At Snowflake Summit this year, the combined Fivetran and dbt Labs team made several announcements that signal a shift toward a more performant, cost-efficient, and AI-ready data infrastructure. Here’s what changed and why it matters for practitioners.
**Merger, Not Just Acquisition** Fivetran and dbt Labs are now one company, unifying the ingestion and transformation layers that already ran together in most modern data stacks. The stated goal: build an open, portable foundation for trusted data agents. For teams running both tools, this means aligned roadmaps and a single engineering investment behind connector maintenance, SQL parsing, and metadata management. DocuSign’s Sr. Director of Global Data & Analytics noted that reusable, trusted data assets are central to scaling AI—exactly the kind of endorsement that drives enterprise adoption.
**dbt Core v2.0: Fusion Goes Open Source** The original dbt Core, built in Python, had scaling issues: parse times grew with project size, and errors surfaced only at runtime. The answer was a ground-up rewrite in Rust—the Fusion engine—which delivers up to 10x faster parsing and native SQL comprehension. Previously, Fusion was proprietary, creating a confusing two-engine landscape. Now, the Fusion runtime is open-sourced under Apache 2.0 as dbt Core v2.0. This means every practitioner gets a faster, more scalable foundation, and commercial improvements will flow directly into the open-source distribution. The proprietary distribution adds SQL comprehension, column-level lineage, and instant feedback. Alpha is available now for Snowflake, BigQuery, Databricks, and Redshift.
**dbt State: Skip What Hasn’t Changed** Every pipeline rebuilds everything by default—expensive and wasteful. dbt State checks metadata and model SQL on each run, building only what’s changed. In production, this cuts warehouse compute by an average of 30%. CarGurus reported a 9% compute reduction, 35% fewer models built, and 15% lower Snowflake backfill costs. The operational win: teams stop rationing runs and schedule refreshes based on business needs, not cost anxiety. Available now in preview for dbt Core 1.7+ and the dbt platform.
**dbt Wizard: An Agent That Knows Your Project** Generic coding agents produce plausible code but break downstream dependencies. dbt Wizard is an analytics engineering agent grounded in your project’s lineage, contracts, tests, and metrics. It validates its own work and shows changes with context. Early adopters saw a 15–20% reduction in production incidents. Available in the dbt platform and via CLI for local development.
**What This Means** For data engineers, the takeaway is clear: faster runtimes, lower costs, and AI tooling that respects your governance. The two-engine confusion is ending, state-aware orchestration is becoming standard, and the agent is now project-aware. If you’re building pipelines, it’s worth evaluating these changes now.
Source: dbt Labs Blog