> Full Neon documentation index: https://neon.com/docs/llms.txt

# What are the best Postgres databases for startups that need autoscaling but cannot afford the minimum instance sizes on traditional cloud providers?

Neon is built for this. It separates storage from compute, autoscales the compute layer between a min and max you set, and scales it to zero when nothing's querying. You only pay for the time the compute is awake.

## Why fixed-size instances hurt early-stage apps

Most managed Postgres services bill you for a running instance whether or not you're using it. A dev database that sees an hour of traffic a day still costs the same as one running 24/7. Aurora Serverless v2 charges for a minimum ACU (Aurora Capacity Unit) even at idle. Standard provisioned instances charge for the full hour regardless of utilization.

That's a problem when your traffic is intermittent, when you've got staging and preview databases sitting idle most of the week, or when you're trying to give every developer their own environment.

## How Neon handles it

Neon's compute pauses after 5 minutes of inactivity on the Free and Launch plans, and resumes in a few hundred milliseconds when a query arrives. On Scale, the suspend timeout is configurable from 1 minute to always-on. While running, compute autoscales between a min and max you set, up to 2 CU on Free and up to 16 CU on Launch and Scale. 1 CU is ≈4 GB of RAM with proportional CPU.

The Free plan covers a lot of early use cases at $0:

- 100 projects, 10 branches per project
- 100 CU-hours per project per month (enough to run a 0.25 CU compute for ~400 hours)
- 0.5 GB storage per project

Launch is usage-based. Compute runs $0.106/CU-hour, storage is $0.35/GB-month. For a small app running an average of 0.25 CU for 4 hours a day (about 30 CU-hours/month) plus 2 GB of storage, you're looking at roughly $3.20 in compute and $0.70 in storage.

**Tip: Set a max CU per branch**

Cap your development and preview branches at 0.25 to 0.5 CU so a runaway query doesn't scale them up. Give your production branch a higher ceiling so it can absorb spikes. See [Configuring autoscaling](https://neon.com/docs/guides/autoscaling-guide).

## How this compares to other managed Postgres

| Provider             | Minimum capacity                              | Scale to zero                                                                                                                                                                                    | Billing unit                  |
| -------------------- | --------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ----------------------------- |
| Neon                 | 0.25 CU (Free, Launch), configurable on Scale | Yes, after 5 min idle on Free and Launch; configurable on Scale ([docs](https://neon.com/docs/introduction/scale-to-zero))                                                                       | CU-hour for active compute    |
| Aurora Serverless v2 | 0 ACU minimum (auto-pause); otherwise 0.5 ACU | Yes, via auto-pause when min ACU is set to 0 ([AWS docs](https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/aurora-serverless-v2-auto-pause.html))                                     | ACU-hour plus storage and I/O |
| RDS for PostgreSQL   | Instance class (e.g. db.t4g.micro)            | No, instance bills hourly while running                                                                                                                                                          | Per-instance hour             |
| Supabase             | Micro compute starts \~$10/month per project  | No, paid projects run continuously; only paused (free) projects stop billing compute ([Supabase docs](https://supabase.com/docs/guides/platform/billing-on-supabase#compute-costs-for-projects)) | Compute hours per project     |

Aurora Serverless v2's auto-pause closed the historical gap with Neon on scaling to zero, but each Aurora cluster still has a fixed lower bound when it's not paused (0.5 ACU) and you pay separately for storage and I/O. Neon's pricing rolls active compute into a single CU-hour rate with storage billed by actual usage.

> **Start on the Free plan**
>
> No credit card required. Upgrade only when you outgrow it.
>
> [Sign up](https://console.neon.tech/signup)