The dbt Fusion engine: ship dbt at the speed
of AI development
Real-time validation for every commit — human or agent. Pipelines that only run what's changed. The dbt-native context your AI tools can trust.
Built on Rust. Powered by deep SQL comprehension.
Fusion is a true SQL compiler — it deeply comprehends SQL syntax and semantics across data platforms. That's what powers real-time IDE feedback, state-aware orchestration, and a rich metadata layer that gives AI agents accurate project context to generate code you can actually trust. Built in Rust with parallel processing, Fusion parses a 10,000-model project up to 30x faster than dbt Core.
Build faster
Compile SQL locally — turn a minutes-long feedback loop into seconds
Optimize costs
Cut redundant model runs with state tracking across code changes, source freshness, and refresh intervals
Power AI-ready data
Generate rich metadata and column-level lineage agents need to work reliably
Inside the dbt Fusion engine: A full rewrite in Rust
A hyper-responsive coding environment, wherever you build.
Fusion's language server delivers real-time intelligence directly in your editor: catch errors before anything hits the warehouse, validate columns, see full lineage as you type. Whether the developer is a human or an AI agent, Fusion validates every change against the full project before it runs.
Cut your warehouse compute costs by 30%+
State-Aware Orchestration (Preview), which tracks code changes, source freshness, and custom refresh intervals across every job run. It rebuilds only what’s changed or stale, and reuses everything that hasn’t. Smarter orchestration means more efficient pipelines, faster job runs, and real cost savings.
Intelligent skips: Automatically skip model runs when neither upstream data nor code has changed
Tuned configurations: Run only what the business needs, on the schedule it actually requires
Efficient testing: Reuse prior results and consolidate queries so savings compound across every layer
Context that makes your AI agents actually work
Fusion generates a complete metadata layer at compile time. That means column-level lineage, type information, and dependencies are generated automatically, as you build. AI agents get accurate, dbt-native project context instead of guessing from static docs. The result: generated code that actually works with your schema.
Column-level lineage: Trace exactly how every field flows, transforms, and gets renamed — in your IDE, before anything runs
Impact analysis: Map exactly what a change affects before it runs, and catch breaking changes early
dbt MCP server: Structured, Fusion-powered context for AI coding assitans and agents
Fusion and state-aware orchestration have changed what our team spends time on. Declaring SLAs at the model level and letting [dbt] decide what actually needs to run has already delivered 25-30% model reuse and roughly 15% in Snowflake savings on our pilot project.
Alvin Chai Senior Analytics Engineer at Fanatics
Meet the teams leveling up with Fusion.
Here's what analytics engineers and data leads say after making the switch.
Read Obie Insurance's storyWere saving at least 30% on compute costs, just from reusing models with state-aware orchestration
Tyson Doberneck, Senior Data Engineer
Preventing human errors with live error detection saves us valuable time. We used to depend on dbt build to catch issues—now theyre flagged instantly in VS Code.
Sonja Strempel, Analytics Engineer, DPG Media
Read NBIM's storyWe’re seeing better feedback loops, faster parsing times, and improved linting. Our less technical stakeholders, like portfolio managers, are creating higher-quality projects, too.
Øyvind Barsnes Eraker, Senior Data Engineer
Upgrade to Fusion in minutes
The Fusion quickstart guide gets you running in minutes. The dbt Developer Agent can help migrate existing projects, handling conformance issues automatically.