European Alternatives to Databricks
Looking for a European alternative to Databricks? Databricks, Inc. is a San Francisco company and the party to its Master Cloud Services Agreement. European customers get Irish law and Dublin courts under that contract, but its privacy notice names Databricks, Inc. in San Francisco as the controller for EEA, UK and Swiss users and relies on the EU-U.S. Data Privacy Framework and standard contractual clauses for transfers.
Eleven European tools cover the data engineering, lakehouse and ML platform side, as building blocks rather than one-for-one replacements: lakehouse and warehouse engines, ETL and orchestration, and ML platforms, from companies in Germany, France, the Czech Republic, Sweden and Finland.
How we rank these tools — 4-step process
-
1
European ownership, verified
The company is headquartered and incorporated in the EU, EEA or Switzerland, and processes customer data in Europe. A US parent company disqualifies a tool from this page regardless of where its servers are.
-
2
Category fit and hands-on review
What the tool actually does, who it suits, and where it falls short — checked against the vendor’s own documentation, changelog and pricing page rather than its marketing copy.
-
3
Compliance and pricing check
GDPR posture, hosting location and the prices quoted on this page are verified against the vendor’s public pricing before publication, and re-checked when we revisit the category.
-
4
Position on this page
Placement on this page can be paid, and that can affect which tools appear here and the order they appear in. It never buys a good review: a tool that fails the checks above is not here at any price, and payment does not change the shortcomings we write about. A vendor can ask us to correct a factual error — not to remove a criticism.
Vendors can pay for visibility on this page. It never changes what an entry says about a product, including the criticism, and we earn nothing when you click through to a vendor. Paid placement can affect which tools appear here and the order they appear in. Editorial policy
11 European Alternatives to Databricks
Stackable Data Platform
Open-source Kubernetes data platform with Trino, Iceberg and Spark, plus a managed lakehouse from €1,499 a month
Keboola
Prague data platform for ETL, SQL and Python transformations with a free plan and published compute rate
Exasol
Nuremberg analytics database for fast SQL and lakehouse acceleration, with a free Exasol Personal edition
Scaleway Data & AI Platform
Paris cloud with managed ClickHouse warehouse, Apache Spark clusters, a data orchestrator and GPUs
Hopsworks
Stockholm AI lakehouse with feature store, model registry and serving, and a free plan with no card
Kestra
Open-source orchestration platform from a French SAS, with 2,100+ plugins and unlimited flows
ZenML
Munich open-source MLOps framework for Python pipelines on your own infrastructure, with a published SaaS price
Aiven
Finnish managed open-source data services (Kafka, ClickHouse, OpenSearch, Postgres) with free plans that have no time limit
CloverDX
Prague data integration platform for building pipelines, run on your servers or on AWS, Azure or Google Cloud
Valohai
Turku MLOps platform that orchestrates ML pipelines across clouds and on-prem, with versioning and lineage
OVHcloud Data Platform
Roubaix cloud with an Iceberg-based Data Platform listed as coming soon, plus managed Kafka and ClickHouse
Key takeaways
- Databricks, Inc. in San Francisco is the contracting party, with Irish law and Dublin courts for European customers, but its privacy notice names the San Francisco company as the controller and relies on the Data Privacy Framework.
- Stackable (Wedel, Germany) is the closest in shape to a lakehouse: Trino, Iceberg, Spark, Airflow, NiFi and Superset, free as a Community Edition or managed from €1,499 a month.
- Keboola (Prague) is the easiest to try for pipelines: a free plan with SQL and Python transformations, 120 free compute minutes in the first month and $0.14 a minute after that.
- Exasol (Nuremberg) is an analytics database with a free Exasol Personal edition, and Hopsworks (Stockholm), ZenML (Munich) and Valohai (Turku) cover the machine-learning side.
- Scaleway and OVHcloud assemble lakehouse parts on European clouds, but Scaleway's Spark page reads like early access and OVHcloud lists its Data Platform as coming soon.
- None of the eleven replaces Databricks on its own. Seven of the eleven can be tried free, and you will usually combine two or three.
Why people leave Databricks
Databricks is the platform most large data teams meet when they outgrow a single database.
