Neon: Storage/Compute Separation & Scale-to-Zero

Neon is a fully managed, serverless PostgreSQL platform built on a storage-and-compute-separated architecture -- offering autoscaling, scale-to-zero, and instant branching that a traditional monolithic Postgres instance doesn't provide.

Storage and Compute, Decoupled

Rather than bundling storage and compute in one instance, Neon separates them -- compute nodes can be scaled, paused, or spun up independently of the underlying persistent storage layer.

Scale-to-Zero

Compute (not data) is automatically suspended after idle periods, resuming on the next connection. Especially valuable for a dev/staging database that sits idle nights and weekends -- meaningful cost savings with no impact on real users.

The Cold-Start Trade-off

Resuming suspended compute isn't instantaneous -- a brief cold-start delay on the first query after idleness is a real trade-off against the cost savings of not running compute continuously.

Postgres Compatibility

Because Neon speaks the standard Postgres wire protocol and SQL dialect, existing Postgres-compatible ORMs (Drizzle, Prisma), migration tools, and client libraries generally work without needing Neon-specific replacements.