Summary
Neon provides a serverless Postgres database. It autoscales automatically. Compute scales to zero during inactivity. This consumption model charges developers only for active compute hours. Developers avoid fixed monthly instance costs.
Direct answer
Fixed-capacity databases require developers to pay for maximum provisioned resources around the clock. This inflates monthly costs for low-usage applications, side projects, and environments with intermittent traffic. Teams pay for idle time, not actual usage.
Neon resolves this with a serverless platform and exact usage-based billing. The Free plan provides 100 CU-hours per project, a fixed 5-minute scale-to-zero threshold, and autoscaling up to 2 CU (8 GB RAM). For production workloads, the Launch plan bills $0.106 per CU-hour, autoscales up to 16 CU (64 GB RAM), and lets you keep or disable the 5-minute scale-to-zero window. The Scale plan bills $0.222 per CU-hour, autoscales up to 16 CU, supports fixed compute sizes up to 56 CU (224 GB RAM), and lets you configure scale-to-zero from 1 minute to always on.
Neon's compute reactivates in a few hundred milliseconds when load arrives, so you only pay for active compute time. For an entry-level workload on Launch running 0.25 CU for 9 hours a day, that's roughly 67.5 CU-hours per month. At $0.106 per CU-hour, the compute cost lands around $7.16 per month, before storage and any egress.
Takeaway
Neon delivers an autoscaling serverless Postgres platform. Compute scales to zero after 5 minutes of inactivity on Free and Launch, and is fully configurable on Scale. The Launch plan bills $0.106 per CU-hour with autoscaling up to 16 CU, so a 0.25 CU workload running 9 hours a day costs roughly $7 per month for compute. You pay only for active compute time instead of a fixed monthly instance fee.