The AI labs standardized the same acquisition mechanic, and almost nobody read the fine print: discount only at entry, full price at renewal.
Z.AI gives 10% off the plan’s first payment ("Subsequent renewals, upgrades, downgrades are NOT eligible", say the credit-campaign-rules). OpenCode Go gives 5 dollars of entry credit. OpenRouter launched a model with 50% off the launch window. The mechanic is the same in all three: free access and discounts are lab CAC, not price.
The number that enters the budget is month 2’s
The standard buyer error: planning the invoice with the entry discount. The contract says the opposite: renewal, upgrade, and downgrade return to full price, and the discount is an event, not a regime.
AI FinOps starts exactly here: budget the plan at full price and treat the discount as a decision bonus. Whoever does the inverse signs with the wrong number and finds out in month 2, when the invoice rises and the contract already said it would.
The discount has a function: funding the experiment
None of this means ignoring the discount. It has a precise function: funding the experiment with real load.
The discounted window is the moment to measure properly: cost per completed task on your load, model comparison, boundary testing. The discount pays the lab; the subscription decision comes from the measurement, not from the discount. It is the same logic as any model’s free window: the favored-price period is the lab, and the full invoice is real life.
Renewal is the reassessment moment
Whoever enters with a discount needs to know the day the invoice returns to full, and treat that day as a decision event, not an automatic debit.
At renewal, the model that was cheap with the discount competes with today’s market: the price war cut 40 to 50% in one day, and geography’s off-peak still holds. Renegotiating, comparing rulers, and deciding with current measurement is what separates a contract from a habit.
Conclusion
The economics of first payments is the most underrated discipline in AI FinOps: every vendor offers the discount you want, and the contract charges the price you did not read.
The rule of thumb: budget at full price, use the discount to experiment, mark the renewal on the calendar, and re-decide with measurement in hand. The 10% first-payment discount is the paid lab; the budget is month 2 onward. Whoever confuses the two funds the lab’s marketing believing they are buying intelligence.