Rollover and carryover let unused commitment count toward the next period instead of vanishing. They are negotiable, they are rarely offered by default, and they can save a forecast miss. Here is how they work.
The default on a committed spend agreement is unforgiving. You promise a number, and anything you do not consume by the period end is gone. Carryover and rollover terms change that default. They let unused commitment flow forward, so a slow quarter does not become a permanent loss. These terms are among the most valuable a buyer can win, and they are almost never in the first draft. This is what they mean, how they differ, and how to negotiate for them.
Rollover and carryover are often used interchangeably, but it helps to be precise about what you are asking for. Carryover usually means that commitment unused in one period is added to the available balance in the next, within a cap. Rollover often describes the same idea applied across a renewal, where unspent commitment from the ending term carries into the new one rather than resetting to zero. In both cases the principle is identical: spend you paid for but did not use is treated as a balance you still hold, not as a shortfall the vendor keeps.
The reason this matters is that forecasts are never perfect. Even a careful buyer can be off by a quarter or two when a product launch slips or a use case is cut. Without a carryover term, that timing miss is a full loss. With one, it is just a delay, and the commitment you paid for is still yours to use.
A vendor prices a commitment on predictability. The discount you receive is justified by the certainty of the revenue. Carryover weakens that certainty, because it means revenue the vendor expected in one period might not arrive until later. That is why the standard agreement excludes it and why a seller will push back when you ask.
They concede anyway, on larger deals, for two reasons. First, a buyer who is protected against the downside is willing to commit to a higher number, which the vendor wants. Second, carryover with a cap and an expiry is a limited concession that still preserves most of the predictability. A term that lets you carry, say, a quarter of the annual commit for one additional period is far less risky for the vendor than open ended rollover, and far more likely to be agreed. The art is in the limits.
A workable carryover term has a few moving parts, and each is negotiable.
The cap is how much unused commitment can carry forward, usually expressed as a percentage of the period commit. A higher cap protects you more. A seller will want it low. Somewhere in the middle, often a meaningful fraction of a quarter or more, is a realistic outcome on a sizable deal.
Carried commitment usually has to be used within a defined window or it expires. A longer window is better for you, because it gives a delayed product more time to ramp into the balance. Push for the longest expiry you can, since a short one can mean the carried balance vanishes before your usage recovers.
Clarify whether carryover survives a renewal or only applies within a term. Carryover that resets at renewal is useful but limited. Rollover that crosses the renewal boundary is more valuable, because it means the timing of your renewal does not force you to forfeit a balance you have already paid for.
Carryover and a ramped commitment solve related problems and work best together. A ramp sets a lower floor early and grows it as your usage scales, which reduces the chance of underusing in the first place. Carryover catches the residual, the gap between even a ramped forecast and reality. A buyer who structures a ramp and adds a carryover cap has protected both the predictable risk and the surprise. Neither alone is as strong as the pair.
It helps to be clear about the limits. Carryover is not a refund. It does not give you cash back for unused commitment, it gives you more time to use what you paid for. It is also not a substitute for an honest forecast. A buyer who overcommits wildly and relies on carryover to save them will still lose, because the carried balance has a cap and an expiry. Carryover is insurance against a reasonable miss, not a license to commit to a number you have no path to reaching.
There is a right way and a wrong way to raise carryover. The wrong way is to ask for it as an afterthought near signing, when the rate is settled and the seller has no reason to give ground. By then carryover looks like a free concession with nothing offered in return, and it gets refused. The right way is to raise it early, as part of the structure, and to tie it to the size of your commit. A buyer who says they will commit to a larger number if carryover protects the downside is offering the seller something real, more committed revenue, in exchange for the term. That trade is far more likely to land, because it improves the deal for both sides rather than simply shifting risk.
Frame carryover as the thing that lets you commit confidently rather than as insurance against your own forecast. The seller wants a bigger commit and more predictability. A capped, expiring carryover term gives them most of the predictability while giving you the room to commit higher without fear. Presented that way, it is a structural improvement to the deal, not a one sided ask, and it negotiates accordingly.
A carryover clause is only as good as its parameters, so do not accept the principle without negotiating the figures. A term that allows ten percent of the annual commit to carry for thirty days is technically carryover, but it protects almost nothing. A term that allows a quarter of the commit to carry for a full additional quarter is meaningful protection against a real forecast miss. The cap and the expiry are where the value lives, and a seller who concedes the principle will often try to set the parameters so low that the clause is decorative. Push the cap up and the window out until the term would actually have covered the kind of miss you are realistically exposed to. If it would not have helped in your worst plausible quarter, it is not yet worth having.
Take a company that commits to a year of spend and, because of a delayed launch, ends the first half well below plan. Without carryover, that first half shortfall is simply lost when the period closes, and the company spends the second half racing to use a commit it can no longer fully reach. With a carryover term that lets a quarter of the commit flow forward for an additional period, the early shortfall becomes a balance the company draws down as the delayed product finally ramps. The same delay that would have cost real money under the default costs almost nothing under the term. The product was late either way, but only one of the two contracts punished the company for it.
Carryover only helps if your usage eventually grows into it. The fastest way to consume a balance efficiently, and to need a smaller commit in the first place, is to optimize the workload. Routing across Opus, Sonnet, and Haiku so each task runs on the cheapest capable model, caching stable context at up to 90 percent off, and moving asynchronous jobs to batch at 50 percent off can cut aggregate spend 40 to 70 percent versus running everything on Opus. Counterintuitively, optimization can mean you consume a carried balance more slowly in dollar terms while doing more work, which is exactly the position you want: more output, less spend, and a commitment that stretches further.
Get a quote for a bounded engagement. Fixed fee or gainshare, no risk to you.
Get a QuoteWeekly intelligence on Anthropic pricing moves and the buyer side counters that work.