If your team mainly writes SQL and wants less to operate, look at Snowflake or, on Google Cloud, BigQuery. If you are a Microsoft organisation using Power BI, Microsoft Fabric offers Spark, a warehouse and BI under one capacity. If you want Spark without a separate platform fee on top of AWS, Amazon EMR and AWS Glue run open source Spark as AWS services. If you want full control, Apache Spark with Apache Iceberg is the self-managed open source route. For SQL over a lake you already have, Starburst (Trino) and Dremio; for real-time analytics, ClickHouse; and for data that fits on one machine, DuckDB.
Databricks is a lakehouse platform on AWS, Azure and Google Cloud. Data lives as Delta Lake or Apache Iceberg tables in cloud object storage, governed by Unity Catalog, and is processed by Apache Spark notebooks and jobs, pipelines, Databricks SQL warehouses and ML tooling. Databricks bills in Databricks Units (DBUs), pay as you go per second, with discounts for committed use. Databricks also bought Tabular, the company founded by Iceberg's original creators, in 2024, and agreed to acquire the serverless Postgres company Neon in May 2025.
People who move some or all of their work elsewhere usually have one of these reasons. The first is documented by Databricks; the others are our editorial reading, not survey data.
- Two bills on classic compute. Databricks states that when it runs in your own cloud account you are also charged by your cloud provider for compute instances, storage and networking. Reconciling DBUs with cloud invoices is a common cost-control complaint. Serverless compute avoids provisioning in your account but is a different billing line to track.
- More platform than the team needs. Clusters, jobs, notebooks, pipelines and catalogs suit data engineering teams; a team of analysts who only write SQL may prefer a simpler warehouse (editorial).
- Spark for small data. Many workloads fit on one machine, where a distributed engine adds cost and start-up time without benefit (editorial).
- Wanting to own the stack. Delta Lake, Spark and Iceberg are open source, so some teams prefer to run them directly, or on a cloud provider's managed Spark service, rather than through a platform subscription (editorial).
For head-to-head pages, see Snowflake vs Databricks, Databricks vs BigQuery, Databricks vs Redshift, Databricks vs Microsoft Fabric and ClickHouse vs Databricks.
Quick picks
A managed warehouse on AWS, Azure and Google Cloud with per-second warehouse billing and Iceberg table support.
Spark, a T-SQL warehouse, pipelines and Power BI on one capacity, with OneLake storing Delta or Iceberg tables.
Open source Spark, Iceberg and other frameworks as AWS services, billed per second.
Apache 2.0 licensed engine and table format, run wherever you choose.
Trino-based engine that queries Iceberg, Delta, Hive and Hudi tables and other databases in place.
How we chose
We considered Snowflake, Google BigQuery, Microsoft Fabric, Amazon EMR, AWS Glue, Amazon Redshift, self-managed Apache Spark with Apache Iceberg, Starburst, Dremio, ClickHouse and DuckDB. To be included, a product had to be sold or actively developed in October 2026, cover at least part of what teams use Databricks for (data engineering with Spark, SQL analytics on a lake, or warehousing), and publish its pricing model and deployment options on an official page. Amazon Redshift is covered in our Snowflake alternatives and BigQuery alternatives guides and in Databricks vs Redshift, so it is not repeated here.
For each product we recorded the billing unit, the clouds it runs on, whether it uses open table formats (Iceberg, Delta Lake), its free tier or trial, and its main limitation. Exact prices are not quoted because they vary by cloud, region and configuration. None of this comes from our own use of the platforms.
The order is not a ranking. Entries start with managed platforms that replace most of Databricks, then managed and self-managed Spark, then SQL engines over a lake, then specialised and single-machine engines.
