Airbyte vs Databricks Lakeflow at a glance
| Airbyte | Databricks Lakeflow |
|---|
| Category | Data movement (ELT and CDC) platform | Ingestion, transformation and orchestration built into Databricks |
| Destinations | Any warehouse, lakehouse, database or vector store, on any cloud | Databricks (Delta tables governed by Unity Catalog) |
| Prebuilt connectors | 700+, including SaaS APIs, databases, files and unstructured sources | Managed connectors for major SaaS apps (such as Salesforce, Workday, ServiceNow, Google Analytics) and databases (MySQL, PostgreSQL, SQL Server, Oracle) |
| Custom sources | No-code Connector Builder with AI assist, low-code CDK, Python CDK | Standard connectors you code yourself in Spark, SQL or Lakeflow pipelines; community connectors |
| Customizing a managed connector | Open-source code you can fix or extend | Not supported; fully managed |
| Change data capture | Log-based CDC on every plan | Database connectors with a continuously running ingestion gateway |
| Pricing model | Volume-based (Standard, Plus) or capacity-based Data Workers (Pro, Enterprise Flex) | Compute-based, in DBUs: $0.35/DBU (Premium) or $0.45/DBU (Enterprise) for managed connectors on AWS |
| Free option | Airbyte Core (self-hosted) | 100 free DBUs per workspace per day for managed SaaS and database connectors |
| Deployment | Self-hosted (Core), managed cloud, hybrid or on-prem (Enterprise Flex) | Inside your Databricks workspace on AWS, Azure or GCP |
| Open source | Yes | No (Spark Declarative Pipelines is open source; managed connectors are not) |
| Who can use it | Analysts (UI) and engineers (API, Terraform, PyAirbyte) | Databricks users (UI, API, notebooks) |
About Airbyte
Airbyte is the open standard for data movement, used by 18% of the Fortune 500. It replicates data from 700+ sources into warehouses, lakehouses, databases and AI context stores, and runs self-hosted, fully managed, or hybrid inside your own boundary.
About Databricks Lakeflow
Lakeflow is Databricks' unified data engineering offering. It has three parts: Lakeflow Connect for ingestion, Lakeflow Spark Declarative Pipelines (formerly Delta Live Tables) for transformation, and Lakeflow Jobs for orchestration. Lakeflow Connect provides managed connectors for SaaS applications and databases that write to Delta tables governed by Unity Catalog, plus standard connectors such as Auto Loader that you configure in code.
Databricks-native vs. destination-neutral
Lakeflow Connect is built to land data in Databricks. If Databricks is your only analytics platform, that tight integration with Unity Catalog, serverless compute and Lakeflow pipelines is a real advantage.
Most data teams have more than one destination, though: a warehouse alongside the lakehouse, an operational database, or a vector store for AI. Airbyte writes to Databricks and to Snowflake, BigQuery, Redshift, Postgres, S3, Iceberg lakehouses and vector stores from the same connectors, so ingestion doesn't depend on which platform you choose next.
Connectors: managed catalog vs. 700+
Lakeflow Connect's managed connectors cover major SaaS applications and the most common databases. They're fully managed, which also means you can't customize how they ingest. For anything outside the managed catalog, you build your own pipeline with standard connectors, Spark or SQL.
Airbyte's catalog covers 700+ sources, including the long tail of SaaS tools, APIs and files. When a source is missing, the Connector Builder creates one from API docs with AI assist, and because connectors are open source, you can fix or extend any of them.
Pricing: DBUs, the free tier and the gateway
Lakeflow Connect is billed in DBUs. On AWS, managed connectors list at $0.35 per DBU on Premium and $0.45 on Enterprise. Since March 2026, every workspace gets 100 free DBUs per day for managed SaaS and database connectors. The free tier excludes Zerobus Ingest, Kafka-style streaming sources, community connectors and file connectors such as Auto Loader.
Database CDC adds a second cost. Each database connector runs an ingestion gateway continuously on classic compute, so you provision and pay for that compute even when the ingestion pipeline is idle. Gateways can't run in serverless-only workspaces.
Airbyte Core is free to self-host. Cloud starts at $20/month (Standard) and $189/month (Plus). Pro and Enterprise Flex price on Data Workers, dedicated compute capacity, so cost doesn't depend on DBU consumption or always-on gateways.
Deployment and lock-in
Lakeflow runs inside your Databricks workspace on AWS, Azure or GCP, and its managed connectors only exist inside Databricks.
Airbyte runs four ways on one codebase: free self-hosted Core, managed cloud, hybrid with an Airbyte-managed control plane, or fully on-prem with Enterprise Flex. Pipelines keep working if you add or switch destinations.
Using them together
Many Databricks teams use both. Lakeflow Connect handles the sources it covers natively, Airbyte ingests the long-tail SaaS and API sources into the same lakehouse, and Lakeflow Spark Declarative Pipelines or dbt transform everything downstream.
Where Databricks Lakeflow leads
Lakeflow is a strong fit for teams fully standardized on Databricks whose sources are in the managed catalog. Ingestion, transformation, orchestration and governance live in one platform, and the free tier makes small to mid-size ingestion effectively free.
When to choose Airbyte vs Databricks Lakeflow
| Your situation | Better fit |
|---|
| You need SaaS or API sources outside Lakeflow Connect's managed catalog | Airbyte |
| You load into more than one destination, or not only Databricks | Airbyte |
| You want to fix or extend a connector rather than wait for a managed one | Airbyte |
| You want database CDC without paying for an always-on classic-compute gateway | Airbyte |
| You need to self-host or run fully on-prem | Airbyte |
| Databricks is your only destination and your sources are in the managed catalog | Databricks Lakeflow |
| Your ingestion volume fits within 100 DBUs per workspace per day | Databricks Lakeflow (free tier) |
Build with Airbyte
Ship pipelines in minutes, not sprints. Try Airbyte free or talk to sales.
