The Commit Size That Quietly Locks You In

There is a particular Claude commitment size that feels like the smart move and behaves like a cage. It is large enough to earn a satisfying discount, large enough to make the account team enthusiastic, and just large enough that you cannot easily walk away at renewal. The lock in is not in any single clause. It is in the size itself, in the dependence a big commitment creates and the leverage it quietly transfers to the seller. This is the buyer side view of how commit size becomes lock in, and how to size your Anthropic commitment so the discount does not come with a leash.

We negotiate Claude contracts for enterprise buyers and study nothing else. The oversized commit is one of the most common patterns we are called in to unwind, usually at the renewal, when the buyer discovers that the number that felt prudent last year has become the floor they are negotiating up from this year.

How a commit becomes a lock in

A commitment is supposed to be a trade: you promise volume, you get a rate. The trade goes wrong when the volume you promise is larger than the volume you can comfortably consume or comfortably leave. At that point the commitment stops being a discount mechanism and starts being a dependence mechanism. You have built so much spend into one vendor that the cost of changing course, renegotiating hard, or walking away exceeds the savings the discount ever provided.

The seller understands this perfectly. A large committed customer is a predictable customer, and a predictable customer who depends on the relationship is one who renews on the seller's terms. The discount that won the large commit was the bait. The dependence is the hook. None of it requires bad faith from the account team. It is simply how committed spend works when the commitment is sized past the point of comfort.

The signs you are heading for it

The oversized commit announces itself if you know what to listen for. The discount jumps disproportionately at a size just above your real forecast, which is the seller pulling you across a threshold. The proposed number assumes growth you have not yet demonstrated, baking optimism into an obligation. The term is long and the structure offers no way to reduce the commitment if your usage comes in lower. And the unused commitment treatment is unfavorable, so anything you do not consume simply disappears at the period boundary rather than rolling forward.

Put those together and you have a deal that bills you for optimism, punishes you for shortfall, and leaves you too embedded to push back next year. Each element looks reasonable alone. Together they are the anatomy of a lock in, and a buyer who recognizes the pattern can refuse the parts that create it while keeping the rate.

Size to a forecast you control

The defense is to size the commitment to a forecast you actually believe, built from your own usage data, not to the number that maximizes the headline discount. Pull your real consumption, project it honestly, and commit to the volume you are confident you will consume. A slightly smaller discount on a number you will fully use beats a richer discount on a number you will partly waste, every time, because the wasted commitment is pure loss when it does not roll over.

Crucially, your forecast should be built on optimized consumption, not your current spend. Routing across Opus, Sonnet, and Haiku rather than running everything on Opus typically cuts aggregate spend 40 to 70 percent. Prompt caching returns up to 90 percent on stable context. Batch runs asynchronous jobs at 50 percent. If you commit to today's wasteful number and then optimize, you have locked yourself into paying for tokens you have engineered away. Optimize first, forecast from the efficient baseline, and commit to that.

Structure out the lock in

Even at the right size, structure protects you. A ramp lets you start the commitment lower and step up as adoption grows, so you are never paying ahead of real usage. A true forward lets you convert overage into committed spend at the committed rate rather than paying a penalty, so growth does not punish you. And renewal terms that price from real usage rather than from your peak stop a single big year from becoming a permanent floor. These provisions keep a commitment from hardening into a lock in even as it grows.

Keep an alternative alive

The deepest protection against lock in is not contractual at all. It is keeping a credible alternative. A buyer who could route some workloads through a cloud commitment they already hold, or who has optimized enough to need far less Claude than the seller assumes, is a buyer who cannot be locked in, because dependence requires the absence of options. You do not have to exercise the alternative. You only have to have one, and to make sure the seller knows it is real. That single fact changes every renewal conversation from a negotiation about how much more you will pay into a negotiation about whether you stay at all.

A worked example of the trap

Picture a company forecasting a year of Claude usage at a level its data supports. The account team proposes a commitment about a third larger, pointing to a discount tier that opens just above the higher number, and frames the extra as headroom for growth. The deal looks attractive, because the larger commitment carries a visibly better rate, and committing to growth feels prudent rather than risky. So the company signs the bigger number.

Two things then happen. The forecasted growth arrives slower than hoped, so a meaningful slice of the commitment goes unconsumed, and because unused commitment does not roll over, that slice is simply lost. And the company, now carrying a large commitment it leaned on to build its products, arrives at renewal embedded and dependent, negotiating up from the inflated number rather than down from a clean one. The discount that justified the larger commit has been more than erased by the waste and the weakened renewal position. The smaller commitment, fully consumed, would have cost less in total and left the company freer.

This is the trap in its ordinary form. It does not require anyone to act in bad faith. It only requires a buyer to accept a number sized to the seller's optimism in exchange for a rate, and to discover the cost of that choice a year later when the optimism has not materialized and the commitment has hardened into a floor.

The renewal is where the size is tested

Everything about commit size comes due at renewal. A commit sized to reality, ramped sensibly, and protected with true forward and renewal terms arrives at renewal as a clean number you can defend or reduce. A commit sized to optimism arrives as a peak the seller renews from, with your own dependence as their leverage. The size you choose this term is not just this term's cost. It is the opening position for every negotiation that follows, which is exactly why the seller wants it large.

The buyer side summary

The commit size that quietly locks you in is the one sized to the seller's optimism rather than your reality, stretched past a threshold for a discount you pay for in dependence. Size to a forecast you control, built from optimized consumption. Ramp it so you never pay ahead of usage. Protect it with a true forward and renewal terms priced from real use. And keep a credible alternative alive so dependence never becomes the seller's leverage. A commitment built this way earns its discount without becoming a cage, which is the only kind worth signing.

If a Claude commitment is on the table and the proposed number feels larger than your forecast supports, that gap is exactly where lock in lives. Get in touch or see how we are paid, fixed fee or gainshare, and we will size the commitment to what you will actually use before you sign it.