It gives you one workspace for Spark and SQL pipelines, a serverless data warehouse, data engineering and orchestration, governance, and machine-learning and GenAI tooling, on AWS, Azure or Google Cloud. Its price is based on compute usage measured in DBUs, with storage, networking and instances billed by your cloud provider on top. There is a free trial, a pay-as-you-go option by credit card, and a free, limited Community Edition meant for learning Spark.
It is run by Databricks, Inc. of San Francisco. Its Master Cloud Services Agreement is a contract with that company, and the governing law and venue depend on where your principal place of business is: Delaware for the Americas, Ireland and Dublin for Europe, England and Wales for the UK.
So a European customer is not pushed into American courts. The privacy notice, however, names Databricks, Inc. in San Francisco as the controller for users in the EEA, the UK and Switzerland, and the transfers rest on its EU-U.S. Data Privacy Framework certification and standard contractual clauses.
This page covers the data engineering, lakehouse and ML platform side of Databricks. No European product does all of it. Eleven tools qualify, and each covers a layer: lakehouse and SQL engines, ETL and orchestration, or ML platforms. You will usually combine two or three of them.
- The contract is American even when the law is Irish A European customer gets Irish law and Dublin courts under the MCSA, which is better than a US venue. The counterparty is still Databricks, Inc. in San Francisco, and the privacy notice names that company as the controller and relies on the Data Privacy Framework and standard contractual clauses for transfers. If your buyers or regulators want a European contracting company and a European controller, the tools below offer that. Check the entity for your own plan, because a few of them have group companies in the US.
- You pay two bills, and neither one is a fixed price Databricks prices compute in DBUs, and says that storage, networking and related costs vary by service and cloud provider, and that prices can vary by region. If you buy through Azure, Microsoft bills it under your Azure subscription. That is hard to forecast from the page alone. Several European options are more explicit about at least part of the bill: Stackable lists a managed lakehouse from €1,499 a month, Keboola lists extra compute at $0.14 a minute, and ZenML lists a SaaS plan at $999 a month.
- One platform is also one dependency Databricks builds on open-source engines, so leaving it is easier than leaving some closed rivals. But the features that exist only on its platform, such as its governance layer, its SQL warehouse and its AI assistants, have no drop-in equivalent elsewhere. A European stack built from open parts (Trino, Iceberg, Spark, Airflow, Kafka) gives you the same freedom to switch later, at the cost of operating more pieces yourself or paying someone to.
- The European answers split the job by layer Stackable, Scaleway and OVHcloud assemble lakehouse parts (Iceberg, Trino, Spark, ClickHouse). Exasol is a fast analytics database. Keboola, Kestra, CloverDX and Aiven handle ingestion, orchestration and streaming. Hopsworks, ZenML and Valohai cover machine learning. Choose by the layer you actually use on Databricks, not by the brand.
What you have to replace, not just match
List what you run on Databricks today: batch pipelines, streaming, SQL dashboards, notebooks, model training, feature tables, scheduled jobs and who has access to what. Then map each to a layer. Pipelines and jobs point to Keboola, Kestra or CloverDX. Spark and SQL on a lakehouse point to Stackable, Scaleway or OVHcloud. SQL speed on large data points to Exasol. Model work points to Hopsworks, ZenML or Valohai. Do not look for one product to replace all of it.
Keep your data in object storage you control and in an open table format such as Iceberg, so the engine on top can change without copying everything. Run the new stack beside Databricks on one real workload for a few weeks, compare run time and the total cost including cloud bills, and only then move the rest. Check what you rely on that has no equivalent, such as its governance layer or a specific AI feature, before you commit.
