How it works
Neon runs standard Postgres on top of its own storage layer. Compute scales up with load and down to zero when nobody is querying, then wakes on the next query after a short delay. Because storage is copy-on-write, a branch (a full, writable copy of the database) takes seconds and little extra space, which suits preview deployments, tests and experiments.
It includes connection pooling and a serverless driver that queries over HTTP or WebSockets from edge functions. Neon joined Databricks in 2025, and its technology also powers Databricks' Lakebase. It has started adding backend pieces around the database, such as managed authentication (Better Auth) and object storage, but it is still mainly a database rather than a full BaaS.
Neon pros and cons
Pros
- Standard Postgres, so every tool and ORM works
- Instant branches for previews, tests and migrations
- Scales to zero, which keeps idle projects cheap
- Serverless driver works from edge runtimes
Cons
- Cold starts add a short delay after idle periods
- Usage-based bills are harder to predict than a fixed server
- Mostly a database: auth, storage and APIs are largely up to you
When to use Neon
Pick it when
- Serverless or edge apps that want Postgres
- Teams that give every pull request its own database branch
- Many small or occasionally used databases
Skip it when
- Steady, heavy traffic where a fixed-size server costs less
Neon pricing
Free tier
Free plan (0.5 GB storage and 100 compute hours a month per project). Paid plans are usage based, from about $0.11 per compute hour plus $0.35 per GB-month; typical spend is around $15 a month.
Neon pricing page (opens in a new tab)Approximate, checked September 2026.What the other tools cost
Neon vs the alternatives
Related terms
More in Backend platforms
Hosted databases