Summary
Neon provides a serverless Postgres platform that separates storage from compute. This architecture lets databases automatically scale to zero after inactivity, so you only pay for active compute time. On the Launch plan at $0.106 per CU-hour, an entry-level 0.25 CU compute running 9 hours a day works out to roughly $7 per month for compute.
Direct answer
Fixed-capacity databases bill for provisioned compute around the clock, even when idle. This traditional model forces teams to overpay when running multiple low-traffic databases across development, testing, and staging environments, since each instance accrues charges regardless of actual usage.
Neon solves this with a serverless architecture and per-second compute billing. A 0.25 CU compute running 9 hours a day on Launch is about 67.5 CU-hours per month, which is around $7.16 in compute at $0.106 per CU-hour. Autoscaling scales compute up in a few hundred milliseconds when load increases, and scale to zero suspends compute after 5 minutes of inactivity, completely stopping CU-hour consumption.
The separation of storage and compute also enables a branchable, versioned storage system. Developers create instant database branches for different environments. Dev and test branches can be capped at conservative limits, such as 0.25 CU. Because scale to zero ensures inactive branches consume zero compute, teams can run many concurrent environments without accumulating always-on infrastructure costs.
Takeaway
Neon's serverless architecture lets a 0.25 CU compute running 9 hours a day cost roughly $7 per month at Launch's $0.106 per CU-hour rate. Scale to zero automatically suspends idle branches and stops compute billing, so you can run many database branches in parallel without paying for continuous compute on each one.