Neon

Serverless PostgreSQL database with support for auto-scaling and on-the-fly branching

cloud services

https://neon.com/

Visit Website

In-depth Report

Neon Serverless Postgres is a serverless PostgreSQL database service developed by Neon Corporation and now part of Databricks. The platform achieves automatic expansion, instant database branching and unlimited storage capabilities through an innovative storage and computing separation architecture. Neon combines the reliability of traditional PostgreSQL with the flexibility of cloud native, providing developers with a new database experience. With more than 150,000 new Postgres compute endpoints added every day, Neon has become one of the most talked about products in the serverless database space.
Neon was founded by PostgreSQL contributors and technical veterans with decades of database domain expertise. In May 2025, Neon was acquired by Databricks, further strengthening its enterprise-level capabilities and social recognition. From the perspective of technical architecture, Neon adopts a design idea similar to Amazon Aurora, separating the storage layer from the computing layer, and replacing the traditional PostgreSQL storage layer with a distributed storage system. This innovative architecture enables Neon to achieve a truly serverless experience without compromising data durability. Unlike traditional PostgreSQL hosting services, Neon's computing resources can automatically scale up and down based on workload, and can be completely suspended (Scale to Zero) when idle, enabling pay-per-use cost optimization. The core design philosophy of Neon is to make the database as easy to use as cloud functions while retaining all the functional features of PostgreSQL. This positioning makes it particularly suitable for modern application development scenarios, especially entrepreneurial projects that require rapid iteration and enterprise-level applications that require elastic expansion.
Neon offers several innovative features that are hard to find in traditional PostgreSQL hosting services. In terms of automatic expansion, Neon's computing resources can automatically expand CPU, memory, and storage to adapt to workload changes without manual configuration. According to official data, this feature prevents approximately 54,000 performance degradation events per day. For developers, this means no need to worry about database bottlenecks caused by sudden traffic surges, and no need to pay extra for idle periods. Instant Branching is one of Neon’s most distinctive features. Developers can create editable copies of databases in less than a second through Copy-on-write technology. This is especially practical for development and testing scenarios, where team members can each have independent database branches for experimentation without affecting production environment data. In addition, the branch function also supports data anonymization (Anonymization), which can use virtual values to mask sensitive data, and set the expiration time to automatically clean up temporary branches. In terms of developer experience, Neon provides complete CLI tools and API support. Database creation, connection and branch management can be completed through simple commands. The platform supports the establishment of private network connections through PrivateLink to ensure the security of data traffic. Logs and metrics can be exported to Databricks or any OpenTelemetry-compatible monitoring service, making it easier for enterprises to conduct unified operation and maintenance management. Neon also pays special attention to AI application scenarios. The platform provides ideal backend support for AI Agents, allowing developers to deploy thousands of databases and automatically pause when idle. The millisecond-level database creation speed allows AI Agent to dynamically create and destroy database instances to achieve a truly elastic architecture. In terms of enterprise-grade features, Neon offers HIPAA and SOC2 compliance certification and supports advanced features such as private networking, uptime SLA (99.95%), point-in-time recovery, and single sign-on. It is worth noting that these enterprise-level features are not charged separately and are open to all users.
Neon adopts a purely pay-as-you-go serverless pricing model with no platform fees and no monthly minimum consumption restrictions. The free plan provides a basic usage limit and unlimited team members can join, which is very friendly to small teams and start-ups. Neon's cost advantage over traditional PostgreSQL hosting services like AWS RDS, Aurora, or Supabase is its true Scale-to-Zero capabilities. When the database is idle, computing resources are completely suspended and no charges are incurred. Traditional services usually require minimal computing resources to be reserved, and charges will be incurred even during periods of no traffic. However, it’s important to note that Neon’s storage costs are calculated independently, as its decoupled storage architecture requires ongoing maintenance of copies of data. For applications with large storage requirements, this cost needs to be evaluated.
According to data from multiple third-party review platforms, Neon enjoys a high reputation among the developer community. Users generally recognize its innovative architecture and excellent development experience. In particular, the instant branch function has been widely praised. Positive comments mainly focus on the following aspects: the instant branch function greatly improves the efficiency of development and testing, and is especially suitable for projects that require frequent creation of test environments; automatic expansion and scale-to-zero capabilities significantly reduce database costs; it is fully compatible with PostgreSQL, and existing PSQL tools and ORM libraries can be used directly; the free plan has a generous quota, and small teams can start at zero cost. Negative comments mainly focus on the following aspects: As a relatively new product, Neon's documentation and community support still have room for improvement compared to mature services such as AWS; some users reported that the performance under extreme loads is different from traditional hosting services; the transparency of storage costs needs to be improved. Reviews of Neon in the industry press have been generally positive. Many reviews consider Neon "one of the best choices for serverless Postgres" and is particularly suitable for modern application development and AI application scenarios.
Professional review websites have given high recognition to Neon’s technical architecture. CSDN technology blog analysis pointed out that Neon provides a revolutionary solution to the pain points of traditional databases through an innovative storage and computing separation architecture. Compared with competing product Supabase, Neon focuses more on the database itself, while Supabase integrates more functions (such as user authentication, file storage, etc.), and the two have slightly different positioning. AdTools' in-depth review article pointed out that Neon will still be the most noteworthy choice among the "most hyped serverless platforms" in 2026. Its branching capabilities and automatic expansion capabilities lead the industry.
Despite Neon's performance, there are still some potential risks to be concerned about. The first is the risk of data lock-in. Since Neon was acquired by Databricks, the long-term direction of the product is likely to be aligned with Databricks' strategic priorities rather than evolving independently. This may result in feature adjustments or pricing changes. Second is the cold start delay. While Neon claims to support millisecond database creation, there may be a cold start delay of several seconds when reactivating compute resources after long periods of inactivity, and latency-sensitive applications need to evaluate this impact. The third is maturity consideration. Compared with mature services such as AWS RDS or Google Cloud SQL, Neon's enterprise-level use cases and long-term stability data are relatively limited. Database selection for critical business systems requires careful evaluation.
Neon is particularly suitable for the following scenarios: start-ups and small teams need a low-cost PostgreSQL database to start; development and testing environments need to frequently create and destroy database copies; AI applications and Agent projects need a database backend that can be quickly and elastically expanded; multi-tenant applications that need branch functions to support data isolation; developers who pursue a modern development experience and want to reduce the burden of operation and maintenance. It is recommended to consider carefully for the following scenarios: real-time transaction systems that require extremely low latency; business-critical systems that have extremely high requirements for data durability; enterprises that have already made extensive use of the AWS or Google Cloud ecosystem.
Neon Serverless Postgres represents the development direction of serverless databases. Its innovative storage and computing separation architecture, instant branching function and true scale-to-zero capability provide a new database experience for modern application development. As a member of the Databricks family, Neon is guaranteed enterprise-grade capabilities and long-term growth. For teams pursuing development efficiency, cost optimization, and modern architecture, Neon is an option worth considering. Its free plan lowers the trial threshold so developers can quickly start experiencing its core features. Of course, database selection for key business systems still requires a comprehensive assessment based on specific needs and risk tolerance.

