The Fivetran and dbt Labs merger, first announced in late 2025, formally closed on June 1, 2026, the same morning Snowflake opened its Data Cloud Summit in San Francisco. The combined company will retain the Fivetran name as the corporate parent, with dbt continuing as a distinct product brand. Anjan Kundavaram, Chief Product Officer at Fivetran, laid out the joint product roadmap in a launch post the same morning. For most operators running a Snowflake, Databricks, or BigQuery warehouse with dbt Cloud on top, this is the largest structural change to their stack since Databricks acquired Tabular last year.
dbt Core v2.0 and the Fusion runtime go Apache 2.0
Alongside the close, the combined company shipped dbt Core v2.0, open-sourcing the Fusion runtime under Apache 2.0. The new Rust-based engine claims up to 10x faster parse times, a cleaner adapter contribution model, and modern docs. dbt platform ecosystem stats now sit at over 1 billion PyPI downloads and 100,000 weekly active projects. Three additional products launched the same day: dbt State, which reduces dbt-generated compute by 30 percent on average by skipping unchanged models; dbt Wizard, a personal dbt agent scoring 76 percent on ADE-bench tasks; and a Fivetran AI Connector Agent that generates managed connectors from API documentation in minutes.
The contract math just changed for every joint customer
The strategic logic is straightforward. Fivetran has owned the E and L of ELT for the better part of a decade, moving rows out of Salesforce, NetSuite, Postgres, and a long tail of SaaS sources into the warehouse. dbt has owned the T, providing the SQL-first transformation framework that almost every analytics engineering team we work with has standardized on. The two products have always sat next to each other in the pipeline, but the handoff between them has been brittle. Schema changes upstream in Fivetran would silently break dbt models downstream. Lineage stopped at the warehouse boundary. Cost attribution was a spreadsheet exercise. The merger pitch is that all of that becomes one graph, with one set of contracts, one observability surface, and one bill.
Both vendors have been aggressive on price increases over the last 24 months, and customers running both products at enterprise scale are routinely spending seven figures combined. The combined company has every incentive to bundle, which is good news on paper. In practice, we expect renewal repricing within the next 12 months that nets out higher for most accounts, particularly those who have been holding the line on dbt Cloud seat counts by running dbt Core in production. The leverage of we will switch to the other vendor is gone. Start modeling a 15 to 25 percent uplift at next renewal and negotiate hard against multi-year commits.
The open-source hedge every analytics team should formalize
dbt Core is the foundation that the entire analytics engineering profession is built on. Fivetran has historically been a closed-source company with a transactional view of community. The combined entity has published a statement reaffirming commitment to dbt Core and the Fusion engine, and shipping the runtime as Apache 2.0 on day one is a credible down payment. The proof will be in commit velocity over the next two quarters. We are telling clients to pin their dbt Core version, audit their reliance on Cloud-only features, and quietly evaluate SQLMesh as a hedge. The cost of switching transformation frameworks is real, but it is bounded. The cost of being captive to a vendor that decides to sunset the open-source path is much higher.
Native warehouse ELT just got a single target to compete against
Snowflake, Databricks, and BigQuery all have first-party ingestion and transformation now. Snowflake Openflow, Databricks LakeFlow, and BigQuery Data Transfer Service plus Dataform cover most of what Fivetran and dbt do, with the substantial advantage of being inside the same billing and security perimeter as the warehouse itself. The merger gives the new Fivetran-dbt one larger story to tell against those native stacks, but it also gives the warehouse vendors a single target to compete against. Expect Snowflake and Databricks to ship aggressive Fivetran and dbt migration tooling in the second half of 2026. If you are starting greenfield, the calculus for going native has shifted in favor of the warehouse vendor.
The trusted AI agents framing, separated from the marketing
Both companies have been positioning their products as the substrate for agentic workloads, where LLM-driven systems read and write data, trigger transformations, and act on results. The argument is that agents need verifiable lineage, contracts, and freshness signals to be trustworthy, and that a unified pipeline and transformation layer is the only way to deliver those guarantees end to end. There is real substance here. The new Agents Schema standard, available via Git, lets users designate a warehouse schema as the context layer agents read from, combining metric definitions, semantic models, and dbt lineage in plain SQL tables. But the marketing is running ahead of the engineering. We have not seen a production agentic workload in any of our client environments that actually consumes dbt semantic layer metadata at decision time. That will come, but treating it as a 2026 deliverable is optimistic.
For data leaders evaluating their 2027 stack, the recommended actions are: freeze any in-flight Fivetran or dbt expansion commitments until pricing clarity emerges in Q3, formalize a dbt Core fallback plan that includes Fusion runtime version pinning, and put a native warehouse ingestion proof of concept on the roadmap for H2. If the combined company ships a unified billing surface before the end of 2026 and dbt Core v2.0 sees consistent external commit velocity, the merger thesis works and the price uplift is defensible. If commits slow and Fusion drifts behind the platform product, evaluate SQLMesh seriously by Q1 2027.