At a glance
| Platform | Pricing model | Clouds | Open formats | Best for | Main trade-off |
|---|---|---|---|---|---|
| Snowflake | Credits by edition, cloud and region, plus storage | AWS, Azure, Google Cloud | Iceberg tables (Snowflake or external catalog) | SQL-first analytics teams | Less flexible for Spark-style engineering and ML |
| Google BigQuery | Per TiB scanned or slot capacity editions | Google Cloud; Omni for some AWS and Azure regions | Iceberg managed tables in your Cloud Storage | Serverless SQL on Google Cloud | Tied to Google Cloud |
| Microsoft Fabric | Capacity units (F SKUs), OneLake storage separate | Microsoft-hosted SaaS on Azure | Delta Parquet and Iceberg in OneLake | Power BI and Microsoft 365 shops | Shared capacity; Power BI licences below F64 |
| Amazon EMR and AWS Glue | EMR fee plus EC2, or serverless; Glue DPU-hours | AWS only | Iceberg (and other lake formats) | Spark jobs on AWS without a platform layer | You assemble notebooks, catalog and governance |
| Apache Spark with Apache Iceberg | Free software; you pay for infrastructure | Any cloud or on premises | Iceberg (open spec) | Teams that want to own the stack | All operations, tuning and upgrades are yours |
| Starburst | Credits (Galaxy) or licence (Enterprise) | Galaxy: AWS, Azure, Google Cloud; Enterprise: anywhere | Iceberg, Delta Lake, Hive, Hudi | Federated SQL over an existing lake | SQL only; no notebooks or ML platform |
| Dremio | Dremio Compute Units (Cloud); Enterprise by quote | Cloud: AWS; Enterprise: AWS, Azure, Google Cloud, on premises | Iceberg, Apache Polaris catalog | Iceberg lakehouse SQL and BI acceleration | Dremio Cloud is AWS only for now |
| ClickHouse | Free open source; Cloud billed in compute units plus storage | Self-hosted; Cloud on AWS, Google Cloud, Azure | Reads and writes Iceberg (with limits) | Real-time analytics on event data | Not a Spark or ML platform |
| DuckDB | Free (MIT); MotherDuck usage-based | Any machine; MotherDuck on AWS | Parquet; Iceberg (read, and write via REST catalog) | Single-machine analytics and local development | Single node, not a shared platform |
Snowflake
Snowflake is a managed cloud data warehouse. Compute runs in virtual warehouses that consume credits, billed per second after a 60-second minimum each time a warehouse starts, and storage is billed on average compressed volume. Credit prices depend on the edition (Standard, Enterprise, Business Critical, Virtual Private Snowflake), the cloud and the region. There is no separate cloud provider invoice for the warehouse, which is the main contrast with Databricks classic compute.
Snowflake supports Apache Iceberg tables with Snowflake or an external catalog (such as AWS Glue or Snowflake Open Catalog). The trial lasts 30 days or until the free usage balance is used. See Snowflake vs Databricks.
- Snowflake documents that third-party engines cannot write to Iceberg tables when Snowflake is the catalog.
- Spark-style data engineering and ML need Snowpark or other tools, which differ from Databricks notebooks and jobs.
- Credit spend depends on warehouse sizing and running time, so it still needs active cost management.
Google BigQuery
BigQuery is Google Cloud's serverless warehouse, with no clusters to size. On-demand queries are billed per TiB scanned, with 1 TiB of queries and 10 GiB of storage free each month as listed on Google's pricing page on 7 October 2026; capacity pricing uses slots through the Standard, Enterprise and Enterprise Plus editions with autoscaling. BigQuery ML (Enterprise and above) runs some machine learning in SQL.
For lakehouse users, BigQuery offers Apache Iceberg managed tables stored in your own Cloud Storage bucket, which Google documents as readable by engines such as Spark. That lets a team keep Spark jobs on the same tables while moving SQL analytics to BigQuery. See Databricks vs BigQuery.
- Mainly a Google Cloud service, so it suits Databricks on GCP users more than those on AWS or Azure.
- On-demand cost follows bytes scanned, which needs partitioning and clustering discipline.
- Notebook-based Spark engineering is handled by other Google Cloud services, not BigQuery itself.
Microsoft Fabric
Microsoft Fabric is the closest like-for-like alternative in shape: it combines Spark notebooks and jobs, Data Factory pipelines, a T-SQL warehouse, real-time analytics and Power BI. Compute is a single capacity (F SKUs from F2 to F8192 capacity units) billed per second with a one-minute minimum on pay-as-you-go or discounted through reservations; capacities can be paused, and OneLake storage is billed separately.
