Neon

Neon is a managed, cloud-hosted PostgreSQL service for teams that need elastic database compute and fast isolated branches. It suits growing applications with uneven demand especially well; organizations requiring production-grade availability and compliance should consider the higher Scale tier and its usage charges carefully.

Overview

Neon is a relational database service with JSON support, high availability, read replicas and point-in-time recovery. Its compute can autoscale with workload and suspend after inactivity, then resume when activity returns. That model can reduce always-on resource use for intermittent applications, while cloud-only deployment rules out teams that must host the database on their own infrastructure.

Neon says it joined Databricks in May 2025 and is now part of the Databricks Platform. Data at rest is encrypted with AES-256, and data in transit uses TLS 1.2 or higher. Neon undergoes annual SOC 2 Type II, ISO/IEC 27001:2022 and ISO/IEC 27701:2019 audits and states that it has achieved HIPAA compliance.

Key features

Elastic compute

Autoscaling adjusts database compute to demand, while scale-to-zero suspends compute during inactivity and restarts it when activity resumes. This is a strong fit for development environments, prototypes and workloads with quiet periods. The free allowance still limits sustained use to 50 CU-hours per month per project, so consistently busy services will move quickly into paid consumption.

Branching and pooling

Isolated database branches can be created in about a second and share data through copy-on-write storage. That combination supports separate development or testing environments without duplicating all underlying data. Neon also supports PgBouncer connection pooling, useful when applications open many short-lived connections.

Management and integrations

An API manages projects and databases programmatically. Neon documents workflows with Vercel, Prisma, Drizzle ORM, GitHub Actions and its CLI, giving teams established integration paths for application deployment and automation. Support is directed through Neon's Discord community.

Pricing

Free

The Free plan costs nothing and includes 10 projects, 50 CU-hours per month per project, 0.5 GB storage per project, 10 branches per project and 5 GB monthly egress. It is appropriate for learning, prototypes and small low-traffic applications. The per-project compute and storage ceilings make it unsuitable for a production workload with sustained activity or larger datasets.

Launch

Launch is priced at $0.11 a month, billed at $0.106 per CU-hour and $0.35 per GB-month of storage, with usage-based billing and no monthly minimum. It includes 10 branches per project, autoscaling up to 16 CU and 100 GB monthly egress; branches beyond the included amount cost $0.002 per branch-hour. Neon describes this tier for growing projects and early startups. Compared with Free, it removes the small fixed allowances in favor of metered capacity, so costs track usage but require monitoring.

Scale

Scale is priced at $0.22 a month, billed at $0.222 per CU-hour with usage-based billing and no monthly minimum. It supports up to 1,000 projects, includes a 99.95% SLA-backed uptime commitment, private networking, IP allow lists, 25 branches per project and a 30-day Instant Restore window. It is intended for production workloads that need stronger security, availability and compliance controls. The higher CU rate makes it harder to justify for small or experimental deployments.

Platforms

Neon is offered through api, Linux, macOS, web and Windows platforms. The deployment model is cloud and the service is managed, rather than self-hosted.

Who it's for

Choose Neon for a startup or application team that wants PostgreSQL with autoscaling, scale-to-zero behavior, programmatic administration and rapid database branches. It is particularly well suited to variable workloads and development workflows that benefit from isolated environments. Scale is the relevant option for production teams needing private networking, IP allow lists, an uptime SLA and a longer restore window.

Look elsewhere if self-hosting is mandatory, if predictable flat pricing is more important than consumption billing, or if the Free plan's project, storage, compute and egress caps cannot support your workload.

Pros and cons

• Pros: Autoscaling and suspend-on-idle behavior accommodate uneven demand.
• Pros: Fast copy-on-write branching creates isolated database environments without full data duplication.
• Pros: API access, PgBouncer pooling and documented Vercel, Prisma, Drizzle ORM, GitHub Actions and CLI workflows support automation.
• Pros: Scale adds a 99.95% SLA, private networking, IP allow lists and a 30-day Instant Restore window.
• Cons: Free limits are tight at 50 CU-hours and 0.5 GB storage per project, with 5 GB monthly egress.
• Cons: Launch and Scale use consumption billing, making monthly spend less predictable as compute, storage or branch use grows.
• Cons: Cloud-only deployment cannot meet an on-premises or self-hosted requirement.

Alternatives

For a broader directory of Relational Databases, compare other relational services when Neon’s cloud-only, usage-based model does not fit. The Data Version Control Tools list is more appropriate when versioning datasets, rather than running a managed PostgreSQL database, is the central need.

git-annex is free software for managing large files without storing their contents in Git, so choose it for file-oriented versioning instead of a relational database. Weights & Biases fits machine-learning experiment tracking, with a free tier offering five model seats, 5 GB monthly storage and 1 GB monthly Weave ingestion. Apache Iceberg is a free Apache-licensed table format for data-lake workloads. Apache Paimon is a free open-source data-lake platform with selectable connectors and storage plugins.

ClearML is an alternative for teams seeking an open-source self-hosted machine-learning platform. Dolt is a free, open-source database with Git-like versioning, while its DoltHub Pro plan costs $5 a month for private data. Nile runs entirely on a machine without a cloud account, making it preferable for local data version control. DataChain offers local open-source compute for a single developer and cloud-storage backends, with a Teams plan for collaborative use.

Verdict

Neon is a strong choice for teams building cloud applications that need PostgreSQL, elastic compute and near-instant isolated branches. Its main reason to choose is the combination of scale-to-zero behavior, branching and automation-friendly management. Look elsewhere when self-hosting or fixed, predictable costs are essential, or when the Free plan’s low storage and compute allowances cannot cover real usage.