One year and three year committed use discounts on GCP are the same instrument with different terms: you commit to a level of usage or spend, and in return GCP discounts the rate. The three year term discounts more deeply, up to around 57 percent for resource based commitments against up to around 37 percent for one year, figures indicative of current pricing, but it holds your forecast still for three times as long. Because a CUD bills whether or not you use it, the term decision is really a confidence decision: how sure are you that this exact load will still be running in three years. The disciplined answer is to ladder the terms, three year under the certain baseline, one year over the reasonably stable layer, on demand above that, so the deepest discount sits only where the risk is lowest.
Choosing one term for the whole estate either leaves discount on the table or strands commitments when load moves. Here is how to choose the term workload by workload.
How much more does a three year CUD save?
The three year term is meaningfully cheaper per unit, and the gap is the whole reason to consider the longer lock in. For resource based CUDs the three year discount reaches up to around 57 percent off on demand while the one year reaches up to around 37 percent, with the exact figures varying by machine family and indicative of current pricing. Spend based CUDs follow the same pattern at smaller absolute numbers. The deeper rate compounds over a longer base, so for load that genuinely runs the full three years the three year commitment is clearly the lower lifetime cost. The question is never whether three year saves more on paper, it always does, but whether the load survives the term.
What is the real cost of the longer lock in?
A CUD is an obligation, not a coupon. Once committed you pay for the level you committed to for the full term regardless of usage, so the cost of the three year term is the risk that the load it covers shrinks, migrates to another service, or is rearchitected before the term ends. If demand falls and you cannot backfill the commitment with other eligible usage, the unused portion is stranded spend that can wipe out the discount advantage. The longer the term, the more chances for a reorganization, a platform change, or a product sunset to move the load out from under the commitment. That is why term selection follows the durability of the workload, not the size of the headline discount.
How do you choose the term per workload?
Sort your baseline by how confident you are it will persist, then match the term to the confidence. The table shows the laddering rule.
| Load layer | Forecast confidence | Term to use |
|---|---|---|
| Core baseline that has run flat for a year or more | High, durable, no planned change | Three year CUD for the deepest rate |
| Stable but evolving services | Medium, likely to persist but may shift | One year CUD, renew from strength |
| New, seasonal, or migrating workloads | Low, may move or disappear | On demand or Spot, no commitment yet |
Set coverage targets per layer rather than a single estate wide number, and prefer spend based CUDs for the layers where the specific machine type might change, since they let other usage backfill the commitment if one workload moves.
A worked laddering example
A scaling fintech wanted the headline three year discount across its whole GCP fleet. Its usage history showed a durable core that had been flat for over a year, a middle layer of services still being reshaped, and a fast moving experimentation tier. We placed three year resource based CUDs under the durable core only, one year spend based CUDs under the middle layer so other usage could backfill it, and left experimentation on demand. The estate captured almost the full discount it would have from blanket three year coverage, but with no stranded commitment when two middle layer services were later rearchitected, and the figures are verified against billing data and anonymized.
For any workload you are about to put on a three year CUD, ask whether you would bet it runs unchanged until 2029. If the honest answer is no, use a one year term and renew it from a position of strength.
Where this fits in the GCP estate
Term choice is one part of a wider commitment discipline. Understand the two CUD shapes in spend based versus resource based CUDs, set the right floor with commitment coverage targets on GCP, watch utilization in CUD utilization monitoring, and know the exit in when to let a CUD lapse. The full method lives in the GCP cost optimization guide.
Frequently asked questions
What is the difference between one year and three year CUDs?
Is a three year CUD worth the extra lock in?
Can you mix one year and three year CUDs?
Get the buyer side commitment playbook
We ladder one year and three year CUDs to a forecast you can defend, so you capture the deep discount without stranding spend when load moves. Our guarantee: we reduce your cloud spend or we reimburse our service fee. Pricing is either a Fixed Fee scoped up front or Gainshare, a share of verified savings with no retainer and no risk.
Put a defensible number on your cloud spend.
No provider in the room, no published price list. Tell us your footprint and we will scope the savings against your billing data — we reduce your cloud spend or we reimburse our service fee.
The Cloud Spend Navigator: what changed in cloud pricing, commitments, and FinOps — no vendor spin.