The alternatives compared
| Position | Tool | Headquarters | Pricing | Jurisdiction |
|---|---|---|---|---|
| #2 | Keboola | Prague, Czech Republic | Free plan with 120 minutes of compute in the first month and 60 minutes a month after that; additional compute at $0.14 per minute; subscription and Enterprise tiers quoted | EU (Czech Republic) |
| #4 | Scaleway Data & AI Platform | Paris, France | From about €0.0025/hour (DEV1-S) | EU (France) |
| #6 | Kestra | Lille, France | Free and open source / Enterprise on request | EU (France) |
| #8 | Aiven | Helsinki, Finland | Free tier for small services / from about $20/month, usage-based | EU (Finland) |
| #9 | CloverDX | Prague, Czech Republic | Unit-based licensing: one DX Unit per Designer seat or per Production Server core, sized through an online calculator across Standard, Plus and Enhanced support tiers; no published per-unit price, with multi-year discounts available | EU (Czech Republic) |
| #11 | OVHcloud Data Platform | Roubaix, France | PostgreSQL essential from about €0.16/hour (2 vCPU / 8GB RAM); MySQL, MongoDB, Kafka, Cassandra, OpenSearch, M3DB and Valkey priced separately by engine and tier | EU (France) |
How each alternative compares to Databricks
Stackable Data Platform
the closest in shape to a lakehouse, from a German company
Best for: Teams with Kubernetes skills that want an open-source lakehouse or a managed one
Stackable is run by Stackable GmbH in Wedel, Germany (Amtsgericht Pinneberg, HRB 15351 PI). Its platform is a set of open-source data apps on Kubernetes: Trino for SQL, Apache Iceberg tables on S3 storage, Spark for batch and ML, Airflow for orchestration, NiFi for ingestion and Superset for dashboards. The operators are under OSL-3.0 and the bundled apps under Apache 2.0.
You can run it yourself for free as the Community Edition, buy a Basic, Business or Business+ subscription for support with 9x5 or 24x7 SLAs, or take the Managed Data Lakehouse, which Stackable operates for you and lists from €1,499 a month. The site also advertises a 30-day free trial of the managed lakehouse.
Its terms page returned an error when we checked, so we cannot say which law governs the contract. It does not offer Databricks' serverless warehouse or governance layer as one product.
What Stackable Data Platform does better than Databricks
- German GmbH with a public register entry and an open-source core
- Free Community Edition and a published starting price for the managed lakehouse
- Iceberg, Trino and Spark on open standards, which keeps switching possible
- You can run it on infrastructure you choose
Where Stackable Data Platform is a step down from Databricks
- No single platform with warehouse, governance and ML tooling like Databricks
- Self-managed editions need real Kubernetes skills
- Governing law of the contract could not be verified
- The managed lakehouse covers only a subset of the modules (Trino, Iceberg, Superset, NiFi, Spark, Airflow)
Standout against Databricks. It is the only option here that packages a managed lakehouse from open parts at a published monthly price.
Keboola
pipelines and transformations in one platform with a real free plan
- Which law reaches it. EU (Czech Republic). Databricks is run from the United States, so the CLOUD Act obliges the provider to hand over data on a valid order regardless of which country the servers are in.
- Where the data sits. Multi-cloud on AWS, Azure and Google Cloud; free-plan projects run on Microsoft Azure in the EU, and enterprise contracts can pin the provider, the region or a private cloud.
- Source code. Closed source, as Databricks is.
- Independently checked. SOC 2 Type II, GDPR and HIPAA, listed against the Enterprise tier.
Best for: Data teams that want ETL, SQL and Python transformations and orchestration without assembling tools
Keboola's free-plan terms are a contract with Keboola Czech s.r.o. in Prague, governed by Czech law with the Czech courts. Its privacy policy also names Keboola LLC in Chicago and Keboola Data Services, Inc. in British Columbia as controllers, depending on which entity signs your contract, so check yours.
The Free plan costs $0 and lists unlimited ETL/ELT pipelines, 1,500+ data sources, SQL and Python transformations, Flow Builder, an MCP server and one project. It starts with 120 free compute minutes in the first month and 60 a month after that, and extra minutes cost $0.14. It runs on an extra small Snowflake backend, which is a US vendor's engine. Enterprise is on request and can use other storage such as BigQuery or DuckDB.
Keboola is a data platform for moving and shaping data, not a Spark cluster or a notebook environment for machine learning.
What Keboola does better than Databricks
- Free plan with real compute and a published top-up rate
- Czech company with Czech law on the free-plan terms
- Ingestion, transformation and orchestration in one place
- Data take-out button to export your data and code
Where Keboola is a step down from Databricks
- Not a Spark or ML platform, so heavy data science needs another tool
- The free plan sits on a Snowflake backend
- Group companies in the US and Canada may be the contracting party
- Pricing for the Enterprise plan is on request
Standout against Databricks. It has a free plan that includes SQL and Python transformations and a published price for extra compute.