OneLake stores tables in Delta Parquet or Iceberg format, and Microsoft documents that the T-SQL and Spark engines work on the same copy. Shortcuts can point at ADLS, Amazon S3 and other storage, and Microsoft documents an Azure Databricks integration with OneLake, so the two can coexist during a move. See Databricks vs Microsoft Fabric.
- All workloads draw on one capacity, so a heavy Spark job can throttle reports until you scale or pause.
- On F SKUs smaller than F64, Power BI viewers need Pro or Premium Per User licences.
- Runs only as a Microsoft service, so it is not an option for teams that need AWS or Google Cloud hosting.
Amazon EMR and AWS Glue
Amazon EMR runs open source frameworks including Apache Spark, Hive and Presto on AWS. EMR on EC2 adds a per-second EMR charge to the EC2 and EBS cost with a one-minute minimum; EMR Serverless bills for the vCPU, memory and storage a job uses; and EMR on EKS runs on Kubernetes. AWS documents Iceberg support on EMR, and the latest 7.x release (emr-7.14.0) ships Iceberg 1.10.1.
AWS Glue is the serverless option for Spark-based ETL, billed in DPU-hours per second with a one-minute minimum per run; its Data Catalog stores table metadata (the first million objects are free) and Glue jobs read and write Iceberg tables. Together they replace the data engineering part of Databricks for AWS users, with Athena or Redshift for SQL.
- AWS only.
- There is no single workspace: notebooks, scheduling, catalog, governance and SQL come from separate AWS services that you assemble.
- Glue documents that UPDATE is not supported for Iceberg tables in Glue ETL, so some table maintenance needs other tools.
Apache Spark with Apache Iceberg
Databricks is built around Apache Spark, and the open source engine remains available under the Apache 2.0 licence; the current release is Spark 4.2.0 (14 July 2026). Paired with Apache Iceberg as the table format (current release 1.12.0) and an Iceberg REST catalog such as Apache Polaris, it gives a lakehouse whose tables other Iceberg-capable engines on this page (Trino, DuckDB, ClickHouse, Snowflake and others) can also read.
This route removes the platform fee but not the work. In our view it suits teams with platform engineers who already run Kubernetes or VMs at scale and want no single vendor between them and their data.
- You run clusters, upgrades, security, table compaction and catalog yourself.
- No built-in notebooks, job scheduler, governance UI or SQL warehouse; each needs another project or product.
- Total cost includes engineering time, which is easy to underestimate.
Starburst
Starburst is built on Trino, the Apache 2.0 licensed distributed SQL engine governed by the Trino Software Foundation. Starburst Galaxy is fully managed and billed in credits, with a free tier of up to three clusters, paid Pro, Enterprise and Mission-Critical tiers and a 30-day trial; Starburst Enterprise is self-managed.
Galaxy documents support for Iceberg, Delta Lake, Hive and Hudi tables, so it can query the Delta tables a Databricks estate already holds while engineering jobs move elsewhere, and join them with relational databases. See Trino vs Presto.
- SQL only: there are no notebooks, Spark jobs or ML tooling.
- Storage, table maintenance and catalog remain your responsibility.
- Credit prices vary by tier, cloud and region.
Dremio
Dremio is a lakehouse SQL platform built around Apache Iceberg and the Apache Polaris open catalog. Dremio Cloud is fully managed and billed in Dremio Compute Units, with a 30-day trial; Dremio Enterprise is self-managed on Kubernetes, on premises or in the cloud; and a free Community Edition runs on your own machine or server.
It suits teams whose main Databricks use is SQL and BI over lake tables and who want to standardise on Iceberg rather than Delta Lake.
- Dremio Cloud runs on AWS only; Dremio lists Azure as coming soon.
- Not a Spark or ML platform, so engineering and data science need other tools.
- Enterprise pricing is by quote.
ClickHouse
ClickHouse is an Apache 2.0 licensed column-oriented database for fast aggregations over large append-heavy tables. ClickHouse Cloud offers Basic, Scale and Enterprise tiers, metered per minute in compute units with storage billed separately, plus a bring-your-own-cloud option. ClickHouse can read and write Apache Iceberg tables, with documented limits.
It is an alternative for one specific Databricks use: serving dashboards or product analytics that need low-latency queries on fresh data. Batch engineering and ML usually stay on another platform (editorial). See ClickHouse vs Databricks.
