> ## Documentation Index
> Fetch the complete documentation index at: https://help.nops.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Move Pro SQL Warehouse to Serverless

> Move classic or Pro Databricks SQL warehouses to serverless, which needs no EC2 of your own.

Classic and Pro SQL warehouses bill Databricks DBUs and also launch EC2 instances in your AWS account. A serverless SQL warehouse of the same size bills DBUs only, at a higher rate per DBU, with the compute included. nOps adds up what a Pro or classic warehouse costs in DBUs and EC2, and compares it with the same DBUs at the serverless rate you already pay.

| | |
| - | - |
| **Cloud** | Databricks |
| **Category** | Compute |
| **Service** | Databricks SQL |
| **Action** | Modify |
| **Effort** | Low |
| **Restart needed** | Yes |
| **Can be rolled back** | Yes |
| **Analysis window** | Last 7 days |
| **Source** | nOps |

## Why it matters

The EC2 instances behind a Pro warehouse are easy to miss because they appear on the AWS bill, not the Databricks bill. Serverless also starts in seconds and stops quickly when idle, which can save more.

## How nOps finds it

nOps reads the last 7 complete days of classic and Pro SQL warehouse DBUs and the on-demand EC2 instances tagged with that warehouse. The serverless price is your own billed serverless SQL rate for the same tier and region over the last 30 days (at least 10 DBUs). The saving is the warehouse DBU cost plus its on-demand EC2 cost, minus the same DBUs at the serverless rate, scaled to a month, and must be at least \$25 a month. Warehouses whose EC2 is mostly covered by Reserved Instances or Savings Plans, warehouses with no matched EC2, and workspaces with 90% or more of Databricks spend covered by credits are left out. The extra saving from serverless stopping sooner when idle is not counted.

## Before you start

Serverless SQL enabled for the workspace, and no network or instance-profile setup that serverless cannot use.

## How to implement

1. Confirm serverless SQL is enabled for the workspace and supports the warehouse's network setup (for example private connectivity to data sources).
2. In Databricks, open **SQL Warehouses**, edit the warehouse and change its type to **Serverless**, or create a new serverless warehouse of the same size.
3. Set auto stop to a few minutes.
4. Point dashboards, queries and BI tools at the warehouse (the ID stays the same if you edit it in place).
5. Check that the EC2 instances tagged with the warehouse stop appearing in your AWS bill.

## Risks and considerations

* Changing the warehouse type restarts it, so running queries are interrupted.
* Serverless warehouses run in the Databricks account, so instance profiles and some network setups need serverless equivalents.
* To roll back, change the type back to Pro or classic.

## FAQs

> [accordion] Why can serverless be cheaper when its DBU rate is higher?
>
> The serverless rate includes the compute. With Pro or classic you pay the DBU rate plus the EC2 instances in your own AWS account.

## Related

* [Stop Always-On SQL Warehouse](https://help.nops.io/inform/supported-recommendations/stop-always-on-sql-warehouse)
* [Supported recommendations](https://help.nops.io/inform/supported-recommendations.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.