Exasol
a fast analytics database with a free Personal edition
Best for: Teams whose bottleneck is SQL speed on large data, rather than the whole platform
Exasol is run by Exasol AG in Nuremberg (Amtsgericht Nuremberg, HRB 23037). Its published terms for customers in Europe except the UK are under German and Swiss law, unless your quote says otherwise, while customers in the UK, the US and elsewhere get a different master agreement under US and UK law.
Its homepage offers Exasol Personal for free, Exasol Enterprise for production, and acceleration products for data warehouses and lakehouses. It describes itself as the analytics database, aimed at faster queries and lower costs, with options for AI and ML workloads. No prices are shown.
It is an engine for SQL analytics. It does not replace Databricks' orchestration or ML tooling, so it fits as the query layer in a stack with something like Keboola or Kestra.
What Exasol does better than Databricks
- German AG with a register entry and German or Swiss law for European customers
- Free Exasol Personal edition to test the engine
- Focused on SQL analytics speed and cost
- Offers lakehouse acceleration without replatforming, by its own description
Where Exasol is a step down from Databricks
- Not a full platform, with no orchestration or ML tooling
- No public prices
- Exasol Personal's licence terms were not read
- The applicable law depends on your quote
Standout against Databricks. It focuses on one thing, SQL analytics speed, and lets you test it for free.
- Which law reaches it. EU (France). Databricks is run from the United States, so the CLOUD Act obliges the provider to hand over data on a valid order regardless of which country the servers are in.
- Where the data sits. Paris, Amsterdam, Warsaw — EU only.
- Source code. Closed source, as Databricks is.
Best for: Teams that want data and AI services on a French cloud
Scaleway is run by Scaleway SAS, a simplified joint-stock company registered in Paris (SIREN 433 115 904, 8 rue de la Ville-l'Évêque). Its legal notice and the French register confirm the entity. We did not read the governing law of its customer contracts.
Its Data & AI Platform page puts together a Data Warehouse for ClickHouse, Clusters for Apache Spark, a Data Orchestrator, managed Kafka and AI products. It says the platform is developed and operated on European soil and protects against foreign jurisdiction, which is the vendor's own claim. The Spark page uses the heading "Be the first to use", so check that it is generally available before planning around it.
No prices were read for this page. It is a set of cloud services you assemble, not one workspace like Databricks.
What Scaleway Data & AI Platform does better than Databricks
- French SAS with a public register entry
- ClickHouse warehouse, Spark and an orchestrator on one cloud
- Positions itself on European operation and open standards
- GPUs and AI services on the same cloud
Where Scaleway Data & AI Platform is a step down from Databricks
- Services are separate pieces, not one integrated workspace
- The Spark product may be early access
- Prices and contract law were not verified
- No governance layer comparable to Databricks
Standout against Databricks. A French cloud that offers a ClickHouse warehouse, Spark clusters and an orchestrator under one roof.
Hopsworks
a feature store and ML platform from Stockholm, with a free plan
Best for: ML teams that need real-time features, a model registry and serving
Hopsworks is provided by Hopsworks AB, Asogatan 119, Stockholm. Its terms are governed by Swedish law with Swedish courts, except for European consumers. We did not check the Swedish company register.
Its homepage calls it the AI lakehouse for real-time ML, with a feature store, model registry and model serving. It works with Delta, Iceberg and Hudi tables and with Spark, Flink, Pandas and DuckDB, and it offers on-premises and air-gapped deployment. The pricing page lists a Free plan with one project and no card, a pay-as-you-go SaaS plan with no figure, and a custom Enterprise plan.
It covers the ML and feature side of Databricks, not general ETL or a SQL warehouse.
What Hopsworks does better than Databricks
- Swedish company under Swedish law
- Free plan with no credit card
- Air-gapped and on-prem deployment options
- Works with Delta, Iceberg and Hudi tables
Where Hopsworks is a step down from Databricks
- Not a general data engineering or SQL platform
- Pay-as-you-go and Enterprise prices are not shown
- The free plan covers one project
- No SQL warehouse or general ETL
Standout against Databricks. A feature store built for real-time ML that you can try without a card.