User Reviews

MeganPowell_2020

—

Neon's branching function is really great. Each PR automatically creates a database branch and deletes it directly after testing. There is no need to worry about contaminating production data. Coupled with Vercel's automatic integration, this is the right way to start modern development.

CatherineRoberts8

—

I ran a small project using Neon, and the free version is enough. 512MB is really enough for most hobby projects.

NOeva

—

After migrating from GCP Cloud SQL to Neon, the bill dropped from $80 per month to 0. Scale-to-zero is a real savings.

Jessica_Henderson_X

—

Cold start is a real pitfall. Sometimes you have to wait two or three seconds for the first query. If you are running a user-facing API, it is recommended to go directly to the Launch plan to turn off automatic suspension.

Jeremy.NguyenX7

—

Neon's serverless driver is so comfortable to work with Vercel Edge Functions. HTTP queries do not need to maintain TCP connections.

MeganPedersen

—

Use Neon as a CI/CD test database. Each PR automatically creates a branch and runs integration tests. There is no need to manually maintain seed data, and the development experience is greatly upgraded.

HannahBarnesIII12

—

The free version is 190 hours/month, and an always-on app can be used up in eight days. Fortunately, scale-to-zero makes the development environment almost non-consuming.

LBaker

—

Compared with Supabase, Neon is a pure database. But if you already use another authentication service, Neon's branching function is a win.

Tyler.Mitchell168

—

It smells so good that you can’t go back after using it.

