dbt Core and Cloud Adopt

Overview

dbt (data build tool) models analytics transformations as software: modular SQL, packages, tests, and documentation generated from project metadata. dbt Core runs locally and in CI; dbt Cloud, now called the dbt platform, adds orchestration, job observability, and the MetricFlow-powered semantic layer for consistent business metrics (dbt documentation). Since September 2026 the engine is dbt v2 (the former Fusion engine): the full distribution is called "dbt", its Apache 2.0 subset "dbt OSS", and the dbt Core name remains only for the v1.x line (dbt licensing FAQ).

Adopt for any team building curated tables in a cloud warehouse or lakehouse who needs reproducible pipelines, pull request review for data logic, and contracts between analytics engineering and downstream AI or BI consumers.

Adoption Signals

• Fivetran and dbt Labs completed their all-stock merger on 1 June 2026; the combined company reports more than 100,000 data teams, naming OpenAI, Zendesk, Coupa and HubSpot (dbt Labs).
• dbt v2, the Rust-based Fusion engine under the new name "dbt", became generally available on 16 September 2026; dbt Labs cites a 10,000-node project compiling in 17 seconds instead of 70 seconds on dbt 1.12 (dbt blog).
• The open-source line continues: the "dbt OSS" subset of v2 and dbt Core v1.x are Apache 2.0, and MetricFlow, the engine behind the Semantic Layer, was relicensed to Apache 2.0 in October 2025 (dbt licensing FAQ).
• 44% of respondents in Astronomer's State of Airflow 2026 survey run dbt together with Airflow (Astronomer).

Risks

• Untested incremental models can silently duplicate or drop rows in production.
• Overly permissive warehouse roles in CI service accounts widen blast radius.
• Metric definitions without owners recreate conflicting KPIs across departments.
• Package version drift between projects complicates centralized upgrades.
• Licence split in v2. Features outside dbt OSS ship under the proprietary dbt Product Licensing Agreement, so teams that need a fully open-source toolchain must check which v2 capabilities they depend on (dbt licensing FAQ).

Recommendation

Adopt dbt as the transformation layer for warehouse and lakehouse analytics, with mandatory tests on incremental models and documented owners per semantic metric. Use the dbt platform (formerly dbt Cloud) when you need centralized scheduling and observability; stay on dbt OSS or Core plus your orchestrator when cost or data residency requires it. Test existing projects against dbt v2 before switching, since its built-in SQL compiler catches errors before execution that dbt 1.x only surfaced in the warehouse.

Sources

• dbt documentation
• dbt platform features
• dbt Semantic Layer
• dbt best practices
• Fivetran + dbt Labs complete merger
• dbt v2.0 is GA
• dbt licensing FAQ
• State of Apache Airflow 2026