- Which law reaches it. EU (France). Databricks is run from the United States, so the CLOUD Act obliges the provider to hand over data on a valid order regardless of which country the servers are in.
- Where the data sits. Your own infrastructure, or EU cloud.
- Source code. Open source, where Databricks is not: you can read what it does rather than take the description on trust.
Best for: Teams that need to schedule and orchestrate pipelines across many systems
Kestra is run by Kestra Technologies, 81 rue du Pré Catelan, La Madeleine, France, a simplified joint-stock company (SIREN 900427873) created in June 2021, according to the French register. Its privacy policy names it as the controller. Its terms page returned an error, so we could not read the governing law.
Its pricing page shows an Open Source edition and an Enterprise edition, with 2,100+ plugins, unlimited flows and executions, YAML and no-code editors, an AI copilot and an MCP server. Enterprise is on request and adds security and support, and the FAQ describes a managed Kestra Cloud.
Kestra orchestrates work that runs elsewhere. It is not a lakehouse, and you still need an engine such as Spark or Trino to do the processing.
What Kestra does better than Databricks
- French SAS with a public register entry
- Open-source edition with unlimited flows and executions
- 2,100+ plugins that connect to other systems
- Can run on PostgreSQL or MySQL with nothing else to operate
Where Kestra is a step down from Databricks
- Does not process data itself, so it needs an engine
- Contract terms and law could not be verified
- Enterprise price is on request
- No warehouse or ML tooling
Standout against Databricks. An open-source orchestrator with 2,100+ plugins and a managed cloud edition.
ZenML
open-source MLOps for Python pipelines on your own infrastructure
Best for: ML teams that want to keep their pipelines portable across orchestrators and clouds
ZenML is run by ZenML GmbH, Schellingstr. 36, Munich (Commercial Register Munich HRB 268487). Its terms of service are governed by German law with the courts of Munich. The company states SOC 2 Type II and ISO 27001 compliance.
ZenML is open source under Apache 2.0, and its pricing page lists Open Source as free with unlimited executions and projects, Scale SaaS at $999 a month for 2,000 executions and 3 projects, and Enterprise on request with SSO, audit logs and air-gapped deployment. A separate product, Kitaru, is $39 a month.
It orchestrates ML pipelines on the infrastructure you already run. It does not provide the compute or the warehouse.
What ZenML does better than Databricks
- German GmbH with a register entry, German law and a Munich court
- Open source with a free self-hosted version
- A published price for the paid SaaS plan
- Runs on the orchestrator and cloud you already use
Where ZenML is a step down from Databricks
- Only covers pipelines and the model control plane, not the data platform
- The paid plan is priced in US dollars and starts high at $999
- Needs engineers to wire it to compute and storage
- No SQL warehouse or data platform
Standout against Databricks. It keeps ML pipelines portable and does not tie them to one cloud.
- Which law reaches it. EU (Finland). Databricks is run from the United States, so the CLOUD Act obliges the provider to hand over data on a valid order regardless of which country the servers are in.
- Where the data sits. Your chosen region, including EU-only.
- Source code. Closed source, as Databricks is.
- Independently checked. Services are open source.
Best for: Teams that need managed Kafka, ClickHouse, OpenSearch or PostgreSQL with less operations work
Aiven's privacy policy names Aiven Ltd, with Finnish business ID 2795743-5, as the controller. The Finnish register lists that ID as Aiven Oy, registered in November 2016. Its terms are governed by Finnish law, with disputes settled by arbitration under the rules of the Finland Chamber of Commerce. The terms text we read does not name the entity.
Its products include Apache Kafka, PostgreSQL, OpenSearch, ClickHouse, Valkey and MySQL. The pricing FAQ says free plans exist for PostgreSQL, MySQL, OpenSearch, Kafka and Valkey, with capped resources and no time limit. No figures were read for the paid plans.
Aiven runs the data services around a platform like Databricks (streaming, a ClickHouse analytics engine, search), not the platform itself.
