How to Run Your First Notebook in Databricks Free Edition

Short answer: Sign up at databricks.com for the free edition, create a serverless SQL warehouse, then attach a notebook to it and run a SELECT on the built-in samples.nyctaxi.trips table. Setup takes about 10 minutes and costs nothing — the free tier has no expiring trial, just lower compute limits.

Databricks split the old 14-day community trial into a permanent free edition in 2024. That change is why so many older tutorials now dead-end at a “your trial expired” screen.

What is Databricks?

Databricks is a cloud data platform built on Apache Spark, with a managed notebook workspace, Unity Catalog for governance, and SQL warehouses for BI queries. It is where a lot of enterprise data engineering actually happens. The free edition gives you one metastore, serverless compute, and a limited amount of monthly usage.

It is not a general-purpose app. If you only need a data frame in a script, pandas on a laptop is still simpler. Databricks earns its place when data is too big for one machine or when more than one person needs to share the pipeline.

How do you sign up for the free edition?

1. Go to the Databricks site and choose Get started free. Pick the free edition, not a paid cloud trial.
2. Authenticate with email or a Google/Microsoft account. No credit card is requested for the free edition.
3. Choose your region. Pick the one physically closest to you; latency on notebooks is noticeably better.
4. Wait for the workspace to provision. Usually under two minutes.
5. On first login, the workspace offers a Quickstart tour. Skip it if you have used Spark before — it spends time on UI chrome.

How do you create and run a notebook?

The default free workspace lands you in the SQL persona. For Python, switch to the Data Science & Engineering persona in the left rail.

1. Click New → Notebook, name it first_notebook.
2. Set the default language to Python.
3. On the compute selector at the top, choose Serverless rather than starting a cluster. In the free edition this avoids the per-cluster startup wait, which can run to three or four minutes.
4. Paste a first cell:

display(spark.table("samples.nyctaxi.trips").limit(10))

5. Press Shift+Enter. The first query compiles Spark, so expect 20–40 seconds. Later cells return in one to three.
6. If you need SQL instead, prefix the cell with %sql and the same table works.

The samples catalog ships with every workspace. It contains nyctaxi.trips, nyctaxi.fares and a few others — enough to build a real query without uploading anything.

Is the free edition actually unlimited?

No. Free edition has monthly ceilings on serverless compute and a cap on concurrent users, and warehouses auto-stop after a period of inactivity. You also cannot add collaborators from outside your account in some configurations, which matters if you are trialing this for a team.

There is also no SLA and no support ticket channel. For learning, fine. For anything with a deadline, not fine.

Does Databricks work with dbt and notebooks from other tools?

Yes. Databricks exposes JDBC/ODBC endpoints, a SQL connector for Python, and a native dbt adapter. Most real projects end up writing SQL in dbt and running it against a Databricks warehouse rather than hand-writing Spark.

One practical catch: the Python connector expects a personal access token and a server_hostname. Generate the token under Settings → Developer → Access tokens, and paste the hostname without https://. Half the failed connections I see come from that prefix.

Who is it for?

Reasonable for analysts who need to query warehouse-scale data without provisioning infrastructure, and for anyone learning Spark jobs in a shared workspace. Less reasonable for small projects — the free edition’s compute limits make it a poor place to run nightly batch jobs, and you cannot install arbitrary Python packages outside the cluster’s configured libraries in the free tier.

Two honest drawbacks beyond that. The UI is dense; the SQL editor hides warehouse settings two clicks deep, and the first hour is mostly navigation. And the free edition quietly drops your workspace after extended inactivity, so export notebooks you care about.

Quick Summary

• Cost: free edition is permanent, with monthly compute caps; no credit card.
• Setup time: about 10 minutes to a running notebook.
• Sample data: samples.nyctaxi.trips is available in every workspace.
• Best for: learning Spark and SQL, shared analytics workspaces.
• Main limits: no SLA, tight compute ceilings, idle workspace removal.

Last updated October 2026.