- Not a general data engineering or ML platform.
- Its SQL dialect and table design (table engines, sort keys) differ from Spark SQL.
- Iceberg writes have documented type and DML restrictions.
DuckDB
DuckDB is an MIT-licensed analytical database that runs inside a Python, R or other process with no cluster. It reads Parquet directly, and its Iceberg extension reads Iceberg tables and writes to tables managed by an Iceberg REST catalog. The current stable release is 1.5.6, with DuckDB 2.0.0 scheduled for 21 October 2026, so check the release notes before relying on version-specific behaviour.
For jobs that Databricks runs on small clusters, DuckDB on a single VM or laptop may be enough. MotherDuck, a managed service built on DuckDB on AWS, adds sharing and has a free Lite plan. See DuckDB vs pandas for the dataframe angle.
- Single node and in-process: data must fit the resources of one machine.
- Not a shared, governed platform for many users without MotherDuck or other tooling.
- A major release (2.0) is imminent, so extensions may need updating.
When to stay with Databricks
Stay with Databricks if you use most of it: Spark engineering, pipelines, SQL warehouses and ML on the same governed tables, across AWS, Azure or Google Cloud. Few alternatives cover all of that in one product; Fabric comes closest but only as a Microsoft service. If the problem is cost, check whether serverless compute, cluster policies and auto-termination, or committed-use pricing change the picture before migrating.
Because Databricks stores data as Delta Lake or Iceberg tables in your own object storage, partial moves are realistic: keep engineering on Databricks and point Starburst, Snowflake, BigQuery or DuckDB at the same tables for SQL, or the reverse. A full move means rewriting notebooks, jobs and Unity Catalog permissions, so run both in parallel first.
Frequently asked questions
What is the main alternative to Databricks?
Snowflake is the most direct alternative for SQL analytics, and Microsoft Fabric for a combined Spark, warehouse and BI platform. See Snowflake vs Databricks and Databricks vs Microsoft Fabric.
Is there an open source alternative to Databricks?
Yes, in parts. Apache Spark (Apache 2.0), Delta Lake and Apache Iceberg are open source, and Trino (Apache 2.0) or DuckDB (MIT) can provide SQL. You assemble and operate the platform yourself, including notebooks, scheduling and governance.
Why is Databricks billed twice?
On classic compute in your own cloud account, Databricks charges DBUs and your cloud provider charges for the virtual machines, storage and networking, as Databricks states on its pricing page. Serverless compute runs without provisioning resources in your account.
Is Databricks Free Edition allowed for work?
No. Databricks states that Free Edition is for personal use only and its terms do not allow commercial use. It replaced Community Edition. Use a trial to evaluate it for work.
Can Snowflake or BigQuery read my Databricks tables?
Often yes, through Iceberg. Databricks documents an Iceberg REST Catalog API in Unity Catalog and UniForm, which lets Iceberg clients read Delta tables. Check each engine's supported catalogs and whether it can write as well as read.
Is Databricks better than Redshift?
They target different starting points: Databricks is a multi-cloud lakehouse with Spark, Redshift is an AWS warehouse. See Databricks vs Redshift and, for the concepts, data warehouse vs database.
Sources
- Databricks pricing
- Databricks serverless compute
- Databricks: Iceberg in Unity Catalog
- Databricks: UniForm (Iceberg reads of Delta tables)
- Databricks Free Edition
- Databricks and Tabular announcement
- Databricks agrees to acquire Neon
- Snowflake pricing options
- Snowflake: Apache Iceberg tables
- Snowflake trial accounts
- BigQuery pricing
- BigQuery Iceberg managed tables
- Microsoft Fabric pricing
- Microsoft Fabric licences and capacity
- OneLake overview
- Amazon EMR pricing
- Amazon EMR: Iceberg
- AWS Glue pricing
- AWS Glue: using the Iceberg framework
- Apache Spark downloads
- Apache Iceberg releases
- Starburst pricing
- Starburst Galaxy documentation
- Trino Software Foundation
- Dremio pricing
- ClickHouse Cloud pricing
- ClickHouse GitHub repository (licence)
- DuckDB Iceberg extension
Checked October 2026.
How we research these guides: our editorial method.