Combining CUDs, SUDs and Spot VMs

Combining CUDs, SUDs and spot VMs is how a GCP buyer covers a real workload at the lowest defensible cost, and getting the combination wrong is how buyers quietly overpay. Each tool fits a different shape of demand. As of June 2026, committed use discounts reward a steady committed level over one to three years, sustained use discounts apply automatically to steady usage with no commitment, and spot or preemptible VMs cover interruptible work at a deep discount. They do not all stack on the same resource, and the order in which they apply matters. The goal is to layer them so the steady base is committed, the automatic discounts do their work, and the flexible peaks run on spot, without paying twice or committing to demand that should have stayed variable.

Combining CUDs, SUDs and spot VMs

Combining CUDs, SUDs and spot VMs starts with understanding that they target different demand and do not all double stack. A committed use discount applies to a committed level of resources or spend. A sustained use discount applies automatically to eligible steady usage that is not already under a commitment. A spot VM is a separate purchase of interruptible capacity at a deep discount. As of June 2026, CUDs and SUDs do not double stack on the same resource, so the question is not how to apply every discount at once, it is how to assign each unit of demand to the cheapest tool that fits it.

Map demand to tools by shape. The steady, predictable base belongs on committed use, because it is the usage you will consume regardless and the commitment turns that certainty into a discount. The steady usage you choose not to commit still earns sustained use discounts automatically, so there is a floor of saving even without a commitment. The interruptible, fault tolerant peaks belong on spot, where they run far below on demand rates without any commitment at all.

The combination fails when buyers try to cover everything with one tool. Commit to the whole estate and you overpay for the variable demand and the interruptible work that should have been cheaper elsewhere. Run everything on demand and you forgo the committed and sustained discounts on the steady base. The buyer side discipline is to split the workload by its demand shape and assign each part to committed use, sustained use, or spot, rather than reaching for a single lever.

How the three discounts actually stack

Committed use and sustained use do not stack on the same resource. Committed use applies to the resources you have committed, and sustained use applies automatically to eligible steady usage that sits outside a commitment, as of June 2026. In practice this means committed use is your tool for the base you are confident enough to commit, and sustained use is the automatic reward on the steady usage you leave uncommitted. They are complementary across your estate, not additive on a single unit of usage.

Spot capacity is a separate purchase, not a discount layered on top of committed or on demand pricing. A spot VM is interruptible capacity bought at a deep discount, and it is the right tool for fault tolerant work that can absorb being reclaimed. Because it is a different resource rather than a discount on an existing one, spot does not interact with the commitment math. It simply removes interruptible demand from the pool that would otherwise need committed or on demand coverage.

Because the discounts do not all combine on the same resource, the saving comes from allocation, not from stacking. The buyer who tries to maximize stacked discounts on one workload is solving the wrong problem. The buyer who allocates each unit of demand to the cheapest appropriate tool, committed for the base, sustained automatically for the uncommitted steady usage, and spot for the interruptible peaks, captures the real saving across the whole estate.

Matching each tool to the right demand

Committed use fits the durable base, the steady level of usage you will run for the full term even in a poor scenario. This is the demand worth committing because the commitment converts certainty into a discount, and the risk of shortfall is low when the base is genuinely durable. Size the commitment conservatively to that floor, leaving the automatic sustained use discounts and on demand pricing to absorb variability above it.

Sustained use discounts need no decision, but they should still shape your commitment sizing. Because steady uncommitted usage already earns an automatic discount, the incremental gain from committing that usage is smaller than the headline committed use rate suggests. Always measure the committed use decision as the gain over the sustained use baseline, not the gain over list price, because the sustained use discount is what you would earn anyway and is the right comparison point.

Spot fits interruptible, fault tolerant demand, the batch jobs, processing pipelines, and stateless work that can tolerate being reclaimed. Routing this demand to spot removes it from the committed and on demand pools entirely, covering the peaks at a deep discount without extending a commitment to reach them. The more interruptible work you can move to spot, the smaller and safer your commitment needs to be, because the commitment only has to cover the steady base that genuinely must run uninterrupted.

Where buyers overpay by mixing them wrong

The most common overpayment is committing to demand that should have been spot or on demand. When buyers chase coverage, they fold interruptible and variable workloads into a commitment, paying a committed level for demand that flexes or could have run far cheaper on spot. The fix is to pull that demand back out, route the interruptible part to spot and the variable part to on demand, and shrink the commitment to the steady base that remains.

A second overpayment is ignoring the sustained use baseline when valuing a committed use offer. If you measure a committed use discount against list price, it looks larger than it is, because your steady usage already earns sustained use discounts automatically, as of June 2026. The real value of the commitment is the incremental gain over that automatic baseline, and a buyer who negotiates against list rather than against the sustained use rate is negotiating against an inflated number.

The third overpayment is oversizing the commitment because the combination was not planned first. When the spot and sustained use layers are an afterthought, the commitment ends up carrying demand those layers should have covered, which inflates both the committed amount and the shortfall risk. Plan the allocation across all three tools before fixing the commitment, so the committed level is only the steady base that is left after spot and on demand have taken the demand that suits them.

A practical layering approach

Start by carving out the interruptible work and sending it to spot. Identify every fault tolerant, restartable workload and route it to spot or preemptible capacity, capturing a deep discount with no commitment. This is the first move because it shrinks everything that follows, removing demand from the pool that the commitment and on demand pricing would otherwise have to cover.

Next identify the variable but not interruptible demand and leave it on on demand pricing, earning sustained use discounts automatically where the usage is steady enough to qualify. This layer needs no commitment and no special handling. It simply absorbs the demand that is too important to risk on spot but too variable to commit, and the automatic sustained use discount reduces its effective rate without any lock in.

Finally size the committed use discount to the steady base that remains after spot and on demand have taken their share. Commit conservatively below that floor, measure the gain against the sustained use baseline rather than list price, and prefer flexible and shorter structures where the discount difference is small. An independent buyer side review can confirm the allocation is right before you commit, so you capture the discounts that suit each demand shape without paying for any you do not.

Sources, method, and as of date

The program mechanics and ranges on this page reflect publicly available provider documentation and our buyer side negotiation experience, as of June 2026. AWS, Microsoft, and Google revise their programs frequently, so treat every figure as a point in time reference and confirm the current terms directly with the provider before you act.

This page is commercial negotiation advisory, not legal, tax, or accounting advice. We are independent and buyer side, with no reseller margin and no hyperscaler incentive, and we are paid only by the buyer. For interpretation of any commitment contract or program term, engage your own legal counsel.

Layer committed, sustained, and spot so you never pay twice.

A CONFIDENTIAL COMMITMENT REVIEW BEFORE YOU SIGN