Neon

Neon is a serverless PostgreSQL database acquired by Databricks in May 2025 for approximately $1 billion.

Reviewed by 7wData
API Available

On this page

Publisher review

Neon is a serverless PostgreSQL database acquired by Databricks in May 2025 for approximately $1 billion. It separates compute from storage, allowing databases to pause when idle and wake in 300–500 milliseconds. The platform is engineered for developers building AI agents and multi-tenant applications, offering a familiar Postgres interface with modern cloud-native features like database branching (instant test copies using copy-on-write), point-in-time recovery, and HTTP API access.

After Databricks' acquisition, Neon announced significant price reductions: compute costs dropped 15–25%, storage fell 80% to $0.35/GB-month, and the free tier doubled to 100 CU-hours monthly. The architecture differs fundamentally from traditional managed Postgres—instead of collocating data and compute, Neon routes reads and writes through a custom storage layer, introducing single-digit millisecond latency overhead. For bursty workloads (development databases, preview environments, or databases serving hundreds of idle tenants), this overhead is negligible.

For sustained, latency-sensitive OLTP systems, always-on compute still costs less per month than the serverless premium. The platform restricts superuser access, supports 80+ PostgreSQL extensions (excluding those requiring filesystem access or background workers), and automatically removes inactive replication slots after 40 hours. Developer feedback is consistently positive on developer experience and cost control, though cold-start sensitivity varies by use case.

Get the AI & data signal, daily.

48k+ subscribers read this every morning. One email, both newsletters. Unsubscribe anytime.

How it works

  1. Scale-to-zero compute

    Compute suspends after five minutes of inactivity and resumes in 300–500 ms on the next query, eliminating idle infrastructure costs.

  2. Database branching

    Create instant, isolated database copies using copy-on-write semantics for testing, schema migration, and preview environments without operational overhead.

  3. Storage-compute separation

    Independent scaling of compute and storage via a custom multi-tenant storage layer, enabling cost-effective multi-database architectures.

  4. Logical replication

    Stream data changes to external Postgres instances or data sinks via pgoutput and wal2json decoder plugins; slots auto-expire after 40 hours.

  5. Point-in-time recovery

    Restore databases to any point in time without manual backup management or downtime.

  6. REST API and connection pooling

    Query Postgres via HTTP and benefit from built-in connection pooling to prevent resource exhaustion in serverless environments.

  7. Instant read replicas with autoscaling

    Spawn read replicas on demand and enable autoscaling independently on replicas for read-heavy workloads.

Strengths and trade-offs

Strengths

  • Dramatic cost advantage for bursty, multi-tenant, or development workloads; RDS Aurora costs 3–4× more for idle databases.
  • Database branching with copy-on-write semantics accelerates testing and preview environments without schema synchronization overhead.
  • Pricing transparency: 15–25% compute reductions and 80% storage reductions post-acquisition; usage-based billing with zero minimums on free tier.

Trade-offs

  • Cold-start latency (500 ms to 1 second) unsuitable for real-time systems; warm latency adds 3–10 ms overhead versus co-located Postgres due to storage layer network hop.
  • Operational constraints: no true superuser access, 80+ extensions supported (excluding background workers and custom C extensions), and logical replication slots auto-expire after 40 hours.
  • For sustained, always-on workloads, dedicated instances prove cheaper per month because the serverless per-hour premium becomes uneconomical without idle periods; compute costs $0.106–0.222/CU-hour depending on tier.

Pricing context

Neon offers a freemium model: the Free tier includes 100 CU-hours monthly and 0.5 GB storage per branch at no cost, ideal for prototypes. Launch tier begins at $5/month minimum with compute at $0.106/CU-hour and storage at $0.35/GB-month. Scale tier charges $0.222/CU-hour (higher for SOC 2 Type 2 and HIPAA compliance) with the same storage rate.

A CU-hour represents compute size multiplied by hours running; for example, a 0.25 CU running 4 hours equals 1 CU-hour. Additional branches beyond plan limits cost $1.50/branch-month. The platform provides organization-level spending caps and alerts at 80% and 100% utilization. Post-Databricks integration (May 2025), the company reduced pricing across all tiers and doubled the free compute allowance.

Getting started with Neon

  1. Sign up for Neon

    Go to neon.com and click "Sign up". Create an account using your email or a GitHub, Google, or GitLab account. After verifying your email, you will land in the Neon console with a free tier automatically activated.

  2. Create your first project

    In the Neon console, click "New Project". Give it a name and select a region. Neon provisions a serverless PostgreSQL database with compute and storage separated. Your database connection string is displayed immediately.

  3. Connect your application

    Copy the connection string from the project dashboard. Use it in your application's database client (e.g., psql, Prisma, or a language driver). For serverless environments, enable connection pooling by appending the pooler host to the string.

  4. Create a database branch

    In the console, select your project and click "Branches". Click "New Branch" and choose a parent branch. Name it (e.g., "feature-test"). Neon instantly creates a copy-on-write copy for isolated testing without extra storage cost.

  5. Set up spending alerts

    Navigate to "Organization" > "Spending Limits". Enable alerts at 80% and 100% of your budget. Set a monthly spending cap to prevent unexpected charges. This ensures cost control as your usage scales.