What Aiven does better than Databricks
- Finnish company with Finnish law
- Free plans for five services that do not expire
- Managed Kafka and ClickHouse for streaming and analytics
- Console, CLI, API and Terraform support
Where Aiven is a step down from Databricks
- No ETL builder or ML platform
- Arbitration in Finland rather than a local court
- Paid prices were not read
- The terms we read do not name the contracting entity
Standout against Databricks. Permanent free plans for five managed open-source data services.
- Which law reaches it. EU (Czech Republic). Databricks is run from the United States, so the CLOUD Act obliges the provider to hand over data on a valid order regardless of which country the servers are in.
- Where the data sits. Self-hosted only — CloverDX runs entirely inside the customer's own AWS, Azure or Google Cloud account, or fully on-premise; CloverDX itself never stores or transmits customer data.
- Source code. Closed source, as Databricks is.
Best for: Teams that build data pipelines and want to run them on-prem, on Docker or on a cloud of their choice
CloverDX's privacy policy names CloverDX a.s. in Prague (IČO 28213874, Municipal Court in Prague, B 13411). Its EULA, however, names a different company for UK and EU customers: Javlin Ltd., an English company in Tunbridge Wells, under English law and the courts of England and Wales. Outside the UK and EU, the contract is with CloverDX Inc. in Arlington, Virginia.
It is a data integration platform with an AI assistant, transformations, orchestration, master data management and a data catalog. It runs on your own servers, on Docker, or on AWS, Google Cloud or Azure. Its pricing page charges per DX Unit (one designer seat or one production server core) and shows an estimator that ends in "Let's talk", so there is no price.
It builds pipelines and does not provide a Spark lakehouse or a machine-learning platform.
What CloverDX does better than Databricks
- Czech company with a public register entry
- Runs on-prem, on Docker or on any of three clouds
- Capacity-based units instead of per-row pricing
- Includes a data catalog and master data management
Where CloverDX is a step down from Databricks
- Contracting entity is British for UK and EU customers
- No public prices and no free plan on the pages we read
- Not a lakehouse or ML platform
- Commercial licence rather than open source
Standout against Databricks. It prices by capacity units rather than by row or connector.
Valohai
an MLOps platform that runs your pipelines on any cloud or on-prem
Best for: ML teams that train on several clouds or their own GPUs and need lineage
Valohai is provided by Valohai Oy, Linnankatu 16, Turku, Finland, registered as a limited company in the Finnish register (business ID 2786205-7). Its terms are governed by Finnish law.
Its site describes multi-cloud orchestration, on-prem and hybrid setups, dataset caching, automatic versioning, full lineage and one-click reproduction. It says it has been doing this since 2016. The pricing page says a subscription is bought through sales on a per-user licence basis, and no figure is shown. The site offers a Start free button and a demo.
It manages ML work and does not provide a warehouse or ETL like a Databricks workspace.
What Valohai does better than Databricks
- Finnish company with a register entry and Finnish law
- Runs on several clouds and on-prem
- Strong on lineage and reproducibility
- Per-user licence instead of per-compute billing
Where Valohai is a step down from Databricks
- Focused on ML pipelines only
- No public price
- Subscriptions are sold through sales
- The free offer's limits were not verified
Standout against Databricks. It keeps lineage and reproducibility across clouds, from a company that has done ML platforms since 2016.
OVHcloud Data Platform
a European cloud building an Iceberg-based data platform, not yet released
- Which law reaches it. EU (France). Databricks is run from the United States, so the CLOUD Act obliges the provider to hand over data on a valid order regardless of which country the servers are in.
- Where the data sits. 30+ data centres, majority in Europe.
- Source code. Closed source, as Databricks is.
- Independently checked. SecNumCloud (ANSSI), GAIA-X founding member.
Best for: Teams on OVHcloud who want to follow an open-standards lakehouse as it arrives
OVHcloud is run by OVH SAS in Roubaix, France (SIREN 424 761 419, active in the French register). We did not read the governing law of its customer contracts.
