How much of my cloud spend should I commit?

Commit the floor you incur every month with confidence, add only the probable spend backed by plans already in motion, and leave speculative growth on demand. The exact share is expressed as a coverage ratio set against your tolerance for downside.

Is it worse to commit too much or too little?

Committing too much is worse. As of June 2026, overcommitment costs the full unused amount through AWS EDP shortfall or lost Azure MACC commitment, while undercommitting costs only the discount delta on spend that flowed on demand. The penalties are not symmetric.

What is a sensible on demand buffer?

Hold enough on demand spend that a soft quarter never pushes you below your commitment. The buffer is insurance against the asymmetric shortfall penalty, not waste. Buyers with variable workloads should keep a larger buffer.

Does workload type change the commit versus on demand split?

Yes. Spiky or seasonal workloads should keep a larger on demand layer because committing to an occasional peak guarantees waste in the troughs. Steady, predictable workloads can support a higher commitment share.

How do I express the decision to procurement?

Use a coverage ratio, the share of expected spend the commitment covers. It gives a single defensible number to hold in the room rather than arguing line by line over which workloads to commit.

Should I commit spend that Savings Plans already cover?

Be careful not to double count. As of June 2026 GCP CUDs and SUDs do not double stack on the same resource, and Savings Plans already discount part of your usage. Base the commit decision on net new value, not savings you already have.

COMMITMENT STRUCTURING · BUYER SIDE

How much to commit vs leave on demand

GET A CONFIDENTIAL REVIEW →

PUBLISHED JUNE 16, 2026 · REVIEWED JUNE 16, 2026

Deciding how much to commit vs leave on demand is the heart of commitment structuring. Commit too much and you pay for capacity you never use. Commit too little and you leave discount on the table for spend you were always going to incur. The right split is not a single magic number, it is a function of how confident you are in each layer of your spend and how much you can afford to lose if you are wrong.

The seller wants the commitment as large as possible. Your interest is the opposite, to commit only the spend that is certain and let everything uncertain flow at on demand rates. Here is how to find that line, layer by layer, and how to defend it in the room.

How much to commit vs leave on demand, layer by layer

Picture your cloud spend as three layers. The bottom layer is the floor you incur every month without fail. The middle layer is probable spend that depends on plans already in motion. The top layer is speculative growth that may or may not arrive. Commitment belongs on the floor, is defensible on part of the middle, and is dangerous on the top.

As of June 2026, the penalty math makes this layering essential. AWS EDP shortfall is paid by the buyer when consumption falls below the commitment (source: AWS EDP program terms), and Azure MACC unused commitment is generally lost with no rollover (source: Microsoft MACC documentation). Commit the speculative top and a soft quarter turns into cash out the door. Leave it on demand and the only cost is a thinner discount on spend that might not have happened anyway.

Identify the floor with hard data

The floor is not your average monthly spend, it is the spend that persists in your worst recent month. Pull at least twelve months of usage, find the trough, and treat that as the floor you can commit with confidence. Everything above the trough is variable to some degree. This is the same starting point as how to size a cloud commitment correctly, and it is where most buyers overreach by anchoring to an average rather than a trough.

The asymmetry that should anchor the decision

Overcommitting and undercommitting do not cost the same. Overcommitting costs you the full unused amount, either as a shortfall payment or as forfeited commitment. Undercommitting costs you only the discount delta on the spend that flowed on demand instead of under the commitment. The first number is far larger than the second. When in doubt, commit less. The math of the downside is on your side.

This is the same asymmetry that drives the choice between conservative versus aggressive commitment sizing. A buyer who internalises the asymmetry naturally lands on a conservative split, and stops treating the discount on speculative spend as if it were guaranteed.

Translate the split into a coverage ratio

The cleanest way to express how much to commit is a coverage ratio, the share of your expected spend that the commitment covers. A conservative buyer might target coverage of the floor plus a slice of probable spend, leaving a healthy on demand buffer. We walk through target ranges and how to set them in commitment coverage ratios explained. The coverage ratio gives procurement one defensible figure to hold in the room rather than arguing line by line while the sales team controls the story.

Variable workloads change the answer

If a large share of your spend is spiky or seasonal, the on demand layer should be bigger, because committing to a peak you only hit occasionally guarantees waste in the troughs. We treat this case in detail in commitment structuring for variable workloads. Steady, predictable workloads can support a higher commitment share. Bursty ones should not, no matter how attractive the tier looks.

Build a buffer on purpose

Identify the floor spend you incur every single month and commit that with confidence.
Add only the portion of probable spend that is backed by plans already in motion.
Leave all speculative growth on demand until it proves itself.
Hold a deliberate on demand buffer so a soft quarter never pushes you into shortfall.
Revisit the split at each renewal, when your floor is clearer and your leverage returns.

This buffer is not waste, it is insurance against the asymmetric downside. The full sizing method that produces these layers is in how to size a cloud commitment correctly and the broader framework is in our cloud commitment structuring and sizing guide.

Account for what is already discounted

The commit versus on demand decision gets more subtle once Reserved Instances, Savings Plans, and automatic discounts are in play. As of June 2026, GCP Sustained Use Discounts apply automatically with no commitment, and CUDs and SUDs do not double stack on the same resource (source: Google Cloud documentation). If part of your spend is already discounted by another instrument, the marginal benefit of committing it again is smaller. We untangle the layers in reserved instances vs savings plans vs commitments so your commit decision reflects net new value, not double counted savings.

How to defend the split in the room

Procurement loses the commit versus on demand argument when it tries to litigate every workload. The seller has more product knowledge and more time. The winning move is to set the coverage target first, present it as policy, and require the seller to justify any spend above it. A single defensible number, backed by your own usage trough, is far harder to argue against than a workload by workload debate. Walking in with the number already set changes who is on the back foot.

Reassess the split as the estate changes

The right split is not fixed for the life of the deal. As workloads migrate, retire, or scale, the floor moves and the probable layer firms up or evaporates. A split that was conservative at signing can drift toward overcommitment if spend softens, or leave discount unclaimed if spend climbs. Treat the commit versus on demand decision as a position you manage, not a number you set once. Review it every quarter against realised usage and bring the findings into your renewal preparation.

This is also where flexibility provisions earn their keep. The right to adjust the committed amount, or to shift unused commitment within the term where a provider allows it, turns a static split into one you can correct as reality unfolds. We cover the provisions worth negotiating in building flexibility into a commitment.

The bottom line on how much to commit vs leave on demand

Commit the certain floor, defend a measured slice of probable spend, and leave the speculative top on demand. The asymmetry of the penalties means erring toward less commitment is almost always the cheaper mistake. If a proposal is pushing you to commit your forecast rather than your floor, a commitment structuring and sizing service will rebuild the split from real consumption before you sign.