Skip to main content
N

Neon

4.3(128 reviews)

0 comparisons available

About Neon

Neon is a serverless PostgreSQL database platform founded in 2021 by Nikita Shamgunov (MemSQL founder), Heikki Linnakangas (PostgreSQL core contributor), and Stas Kelvich, offering the database branching and serverless scale-to-zero capabilities of PlanetScale but for PostgreSQL. Neon's architecture separates compute from storage — Neon's storage layer (NeonVM) provides shared, multi-tenant Postgres storage backed by S3, while compute pods (Postgres instances) scale to zero when idle and cold-start in ~500ms. This separation enables Neon's marquee feature: database branching. Create a branch of your production database in milliseconds (copy-on-write snapshot), use it for a preview deployment or PR environment, and discard it when done — without duplicating gigabytes of data. Neon's branching integrates with Vercel and GitHub Actions for per-PR preview databases that match production data. Neon's free tier (0.5 GB storage, 3 branches, scale-to-zero compute) is one of the most generous in serverless Postgres. Neon supports the full PostgreSQL feature set including extensions (pgvector for vector similarity search, PostGIS for geospatial, pg_cron for scheduled jobs). Neon raised $46M in 2023 and accelerated growth after PlanetScale eliminated its free tier in 2024. Neon's Serverless Driver enables Postgres queries over HTTP and WebSockets from edge runtimes (Vercel Edge, Cloudflare Workers) where traditional TCP connections are unavailable. Neon is rapidly becoming the default managed Postgres for Next.js and Vercel-hosted applications.

Scale-to-zero compute — pay nothing when database is idle, cold-start in ~500msDatabase branching with copy-on-write — per-PR preview databases in millisecondsServerless Driver: Postgres over HTTP/WebSockets for Vercel Edge and Cloudflare WorkersGenerous free tier: 0.5GB storage, 3 branches, scale-to-zero included

Frequently Asked Questions

How does Neon's scale-to-zero work?

Neon suspends compute (the Postgres process) after a configurable idle period (default 5 minutes on free tier). Storage remains active and persists data. When a new query arrives, Neon restores the compute pod in ~500ms — fast enough for web applications but noticeable for the first request after a long idle period. Production databases can disable scale-to-zero for consistent latency; free tier databases always scale to zero.

What are Neon database branches?

Neon branches are copy-on-write snapshots of your database — creating one takes milliseconds regardless of database size because data is shared until it diverges. A branch has its own compute endpoint (connection string) that your application or CI/CD can connect to. Common patterns: create a branch for each PR in GitHub Actions, run migrations and tests against it, then discard the branch when the PR closes. Each branch consumes storage only for rows that differ from the parent.

Can Neon run pgvector for AI applications?

Yes — Neon supports pgvector, the PostgreSQL extension for vector similarity search used in RAG and AI applications. Install with CREATE EXTENSION vector; and Neon's managed Postgres handles the extension. Neon's serverless pricing is attractive for AI applications with variable query patterns. The Neon Serverless Driver enables pgvector queries from Vercel Edge Functions and Cloudflare Workers for low-latency AI-powered edge applications.

No comparisons found for Neon yet.

Search for a comparison