Its Public Cloud page for the Data Platform is labelled "Coming soon". It describes a Lakehouse Manager on Apache Iceberg, connectors, a data processing engine for ETL/ELT, and open-source standards such as Iceberg, Trino and Spark, with no egress fees and ISO 27001, ISO 27701, HDS and SOC 2 Type 2 certified infrastructure. Today it already sells managed Kafka, ClickHouse and OpenSearch, and AI Notebooks, AI Training and AI Deploy.
Because the Data Platform is not released, treat it as a roadmap item and build on the services that exist today.
What OVHcloud Data Platform does better than Databricks
- French company with a public register entry
- Managed ClickHouse, Kafka and AI notebooks available today
- Plans to build on Iceberg, Trino and Spark rather than a closed format
- States no egress fees for the Data Platform
Where OVHcloud Data Platform is a step down from Databricks
- The Data Platform is listed as coming soon
- No prices read for this page
- Contract law was not verified
- Nothing comparable to Databricks' integrated workspace yet
Standout against Databricks. A European cloud that has announced an Iceberg-based lakehouse, with the pieces for it already on sale.
What actually breaks when you switch
The hard part of leaving Databricks is the glue, not the data. Notebooks, scheduled jobs, permissions and any features that exist only on its platform have to be rebuilt or mapped on the new stack, and it takes longer if your tables are not in an open format in storage you control. Budget for a parallel run on one real workload before you commit.
Check the contract entity and the region for the plan you would actually buy. Keboola has group companies in the US and Canada, CloverDX's EU contract is with a UK company, and Aiven's terms text does not name the entity. Scaleway's Spark clusters and OVHcloud's Data Platform may not be generally available yet, so confirm in writing.
Compare total cost, not the sticker. Databricks bills compute in DBUs and your cloud bills storage and networking separately, and the European tools price in different ways: per month, per compute minute, per execution or on request. Ask each vendor for a quote based on your own workload.
Is there one European product that works exactly like Databricks?
No. The closest in shape is Stackable's Managed Data Lakehouse, which bundles Trino, Apache Iceberg, Superset, NiFi, Spark and Airflow into a lakehouse that Stackable operates for you. Scaleway and OVHcloud are assembling similar sets of parts on their clouds, and Keboola covers pipelines and transformations in one platform.
None of them has Databricks' one platform for SQL, governance and ML. If you rely on that integration, expect to run two or three tools instead of one.
Which of these can I try for free?
Stackable's Community Edition is free and open source (you run it on your own Kubernetes), and it advertises a 30-day free trial of the managed lakehouse. Exasol Personal is free, Keboola has a $0 plan, Hopsworks has a free plan with one project and no card, and Kestra and ZenML are open source. Aiven has free plans for PostgreSQL, MySQL, OpenSearch, Kafka and Valkey with no time limit.
CloverDX, Scaleway and OVHcloud do not show a free plan on the pages we read, and Valohai has a Start free button but publishes no limits or price. Databricks has its own free trial and a limited free Community Edition, so the free entry point is not unique to the European tools.
Where does the European data actually sit?
It depends on the tool and on where you deploy it. Stackable's platform runs on any Kubernetes, and Hopsworks lists AWS, Google Cloud, Azure and OVHcloud for its Kubernetes installer, so the cloud under it is your choice. Scaleway and OVHcloud run their own services on their own clouds. Keboola's free plan uses an extra small Snowflake backend, which is a US vendor's engine, and we could not confirm the region.
Ask each vendor for the region and the cloud provider of your plan in writing, and keep in mind that a European contracting company running on a US hyperscaler still depends on that hyperscaler.
Does the contract really matter for a data platform?
For a data platform the contract covers your most sensitive records, so it is worth reading. With Databricks, a European customer gets Irish law and Dublin courts, which is a real European term, while the controller and transfer mechanisms point to the US company and the Data Privacy Framework. With the tools here, the contracting entity sits in the EU, in the UK, or in Finland, France, Germany, Sweden or Czechia, and each vendor states its own law.
For a small test it is a minor point. For regulated or customer data it belongs next to data location, subprocessors and exit terms. Databricks' subprocessor page returned an error when we checked, so we could not compare it.
Which one to pick
If you want the nearest thing to a lakehouse, with open parts and a European GmbH behind it, Stackable is the one to look at first, either as the free Community Edition or as the managed lakehouse from €1,499 a month. If your work is mostly pipelines and transformations, start with Keboola's free plan, and if your bottleneck is SQL speed, test Exasol Personal.
