At this week's dbt Summit in Las Vegas, 2,000 data professionals gathered-with thousands more online-as Fivetran and dbt Labs laid out a shared thesis: the data foundation that makes analytics trustworthy is the same one that makes AI trustworthy.
The centerpiece is dbt v2, now generally available. It unifies dbt on one Rust-based engine, replacing the split between the original Python-based standard and last year's Fusion engine. Two distributions ship: the superset "dbt," free with optional paid features like dbt State, and "dbt-oss," limited to Apache 2 licensed components. dbt Core lives on as dbt v1 under Apache 2.0. BigQuery, Databricks, DuckDB, Redshift and Snowflake adapters are GA; ClickHouse and Spark are in beta; Athena, Fabric and Postgres are coming soon.
dbt State, also GA, inspects warehouse metadata and model SQL to decide what to build, skip, clone or defer each run. dbt Labs reports an average 15-30%+ cut in warehouse compute. RxBenefits' Chris Shepherd said scheduled-job warehouse costs fell 59% on a Snowflake adaptive warehouse, with over 700k models reused instead of rebuilt. Virgin Media O2's Gordon Curzon cited 25% savings on job run time and BigQuery compute. Freshness now lives in the model as a lag_tolerance rather than the job. State runs on dbt v1.7 through v2 across local, platform and self-managed orchestration.
Lake Compute enters private beta: a single-node DuckDB SQL engine for dbt that transforms Apache Iceberg tables directly, letting teams tag individual models to run there while the rest stay on the warehouse.
On the AI side, dbt announced out-of-the-box integrations with Anthropic and a ChatGPT plugin (GA) for surfacing models, metrics, lineage and tests through the dbt MCP Server. Fivetran Context Layer, in early access, turns structured and unstructured sources into context for tools like Claude and Slack. dbt Wizard, an agent grounded in your project, is in public preview in the dbt platform, with Explore Mode in public preview, a public beta CLI and a private beta Desktop app. dbt Charts, in public beta, brings dashboards into YAML under version control alongside the models they depend on.
Source: dbt Labs Blog