Neon logo
Hosting & Infrastructure

Neon

Serverless Postgres with branching

What it does

Neon is a serverless Postgres database platform. Full standard Postgres (compatible with any Postgres driver, tool, or extension), but with separated compute and storage that scales to zero when idle, plus git-like database branching (create a branch of your entire database in seconds, use it for a preview deploy or a test, throw it away). Acquired by Databricks in 2025 for around $1B.

Free tier

Yes

Starts at

$19/mo

Setup

easy

Key features

Serverless PostgresDatabase BranchingAutoscalingPoint-in-Time RestoreAPI

Integrates with

Vercel, Netlify, Cloudflare, Prisma

Who this fits

  • SaaS teams needing per-preview-environment databases (Neon branching + Vercel preview deploys is the killer combo)
  • Cost-conscious startups whose databases sit idle overnight (scale-to-zero saves money)
  • Developers using serverless platforms (Vercel, Cloudflare Workers, Lambda) that need a database that fits the serverless model
  • AI apps needing pgvector (Neon supports it natively) for embedding storage
  • Teams migrating off Heroku Postgres, RDS, or Supabase for the branching workflow

Who it does NOT fit

  • Apps that require constant sub-10ms latency (cold starts on scale-to-zero add latency)
  • Very large workloads (100TB+, hundreds of thousands of QPS) that outgrow the serverless model
  • Teams needing MySQL, MongoDB, or other engines (Neon is Postgres only)
  • Teams needing bundled auth + storage + realtime (use Supabase instead; also Postgres, but broader BaaS)

Pricing

- **Free:** 1 project, 10 branches, 0.5 GB storage, autoscaling compute up to 0.25 vCPU, scale-to-zero after 5 minutes idle. Enough for hobby projects and demos. - **Launch:** $19/month base, 10 GB storage included, 300 compute hours, more branches, IP allow list. - **Scale:** $69/month base, more storage and compute, higher branch limits, added features (logical replication, backups). - **Business/Enterprise:** custom, adds SLA, SSO, dedicated support, higher limits, HIPAA on request. - Since Databricks acquisition, expect deeper integration with Databricks tools and possible pricing adjustments.

Real gotchas

  • Scale-to-zero means the first query after idle triggers a cold start (typically under 1 second on Launch and higher, longer on Free). This is fine for most apps but breaks the assumption of sub-10ms P99 latency for the first query of a session [source: neon.com/docs/introduction/scale-to-zero].
  • Connection limits scale with compute size. Serverless environments (Vercel Functions, Lambda) can exhaust connections; use Neon's connection pooler (PgBouncer-based) or a serverless driver.
  • Branching is not backup. If you drop a table on main, you lose it. Use point-in-time restore (paid tiers) for actual recovery.
  • Databricks acquisition (2025) may change roadmap and pricing over the next 12-24 months; verify current plans if committing long-term.
  • pgvector index build times can be slow on Free tier's small compute; upgrade for AI workloads.

Integrations that matter

Vercel (native integration for preview branch per PR), Cloudflare Workers, AWS Lambda, Netlify Functions, Prisma, Drizzle, all standard Postgres drivers, Docker, Kubernetes, Terraform, GitHub Actions, Datadog, Grafana, pgvector (native), PostGIS, all standard Postgres extensions.

Otto’s take on this tool

Recommend Neon for developers building SaaS on serverless stacks (Next.js on Vercel, Cloudflare Workers, Lambda) who want per-preview-branch databases and scale-to-zero cost efficiency. Recommend Supabase when the client also wants auth, storage, realtime bundled. Recommend PlanetScale for MySQL fans. Recommend AWS RDS or GCP Cloud SQL when the client needs enterprise-managed traditional Postgres.

Affiliate disclosure: if you sign up through the link above, Ottomately may earn a referral commission at no cost to you. We only feature tools we have actually used or vetted through our recommendation engine.

More in Hosting & Infrastructure

Not Sure Neon is Right?

Otto can pick for you.

30 seconds, one form, get a personalized stack with pricing and integration paths.

Get my recommendations