# Neon "too many connections" / max_connections reached, explained

> Why a Neon compute runs out of connections, the exact max_connections per compute size, and the three documented fixes: close idle connections, use the pooler, or upgrade compute.

Source: https://scribase.com/neon-too-many-connections

Neon's own docs describe this as happening "when the maximum number of simultaneous database connections, defined by the Postgres max_connections setting, is reached." Unlike a fixed-size server, a Neon compute's max_connections scales with its CU size, so the limit moves depending on how the compute is configured, not just how busy it is.

## What the error means on Neon

Neon's connection-errors documentation describes the same Postgres-level message as every other Postgres host: the database has reached its max_connections ceiling. The twist on Neon is that max_connections is not one fixed number — it scales with the compute size (CU) currently assigned to the project, so a compute that autoscaled down, or that is pinned small, has a lower ceiling than the project might need under load.

## Sources

- [Neon: Connection errors](https://neon.com/docs/connect/connection-errors)

## max_connections by compute size

_Read 2026-11-18 on neon.com/docs/connect/connection-pooling_

| Compute size (CU) | RAM | max_connections |
| --- | --- | --- |
| 0.25 | 1 GB | 112 |
| 0.5 | 2 GB | 225 |
| 1 | 4 GB | 450 |
| 2 | 8 GB | 901 |
| 3 | 12 GB | 1,351 |
| 4 | 16 GB | 1,802 |
| 5 | 20 GB | 2,253 |
| 6 | 24 GB | 2,703 |
| 7 | 28 GB | 3,154 |
| 8 | 32 GB | 3,604 |
| 9–56 | 36–224 GB | 4,000 (capped) |

## Sources

- [Neon: Connection pooling](https://neon.com/docs/connect/connection-pooling)

## The three documented fixes

- Find and close long-running or idle connections — the same ORM-keeps-a-pool-warm or forgotten-cron-job causes that also drive up CU-hour usage (see the Neon cold-start guide for the same culprits).
- Use the pooled connection string (the -pooler hostname). Neon's PgBouncer accepts up to max_client_conn=10,000 client connections and multiplexes them down to default_pool_size, which is 0.9 x max_connections per user/database pair — so at 1 CU (450 max_connections) that is about 405 concurrent app connections sharing a much smaller number of real Postgres backends.
- Upgrade to a larger compute size or raise the autoscaling maximum. Each step up roughly doubles max_connections (112 → 225 → 450 → 901 ...), and a compute that scales up under load gets the higher ceiling automatically.

## Sources

- [Neon: Connection errors](https://neon.com/docs/connect/connection-errors)
- [Neon: Connection pooling](https://neon.com/docs/connect/connection-pooling)

## How Scribase prices and pools connections

Every Scribase project is $25 a month on Pro and runs behind a pooler for application traffic by default, so a burst of short-lived connections (serverless functions, a traffic spike) does not need its own upgrade decision mid-incident. Direct Postgres connections from the internet are not open yet; migrations, the console SQL editor and the HTTP serverless driver work today.

## Sources

- [Scribase pricing](https://scribase.com/pricing)

## FAQ

**Why does Neon's connection limit change?**

max_connections on Neon scales with the compute size (CU) currently assigned to the project — from 112 at 0.25 CU up to a capped 4,000 at 9 CU and above. A compute that autoscales down, or is pinned small, has a lower ceiling.

**What is the fastest fix for "too many connections" on Neon?**

Switch to the pooled (-pooler) connection string. Neon's PgBouncer accepts up to 10,000 client connections and multiplexes them onto a much smaller number of real database connections (0.9 x max_connections per user/database pair), which absorbs far more concurrent app instances than the direct limit.

**Does upgrading compute size fix it permanently?**

It raises the ceiling (each step roughly doubles max_connections) but does not fix a connection leak. Neon's own guidance is to find and close idle or long-running connections first, then use pooling, then consider a larger compute.

## Related

- [Check a connection string](https://scribase.com/tools/connection-string)
- [Supabase too many connections](https://scribase.com/supabase-too-many-connections)
- [Prisma P1001: can't reach database server](https://scribase.com/prisma-p1001-cant-reach-database)
- [Neon compute limit and cold starts](https://scribase.com/neon-compute-limit-cold-start)
- [Neon alternative](https://scribase.com/neon-alternative)
- [Migrate from Neon](https://scribase.com/migrate/neon)