Neon
activeby Databricks (Neon) · since 2022 · also known as Neon Postgres, Neon Serverless Postgres
Serverless Postgres with copy-on-write branching and scale-to-zero, now part of Databricks.
Overview
Neon is a serverless PostgreSQL service that splits Postgres into stateless compute and a custom storage layer (safekeepers for the WAL, pageservers backed by object storage). That design enables instant copy-on-write branches of a whole database, point-in-time restore inside a history window, and computes that autoscale and suspend to zero when idle. It became the default Postgres for preview environments, AI app builders and agent-created databases. Databricks acquired Neon in 2025 and its technology now also underpins Databricks Lakebase. The free plan remains generous, and paid plans are purely usage-based, so it is a strong fit for many small or bursty databases; cold starts and shared-storage latency make it less ideal for steady, latency-critical OLTP.
Strengths & trade-offs
Strengths
- +Instant copy-on-write branches for preview environments, tests and agents
- +Scale-to-zero and autoscaling make thousands of small databases cheap
- +Time-travel queries and instant restore within the history window
- +Serverless HTTP/WebSocket driver works from edge runtimes
- +Real Postgres with common extensions (pgvector, PostGIS)
Trade-offs
- −Cold starts after suspend add latency to the first query
- −Storage over network/object storage is slower than local NVMe for heavy OLTP
- −Single-region projects; no multi-region writes
- −Product roadmap now tied to Databricks priorities; public repo activity has slowed
Great for
- +Per-pull-request preview databases
- +AI app builders and agents provisioning databases on demand
- +SaaS with many small, mostly idle tenant databases
- +Serverless and edge apps on Vercel or Cloudflare
Think twice for
- −Steady high-throughput OLTP with strict latency SLOs
- −Multi-region active-active writes
- −On-premises deployments
Capabilities
Native or official first-party support · 18 of 25 tracked
Scale & consistency
- Comfortable data size
- Medium · ≤ 10 TB
- Horizontal scaling
- None
- Multi-region
- No
- Transactions
- Full ACID
- Consistency
- Strong
- Schema
- Strict
- Architecture
- Serverless, Shared storage, Primary replica
- Query languages
- SQL, PL/pgSQL
Autoscaling single-writer compute with read replicas sharing storage; projects are pinned to one region.
Deployment
available on · AWS · Azure · Vercel
Ecosystem
Alternatives to Neon
Open-source Postgres backend platform: database, auth, storage, realtime and edge functions in one project.
neon vs supabase →Databricks' serverless Postgres built on Neon tech, with branching and sync to Delta tables in Unity Catalog.
neon vs lakebase →Performance-focused managed MySQL (Vitess) and Postgres, with NVMe-backed Metal clusters and branching.
neon vs planetscale →Open-source Postgres platform with storage-level copy-on-write branching, scale-to-zero and anonymized clones.
neon vs xata →Serverless Postgres from the Prisma ORM team, on unikernels, billed per operation with no cold starts.
neon vs prisma-postgres →AWS's MySQL- and PostgreSQL-compatible engine on a shared, six-way replicated storage layer.
neon vs amazon-aurora →Neon FAQ
What is Neon?
Neon is a relational database by Databricks (Neon), first released in 2022. Neon is a serverless PostgreSQL service that splits Postgres into stateless compute and a custom storage layer (safekeepers for the WAL, pageservers backed by object storage). That design enables instant copy-on-write branches of a whole database, point-in-time restore inside a history window, and computes that autoscale and suspend to zero when idle. It became the default Postgres for preview environments, AI app builders and agent-created databases. Databricks acquired Neon in 2025 and its technology now also underpins Databricks Lakebase. The free plan remains generous, and paid plans are purely usage-based, so it is a strong fit for many small or bursty databases; cold starts and shared-storage latency make it less ideal for steady, latency-critical OLTP.
Is Neon open source?
Yes. Neon is open source. License: Apache License 2.0 (storage engine).
Does Neon support ACID transactions?
Yes. Neon supports full ACID transactions, with strong consistency.
Does Neon scale horizontally?
Neon does not scale horizontally; it runs on a single node, so you scale it up rather than out. It is comfortable at medium scale (≤ 10 TB).
How can I run Neon?
You can use the vendor's managed cloud and run it serverless. It is available on AWS, Azure and Vercel.
What is Neon best for?
Neon is a strong fit for per-pull-request preview databases; AI app builders and agents provisioning databases on demand; and SaaS with many small, mostly idle tenant databases. Think twice for steady high-throughput OLTP with strict latency SLOs.
What are the best alternatives to Neon?
Common alternatives to Neon include Supabase, Databricks Lakebase, PlanetScale, Xata and Prisma Postgres. The right choice depends on your workload, scale and hosting; WhichDB's smart match ranks them against your requirements.
How popular is Neon?
Neon ranks #86 of 844 databases on the WhichDB Score (30.4/100), and #26 among relational databases. The score combines public adoption signals such as package downloads, GitHub activity, job posts and developer discussion.