.
Your first five moves
- Check whether your order carries a separate overage rate card, because a single rate card across both is what keeps under commitment bounded and makes the whole sizing argument valid.
- Size the commitment to your trailing twelve month run rate, not above it, since over commitment forfeits 100 percent of the gap while under commitment costs only a rate difference on the excess.
- Request the ramp before the first draft order, as every ramp asked for at that stage was granted in some form, and asking after the draft exists is a much weaker conversation.
- Treat a multi year order as separate credit periods, each with its own expiry and no carry forward by default, and negotiate the years together if you want one pool.
- Model provisioning steps rather than a growth curve, because infrastructure that consumes on allocation can move a quarter of the annual commitment in a week. The Oracle practice runs the sizing with you.
Frequently asked questions
How does a Universal Credits commitment work?
You commit a dollar amount for a defined credit period, Oracle discounts the service rates against that commitment, and metered consumption draws the balance down until it reaches zero or the period ends.
The discount is real, and the commitment is also a bill you have agreed to pay whether or not the workloads land.
What happens if we consume more than we committed?
Oracle invoices the excess monthly in arrears at the rate card established in your order.
That is materially gentler than most buyers assume, and it is the single fact that should reshape sizing, because it means under committing costs a rate difference on the excess rather than a penalty on the whole amount.
What happens to credits we do not use?
They are forfeited at the end of the credit period. That makes over commitment cost 100 percent of the gap, against a bounded rate difference for under commitment, and roughly one annual pool in four in our file reached its anniversary with material value still unspent.
So how large should the commitment be?
Below expected consumption rather than at it, sized to the floor twelve months of telemetry supports, with the growth case carried by a ramp into years two and three.
That follows from the asymmetry rather than from risk appetite: being light costs a margin on the excess, being heavy costs the entire unused balance.
What could break that logic?
A separate overage rate card priced above the committed rate. Where an order carries one, the cost of under committing stops being a bounded margin and becomes an open ended premium, which inverts the calculation.
Ask for a single rate card across committed and overage consumption before relying on the asymmetry argument.
Is a three year commitment one pool of money?
No. It normally contains three twelve month credit periods, each with its own expiry, unless you negotiate the years together. The term is how long you are locked in; the credit period is how long you have to spend.
Buyers who plan a three year burn curve discover the year one shortfall did not carry forward.
Why does the burn curve behave like a staircase?
Because database infrastructure on a hyperscaler consumes on allocation rather than on use, so it draws the full rate from the moment it is provisioned regardless of whether workloads have migrated.
One architectural decision can move a quarter of the annual commitment inside a week, which is why steps have to be modelled and dated rather than smoothed.
Is moving to pay as you go the safe option?
It removes the commitment and the forfeiture risk, and it is not free: pay as you go customers are not eligible for the support offset, so the flexible route gives up a benefit worth real money on an estate carrying a significant on premises support bill.
Price that loss before treating flexibility as costless.
