Scaling 10k to 1M
What the bill, the provider meters and the spend cap do as users grow from 10K to 1M, derived once from one canonical workload at the 2026-10-08 prices, with the bottleneck named at each 10x.
Last updated
After this section you can
- Derive per-user and per-month model cost from a workload at current prices.
- Read a peak load against per-model limits to name the meter that binds first.
- Decide when a stage needs a contract or a second region, and defend it with one number.
17
Scaling 10k to 1M
The bottleneck moves at each 10x: the bill at 10K users, the tier at 100K, the contract at 1M.
Key idea
Every 10x in users moves the bottleneck. At 10K the monthly bill outgrows the spend cap first; at 100K the tier decides which meter binds; at 1M the cheap tier's output meter and the $200K cap both bind.
Each 10x moves the bottleneck
Same workload: 5 queries a day, 70/30 model mix, no cache. Pans sideways on a phone.