6 Questions When Your AI Supplier Can Reprice Before You Can

1. What price and service have you promised the customer?

The first question is the scope of the customer’s price commitment. Record the service, included volume, quality conditions, support and duration. A monthly invoice does not necessarily mean the seller can change the price every month. The actual agreement may contain a longer commitment, a notice requirement or a defined adjustment mechanism.

2. What price commitment has the supplier actually made?

The second question is the supplier’s corresponding commitment. Identify the model or service, billable unit, currency, minimum spend, rate schedule and any negotiated protection. A published price page and a signed enterprise order can govern different circumstances. Preserve the applicable document and version rather than copying a convenient number from a search snippet.

For a concrete example of why this matters, Anthropic’s commercial terms H.1 address fees and rate updates, including a specific notice provision and separately agreed terms. This article does not resolve how that clause applies to a particular customer. The practical task is to identify the actual effective date and governing agreement before placing it in the exposure model.

3. When can each obligation change?

The third question is the timeline. Put customer renewal, customer repricing, supplier update, notice receipt and termination dates on the same calendar. Announcement, notice and effective date are distinct events. Do not assume that reading a new price creates an immediate right to change an existing customer agreement.

Mark the interval during which the supplier’s new rate applies while the customer’s old price remains binding under the modeled terms. That interval is the exposure horizon. A supplier increase arriving one month before customer renewal has a different consequence from the same increase arriving one month after a year-long renewal.

Preserve the operational response time as well. Even when a permitted alternative exists, migration, testing and customer communication can take time. A contractual option to exit is not the same as a ready substitute. Record which response can actually be completed before the effective change, and what new cost or service limitation it introduces.

Build the timeline from documents rather than assumptions about standard notice. Save the applicable rate notice, the contract version and any customer acknowledgment relevant to the decision. When the effective date is ambiguous, record the ambiguity and obtain clarification through the authorized commercial route. The model can compare several possible dates, but a scenario date should not be presented as an established contractual deadline.

4. Which usage and cost assumptions remain stable?

The fourth question is the unit of cost. A model token, search call, storage unit and completed customer job are different quantities. Use a billable unit that matches the supplier rate, then connect it to the customer service through observed or explicitly estimated usage. Do not multiply a per-token price by a job count without a conversion assumption.

The narrow calculator holds monthly usage and other monthly costs constant to isolate the rate change. That is useful for understanding one mechanism, but actual usage can grow, retries can rise and customer mix can change. Run separate scenarios for those effects rather than presenting the constant-volume result as a full forecast.

Keep price and currency effects separate. If the customer pays in one currency and the supplier charges in another, a fuller model must include the conversion rate, timing and fees. This calculator uses one currency throughout. A cash-prepayment obligation is also distinct from expense timing; contribution can remain positive while liquidity becomes strained.

Preserve the distinction between supplier rate and effective cost per accepted outcome. A stable published rate can still produce higher cost if the workflow requires more attempts. Conversely, a higher unit rate can be cheaper per completed job if fewer attempts are needed. This article isolates contractual repricing; a separate task-level observation is needed to establish those performance effects.

5. What happens to contribution over the remaining term?

The fifth question is the resulting contribution. In a synthetic six-month contract, revenue is 1,200 CU per month. Usage is 1,000 billable units per month. The input rate is initially 0.40 CU per unit, with 300 CU of other monthly costs. Monthly contribution is 500 CU before any excluded costs.

If the input rate becomes 0.70 CU from month three, supplier cost rises from 400 to 700 CU and monthly contribution falls to 200 CU. Two months at 500 CU plus four months at 200 CU produce 1,800 CU over the term. With no rate change, the same model would produce 3,000 CU. The change therefore reduces modeled contribution by 1,200 CU.

The constant-volume break-even supplier rate is (1,200 − 300) ÷ 1,000 = 0.90 CU per unit. This is a mathematical boundary, not a recommended procurement price or a guarantee of adequate business margin. It excludes any costs omitted from the entered monthly allowance.

Stress a second input: if usage rises to 1,400 units at the 0.70 CU rate, supplier cost becomes 980 CU and contribution becomes minus 80 CU per month. The same customer fee that survived the price change alone fails under the combined scenario. Preserve both changes so the explanation does not attribute the loss entirely to repricing.

6. Which response is permitted, practical and worth its cost?

The sixth question is the response boundary. Options may include a permitted customer adjustment, a narrower future scope, a negotiated supplier commitment, a tested substitute or declining an uneconomic renewal. Which options exist depends on the actual agreements and delivery requirements. The calculator does not authorize unilateral changes to a customer promise.

Price the response itself. A new supplier may have lower headline rates but require migration, revalidation and changed operations. A longer input commitment may stabilize price while creating minimum spend or exit exposure. Compare those obligations with the customer demand that is actually committed, not with an optimistic sales target.

Set an internal trigger before the margin reaches zero: a contribution floor, an exposure amount or an approaching effective date. Assign an owner to review the source terms and a deadline for the next decision. This makes the rate change an explicit commercial event rather than a surprise discovered after several weak invoices.

At the next sale, choose a commitment horizon the business can support. A fixed price can be valuable to a buyer, but that certainty has a cost when the inputs remain variable. The commercial decision is whether the fee and conditions adequately fund that exposure. Faster cost tracking helps, but it does not remove the underlying mismatch in promises.

A protected rate can also be tied to a commitment that deserves its own stress. If the seller must buy a minimum volume to secure the rate while its customer can reduce usage, the benefit of price stability may be offset by unused capacity. Compare the protected-rate scenario with actual committed demand and the cost of the unused portion. Do not count an attractive unit price as a saving before the minimum purchase and exit conditions are included.

Frequently asked questions

Primary sources

1. Anthropic commercial terms. Commercial terms H.1 permits published-rate changes subject to the exact notice provision and any separately agreed terms. No actual future price change is asserted. Reviewed 2026-11-11.

Related DN research

• The Agent Cost Stack
• Real Cost of an Agent API Call