Data teams have spent the last decade pairing two categories of tool: something to move data into a warehouse, and something to transform it once it lands. Those two categories just merged into one company, and the terms of that merger are worth understanding before renewing either tool.
What just happened
On June 1, 2026, Fivetran and dbt Labs completed an all-stock merger first announced on October 13, 2025 (Fivetran, 2026). George Fraser, Fivetran's co-founder, is CEO of the combined company. Tristan Handy, dbt Labs' co-founder, is president. Financial terms were not disclosed. The companies said the combined business is approaching $600 million in annual recurring revenue and serves more than 100,000 data teams, including named customers OpenAI, Zendesk, Coupa and HubSpot (Fivetran, 2026).
Those figures come from the companies' own announcement and have not been independently audited. What is verifiable is the structural fact: the company that moves data into a warehouse and the company that transforms it once it is there are now the same company.
What each product actually does
Fivetran is a managed data movement tool. It connects to source systems, databases, SaaS applications and event streams, and replicates that data into a warehouse or lake on a schedule, without customers writing or maintaining custom extraction code.
dbt, short for data build tool, is a transformation framework. Once raw data lands in the warehouse, dbt lets analytics engineers write the SQL that turns it into modeled, tested, documented tables that BI tools, and increasingly AI agents, can query with some confidence in what they are looking at.
Those are the E, L and T of ELT: extract and load handled by Fivetran, transform handled by dbt. Before this deal, most teams paired the two themselves, connecting an ingestion vendor to a transformation vendor through a shared warehouse. The merger now packages that pairing as a single purchase, at least in principle.
Why now
Two forces are pushing this. The first is a broader consolidation pattern across the category in 2026, with Snowflake, Databricks and Microsoft each already pulling more of the data stack under their own roofs, and Fivetran and dbt Labs following the same logic from the ingestion side. The second is that both companies frame the deal specifically around AI agents: an agent querying a warehouse directly is only as reliable as the transformation and documentation layer underneath it, and the combined company argues that owning both ends of the pipeline lets it guarantee lineage and data quality end to end (dbt Labs, 2026).
This did not start with the dbt Labs deal. In September 2025, Fivetran had already acquired Tobiko Data, the company behind SQLMesh, an open source transformation framework that competes directly with dbt (Fivetran, 2025). That followed an earlier 2025 acquisition of reverse-ETL vendor Census. The dbt Labs merger is the third and largest step in the same build-out.
The licensing change that came with it
The same day the merger closed, dbt Labs shipped the first alpha of dbt Core v2.0, built on the Fusion engine, a full Rust rewrite of dbt's execution engine, and licensed the code under Apache 2.0 (dbt Labs Developer Blog, 2026).
That matters because of how Fusion arrived in the first place. When dbt Labs announced Fusion in May 2025, it shipped under the Elastic License 2.0, which is free to run internally but bars a third party from offering it as a hosted or managed service without a separate commercial agreement with dbt Labs. Companies that host or manage dbt on behalf of other companies, including Astronomer, Dagster and Paradime, had flagged that restriction as a problem for their business model. Relicensing the code as Apache 2.0 removes it.
What to check before choosing a side
The merger does not force anyone off either product today. Both companies say Fivetran and dbt keep running independently in the near term, with tighter integration and bundled pricing to follow later rather than immediately (Fivetran, 2026). For a team choosing or renewing tools in this layer, three questions are worth asking directly rather than assuming an answer.
First, whether dbt Core stays maintained at parity with the commercial dbt Cloud and Fusion products, or whether engineering investment quietly shifts toward the paid tiers over time. An open license does not guarantee equal investment going forward.
Second, whether the neutrality dbt built its reputation on, working the same regardless of warehouse or loader, survives being owned by a company that competes directly against Airbyte, Matillion, Stitch, Rivery and Estuary in ingestion. Fraser has estimated that 80 to 90 percent of Fivetran's customers already use dbt for transformation, which is exactly the overlap that makes bundling attractive to the vendor and makes independence harder for a customer to verify from the outside.
Third, what happens to SQLMesh. Fivetran now owns both dbt, through this merger, and SQLMesh, through the Tobiko Data acquisition nine months earlier. Teams already running SQLMesh should ask Fivetran directly how it plans to support two transformation engines that compete with each other under one roof.
What commonly goes wrong here
The mistake is treating "still open source" as equivalent to "still neutral." dbt Core's license is genuinely more permissive than it was a year ago, and that is real progress for anyone building a hosted product on top of it. But a license controls what can legally be done with the code. It does not control where a combined company points its engineering headcount, and that is the thing worth watching over the next several product releases, not the merger announcement itself.