WThompson_2021

—

After Neon was acquired by Databricks, storage dropped by 80% and is now $0.75/GiB per month, which is much cheaper than before. However, the long-term direction after the acquisition is still a bit worrying. If it becomes an exclusive component of Databricks one day, it will be troublesome. After all, long-term stability must be considered when making technology selections.

MrVivienVidal_x

—

After running the production environment for five months, RDS costs 43 dollars a month, and the Neon Launch plan costs 19 dollars. The branch function reduces migration accidents from three times to zero, and the ROI is simply outrageous.

SAkel

—

There is no problem in docking with Prisma, but be careful to use pooled connections, otherwise the number of connections will explode if the concurrency is high in the serverless scenario. In addition, the latency of Drizzle ORM and HTTP driver on Cloudflare Workers is only about 80ms. This combination is recommended.

JustinLee_X12

—

The PPT is stuck. It’s really cold when it gets cold.

TBrown_77

—

Neon's pgvector supports RAG, which is very suitable. Vector retrieval and PostgreSQL can be done in one stop, without the need for an additional vector database. Moreover, the serverless architecture is naturally suitable for the elastic needs of AI startups, and pay-as-you-go does not waste resources.

es4pwimmo

—

After using Neon, the development process has completely changed. Each functional branch has an independent database environment, and there is no need to compete with colleagues for a dev database. It is perfect when used with Vercel preview deployment to automatically assign branches to each PR.

HashWave007

—

The free version has less storage and less projects, but it’s enough to get you started, and it’s not too late to upgrade when your business picks up.

51zhgncriy

—

Neon's PITR is a basic function without extra charge, which is much better than RDS. Once I ran a wrong migration and directly restored it to a few minutes ago, it saved my life. The PITR of RDS also costs more, and the gap is too big.

SRamirez_2021

—

It's free, and everyone who has used it says it smells great.

JesseClark520

—

Serverless driver's HTTP query is about 30%-50% slower than direct postgres connection. This gap is quite obvious in high-frequency query scenarios. It is recommended to use TCP connection directly for latency-sensitive APIs. The HTTP driver is suitable for edge computing but comes at a cost in performance.

秋叶_13

—

A 12GB production repository branch is created in less than a second, and the copy-on-write is awesome. Each developer in our team now has his own independent branch, which is completely isolated and does not interfere with each other.

TAlvarez_2020

—

Vercel + Neon are a perfect match, automatically assigning independent database branches to preview deployment, and you can see the preview environment with real data after the front-end modification. After the PR is reviewed and merged, the branches are automatically cleaned up, so there is no need to manually maintain the test database anymore. It’s so smooth.

KWrightIII

—

The two outages in May added up to 5.5 hours, which is a bit unsafe as a database service. Although the SLA promises 99.95%, continuous problems are too painful for the production environment.

ANr_ui

—

An automatic refetch on the front end ate up most of the compute time without turning it off properly. Later, I adjusted all polling intervals to more than 5 minutes to solve the problem. This is a bloody lesson.

CooperJensen

—

Neon yyds, whoever uses them knows.

HJenkins_2020

—

There are no nodes in China, and latency is a problem for Asian users. The delay in Singapore is acceptable, but it is incomparable with domestic cloud vendors. If the target users are mainly domestic, Neon may not be the best choice.

Michelle.Patel124

—

Create a branch directly from the production database to run complex analysis queries without affecting online business at all. Our data team now relies on this to do ad-hoc analysis every week, and we no longer have to do read replicas separately.

Austin_Sanchez369

—

I was fooled by sslmode=require at first. Sequelize does not use SSL by default, causing the connection to fail. Just configure it and it will be fine. Also note that the environment variable name on Vercel must be DATABASE_URL instead of POSTGRES_URL. It is easy to confuse the two.

AThompson369

—

The autoscaling is very smooth, with 0.25 vCPU during low peak periods and automatically expanded to 2 vCPU as soon as the traffic comes. Compared with the previous RDS manual upgrade instance, the experience is completely different.

翡翠_12

—

190 compute hours cannot be shared across projects. Each project is calculated independently. If one is used up and the other is used up, it will be useless. It would be great if it could be pooled.

梁然飞

—

I can't go back, I really can't go back.

Neon

In-depth Report

User Reviews

Similar Products