Frequently Asked Questions

What is Neon serverless PostgreSQL and how does it work?

Neon is a serverless PostgreSQL database that separates compute from storage. It pauses databases when idle and wakes them in 300–500 milliseconds. This architecture reduces costs for bursty workloads by eliminating idle infrastructure charges.

How much does Neon cost after the Databricks acquisition?

Post-acquisition, Neon cut compute costs by 15–25% and storage by 80% to $0.35/GB-month. The free tier doubled to 100 CU-hours monthly. Launch tier starts at $5/month with $0.106/CU-hour compute, and Scale tier at $0.222/CU-hour.

What is Neon database branching and why is it useful?

Database branching creates instant, isolated Postgres copies using copy-on-write semantics. Developers use branches for testing, schema migrations, and preview environments without operational overhead. This accelerates development workflows and eliminates synchronization delays.

Is Neon good for production workloads or just development?

Neon excels for bursty, multi-tenant, and development workloads due to scale-to-zero savings. For sustained, latency-sensitive OLTP systems, always-on compute costs less per month than the serverless premium. Cold-start latency of 300–500 ms may impact real-time applications.

What are the limitations of Neon serverless PostgreSQL?

Neon restricts superuser access, supports 80+ extensions but excludes those needing filesystem access or background workers. Logical replication slots auto-expire after 40 hours. Warm queries add 3–10 ms overhead versus co-located Postgres due to the storage layer network hop.

How does Neon compare to Amazon RDS Aurora in pricing?

Neon offers dramatic cost advantages for idle or bursty databases, with RDS Aurora costing 3–4× more for similar setups. Entry-level Neon at $7.66/month versus Aurora Serverless. For always-on workloads, dedicated instances may be cheaper per month than Neon's per-hour premium.

Alternatives

How Neon compares

Direct head-to-head against 2 competitors. Picked by 7wData.

This tool

Neon

Pricing
Neon offers a freemium model: the Free tier includes 100 CU-hours monthly and 0.5 GB storage per branch at no cost, ideal for prototypes. Launch tier begins at $5/month minimum with compute at $0.106/CU-hour and storage at $0.35/GB-month. Scale tier charges $0.222/CU-hour (higher for SOC 2 Type 2 and HIPAA compliance) with the same storage rate. A CU-hour represents compute size multiplied by hours running; for example, a 0.25 CU running 4 hours equals 1 CU-hour. Additional branches beyond plan limits cost $1.50/branch-month. The platform provides organization-level spending caps and alerts at 80% and 100% utilization. Post-Databricks integration (May 2025), the company reduced pricing across all tiers and doubled the free compute allowance.
Target
Neon is a serverless PostgreSQL database acquired by Databricks in May 2025 for approximately $1 billion.
Strength
Dramatic cost advantage for bursty, multi-tenant, or development workloads; RDS Aurora costs 3–4× more for idle databases.
Watch for
Cold-start latency (500 ms to 1 second) unsuitable for real-time systems; warm latency adds 3–10 ms overhead versus co-located Postgres due to storage layer network hop.

Amazon Aurora Serverless

Pricing
$0.12/ACU-hour (Standard) or $0.16/ACU-hour (I/O-Optimized); $0.10/GB-month storage
Target
AWS-native teams needing managed Postgres with auto-scaling and high durability
Deployment
AWS cloud only
Strength
Deep integration with AWS Lambda, S3, and other AWS services
Watch for
Cannot scale to zero; minimum 0.5 ACUs; Serverless v1 deprecated Dec 2024

Google Cloud AlloyDB

Pricing
$54.50/vCPU-month (us-east1); 2vCPU/16GB HA cluster ~$546/month
Target
Enterprise workloads needing high-performance PostgreSQL-compatible database on GCP
Deployment
Google Cloud only
Strength
4x faster transaction processing than standard PostgreSQL per Google benchmarks
Watch for
Premium pricing over Cloud SQL; no free tier; complex migration from other clouds

User reviews

No user reviews yet. Be the first to write one.

Sources

Reporting on this tool draws on these publicly available sources.

  1. www.databricks.com — Databricks acquisition of Neon in May 2025, pricing reductions (compute 15–25%, storage 80%), and free tier doubling
  2. neon.com — Current pricing tiers (Free, Launch, Scale), CU-hour rates ($0.106–0.222/CU-hour), storage cost ($0.35/GB-month), branch costs ($1.50/branch-month)
  3. neon.com — Cold-start latency (300–500 ms wake time), warm latency overhead, comparison to Aurora Serverless v2 (15 seconds)
  4. medium.com — Production trade-offs, performance overhead (3–10 ms on simple queries), logical replication slot expiration (40 hours), extension limitations, and best use cases
  5. neon.com — Logical replication support, decoder plugins (pgoutput, wal2json), DDL limitations, and replication slot auto-removal after 40 hours
  6. dev.to — Cost comparison with RDS Aurora and Supabase; entry-level Neon at $7.66/month versus Aurora Serverless; mid-range Neon 30% cheaper than Supabase