Overview
Neon is a managed Postgres service built on a design that splits storage from compute. Compute autoscales between sizes and suspends after five minutes without queries, and storage is shared copy-on-write across branches, so a branch of a large database starts at zero extra cost. Since May 2025 Neon has been part of Databricks, which it markets as Lakebase Postgres.
The Free plan is unusually wide: up to 100 projects, each with 100 CU-hours of compute a month (about 400 hours of a 0.25 CU instance), 0.5 GB of storage and 10 branches. Scale to zero cannot be turned off on Free, and hitting any monthly limit suspends compute until the next month.
Launch and Scale are both pay-as-you-go with no monthly fee. Launch bills compute at $0.106 per CU-hour and fits startups; Scale bills $0.222 per CU-hour and in exchange includes SOC 2, HIPAA, private networking, IP allow lists, metric export, a 99.95% uptime SLA and a 30-day restore window. Storage is $0.35 per GB-month on both.
Before paying, estimate how many hours your compute stays awake: an always-on database pays for every hour, so the scale-to-zero savings apply mainly to workloads with idle periods. Regions are AWS only for new projects, with Frankfurt and London in Europe.
Pricing and plans
Free tier · no card
100 projects, 100 CU-hours of compute per project per month, autoscaling up to 2 CU, 0.5 GB storage per project, 5 GB egress, 10 branches per project, 6-hour restore window, up to 60,000 auth MAU; compute suspends when a limit is hit
At the limit Pauses Compute is suspended for the rest of the month once a free limit is used; nothing is charged.
| Plan | Price | What you get |
|---|---|---|
| Free | $0 | 100 CU-hours per project per month, 0.5 GB storage per project, 5 GB egress, always scales to zero |
| Launch | $0.106/CU-hour | Storage $0.35/GB-month, 500 GB egress per project then $0.10/GB, autoscaling to 16 CU, 7-day restore window |
| Scale | $0.222/CU-hour | Adds SOC 2, HIPAA (BAA), private networking, IP allow rules, uptime SLA, 30-day restore window, up to 56 CU fixed |
What it does
- Scale to zero
- Compute suspends after 5 minutes idle (configurable on Scale from 1 minute to always on) and costs nothing while suspended.
- Autoscaling
- CPU and memory scale within a range, up to 2 CU on Free and 16 CU on paid plans; Scale also offers fixed sizes up to 56 CU.
- Branching
- Copy-on-write branches of schema and data for previews and tests, with optional anonymization and automatic expiry.
- Instant restore
- Point-in-time restore within a history window of 6 hours on Free, up to 7 days on Launch and 30 days on Scale.
- Managed Better Auth
- User authentication stored in the database, up to 60,000 MAU on Free and 1 million on paid plans.
- Data API
- Query Postgres over HTTP, alongside connection pooling, read replicas and extensions such as pgvector and PostGIS.
Strengths and limitations
Strengths
- Idle databases cost nothing for compute on every plan
- No monthly minimum on paid plans, so small projects pay cents
- Fast copy-on-write branches for preview deployments
- SOC 2, HIPAA, SLA and private networking without a fixed-fee enterprise tier
- Up to 100 free projects
Limitations
- Free plan compute suspends for the rest of the month once 100 CU-hours, 0.5 GB or 5 GB egress is used
- A cold start follows each scale-to-zero suspension
- Region is fixed at project creation; moving means creating a new project and migrating
- New projects are limited to 8 AWS regions; Azure regions are deprecated
- Always-on workloads pay for every CU-hour, which can exceed a fixed-size instance
Who it suits
Good for
- Side projects and prototypes that sit idle most of the day
- Preview environments that need a database branch per pull request
- Platforms and AI agents that create many small databases
- Startups that want usage-based Postgres with compliance options
Look elsewhere if
- Latency-sensitive apps that cannot tolerate cold starts unless scale to zero is disabled
- Teams that want Postgres bundled with storage and functions in one backend platform
Pricing history
-
$5 monthly minimum dropped
Announced December 2025 (month only; no day given): Neon stopped enforcing the $5 minimum on paid plans, making them purely usage-based. Source ↗
-
Compute price cut up to 25%
Effective on announcement: Launch compute went from $0.14 to $0.106 per CU-hour and Scale from $0.26 to $0.222. Source ↗
-
Enterprise features folded into the Scale plan
Announced: SOC 2, HIPAA, SSO, PrivateLink, SLA and point-in-time recovery became part of the usage-based Scale plan, removing a fixed surcharge. Free compute doubled from 50 to 100 CU-hours in October. Source ↗
-
Usage-based pricing replaces fixed plans
Effective 2025-08-14: paid plans became usage-based (a $5 minimum instead of the $19 Launch plan) and storage fell from $1.75 to $0.35 per GB-month. Source ↗
Alternatives to Neon
| Tool | From | Free option | Stages |
|---|---|---|---|
| Supabase Managed Postgres backend platform | $25/mo | Free tier | |
| Firebase Backend platform with document and realtime databases | Usage-based | Free tier | |
| Render Managed PaaS | $7/mo | Free tier | |
| Redis Cloud Managed Redis | $5/mo | Free tier | |
| MongoDB Atlas Managed document database | $8/mo | Free tier | |
| Convex Reactive backend and document database | $25/mo | Free tier |
Questions
Is Neon free?
Yes. The Free plan is permanent, needs no card and gives 100 CU-hours of compute and 0.5 GB of storage per project for up to 100 projects.
How much does Neon cost?
Launch is usage-based with no minimum: $0.106 per CU-hour of compute and $0.35 per GB-month of storage. Scale charges $0.222 per CU-hour and adds compliance, SLA and private networking.
Does Neon scale to zero?
Yes. Compute suspends after 5 minutes of inactivity and accrues no charges while suspended; on Free this cannot be disabled.
Who owns Neon?
Neon has been part of Databricks since May 2025, according to Neon's homepage.
Does Neon have EU regions?
Yes, AWS Frankfurt (eu-central-1) and London (eu-west-2).
Sources read for this page