An AI feature will reach ten million registered users. Estimate its monthly inference cost and defend the assumptions.
Make your call first. Use the published answer and rubric to examine your reasoning when you are ready.
Case briefRead the inputs, then make your call
All figures, including prices, are fictional planning assumptions. Retry calls have the same token shape as primary calls in this base case. Estimate the model bill separately from full operating costs.
| Registered users | 10,000,000 |
|---|---|
| Monthly active share | 20% |
| Tasks per active user per month | 8 |
| Primary model calls per task | 2 |
| Extra retry calls as share of primary calls | 5% |
| Input tokens per call | 1,500 |
| Output tokens per call | 300 |
| Uncached input price per million tokens | $2 |
| Output price per million tokens | $8 |
| Successfully completed share of attempted tasks | 90% |
| Separate cache scenario | 40% of input tokens are qualifying hits at a 75% discount; no additional cache-write or storage charge in this fictional tariff. |
Your task. Calculate monthly calls, input and output volumes, the uncached and cached model bill, and cost per successful task. Identify the next assumptions to measure and the missing capacity and operating-cost inputs.
0 words