For machine learning, Hopsworks, ZenML and Valohai each cover a different part: real-time features, portable Python pipelines and multi-cloud lineage. Kestra, CloverDX and Aiven fill in orchestration, integration and streaming, and Scaleway and OVHcloud are the European clouds to watch if you want the data stack on a European provider.
Be realistic about the gap. None of the eleven gives you Databricks' one platform for SQL, governance and ML, and combining two or three means more integration work. Switch if you need a European contracting party and controller, or if you want open parts you can move later, and keep Databricks if its integrated platform is the reason your team is productive.
Frequently Asked Questions
There is no single one. Stackable is the closest to a lakehouse, with Trino, Iceberg and Spark under a German company, and it has a free Community Edition. Keboola is the better choice for ETL and transformations with a free plan, Exasol for fast SQL analytics, and Hopsworks, ZenML or Valohai for machine learning. The answer depends on which layer of Databricks you actually use.
No. Databricks, Inc. is based at 160 Spear Street, San Francisco, and its Master Cloud Services Agreement is with that company. For customers in Europe the agreement is governed by Irish law with Dublin as the venue, and for customers in the UK by the law of England and Wales with London as the venue.
It is priced by compute usage, measured in Databricks Units (DBUs), and the rates vary by product, cloud provider and region, with a price list on its website. Storage, networking and cloud instances are billed separately by your cloud provider. There is a free trial, and a free, limited Community Edition for learning Spark. If you use Azure Databricks, Microsoft bills you under your Azure subscription.
Stackable lists its Managed Data Lakehouse from €1,499 a month, and a free, self-managed Community Edition with community support. It also sells Basic, Business and Business+ subscriptions with 9x5 or 24x7 support, and does not show a price for those. Keboola's Free plan is $0 and includes 120 compute minutes in the first month and 60 a month after that, with extra minutes at $0.14 each. The Enterprise plan is on request.
Hopsworks has a free plan with one project, a feature store and model registry, and no card. Its pay-as-you-go SaaS plan and Enterprise plan show no figure on the page. ZenML's Open Source version is free, and its Scale SaaS plan is listed at $999 a month for 2,000 executions and 3 projects, with Enterprise on request. ZenML also sells a separate product called Kitaru at $39 a month, which is not a ZenML plan.
Stackable includes Apache Spark in its platform and its managed lakehouse. Scaleway has a page for managed Apache Spark clusters with JupyterLab, though the page reads like an early-access product, so check availability. Hopsworks lists Spark among the frameworks it works with, and OVHcloud names Spark in its Data Platform description, which is listed as coming soon.
Exasol is an analytics database from Exasol AG in Nuremberg, with a free Exasol Personal edition and Exasol Enterprise for production. It also lists a lakehouse acceleration product. It is a fast SQL engine, not an ETL tool or notebook environment, so you pair it with something like Keboola or Kestra for the pipelines.
Stackable's operators are under the OSL-3.0 licence and its bundled data apps under Apache 2.0. ZenML and its sister product Kitaru are listed as Apache 2.0, and Kestra has an open-source edition next to Enterprise. Aiven builds its managed services on open-source engines. Exasol, Keboola, CloverDX, Valohai and OVHcloud are commercial products.
Not as a switch you flip. Code written for Spark and SQL can move to a Spark or SQL engine elsewhere, but notebooks, jobs, permissions and any Databricks-specific features need to be rebuilt or mapped by hand. Moving the tables is easier if they are stored in an open format in your own object storage. Run the new stack beside Databricks on one real workload before you move the rest.
Its privacy policy names CloverDX a.s. in Prague, registered with the Municipal Court in Prague. The EULA, however, defines the contracting party differently by region. For customers in the UK and the EU it is Javlin Ltd., an English company in Tunbridge Wells, under the law of England and Wales. Outside the UK and the EU it is CloverDX Inc. in Arlington, Virginia, under Virginia law.
Related comparisons
Who worked on this review
Explore More European Alternatives
Discover privacy-focused European alternatives to other popular US tech services.