ClickHouse is now on the dbt platform — what that actually means for your pipeline

If you've been running dbt against Snowflake, BigQuery, or Postgres and wondering whether ClickHouse is a realistic target now, the honest answer is: getting there, but not yet for production.

What was actually announced

ClickHouse is now the first partner-built v2 adapter on the dbt platform, and the inaugural member of dbt Labs' Partner Adapter Program. The new adapter runs on dbt's v2 engine, codenamed Fusion — a Rust-based rewrite of dbt's core, built for speed, in place of the previous Python-based engine. It supports both self-managed ClickHouse and ClickHouse Cloud as targets, and (separately) the existing dbt-clickhouse connector already works with Fivetran-managed transformations.

What's ready today, and what isn't

This is the part worth being precise about, since "just announced" and "ready to point your production pipeline at" are very different things here. Per ClickHouse's own guidance:

• The v2 adapter is in beta — install and test it locally, on a self-managed single-node instance or ClickHouse Cloud, in dev or staging.
• It is explicitly not recommended for production workloads yet.
• Against dbt's own v1 integration test suite, it passes more than 81% of the full suite and 92% of the tests that actually apply to ClickHouse Cloud specifically (see the chart above).
• Fusion capabilities dbt is still building out for this adapter include dialect-aware validation, static analysis, the language server, column awareness in the VS Code extension, and engine-side column-level lineage — none of that is fully there yet.

If you already use the older, more established dbt-clickhouse adapter against dbt Core (not the new v2/Fusion path), that's a separate, more mature integration that predates this announcement and isn't affected by it.

Why it's worth paying attention to anyway

Two separate things are genuinely interesting here, independent of the beta status. First, ClickHouse itself is built for fast analytical queries at a cost profile that's hard to match with row-oriented warehouses for the right workload shape (high-cardinality aggregations, real-time analytics) — a first-class dbt integration lowers the barrier to actually trying that fit instead of hand-rolling SQL outside dbt's dependency graph and testing framework. Second, Fusion's Rust engine is dbt's own answer to slow local development loops — if the speed claims hold up as it matures, that's a real quality-of-life change for anyone's dbt workflow, not just ClickHouse users specifically.