Everything we announced at dbt Summit and why it matters

A decade ago, dbt gave people a name for something they were already trying to do: write SQL like software engineers, version it, test it, document it, trust it. That idea became a new practice: analytics engineering. The people who built their careers on it became some of the most capable data professionals in the industry.

Now, those same people are asking a harder question: what impact will AI have on the practice of analytics engineering? AI needs an engine that's fast enough to keep up, context that's‌ trustworthy, and agents that understand your business instead of guessing at it. The question isn't whether analytics engineers still matter; it’s how do they level up for this new era?

This week at dbt Summit, we welcomed 2,000 data professionals to Las Vegas (and thousands more online) to discuss the future of analytics in the AI age. Fivetran and dbt Labs are coming together around one thesis: the data foundation that makes analytics trustworthy is the same foundation that makes AI trustworthy. To help our users level up on both fronts, we announced a series of new features across the dbt and Fivetran product portfolios.

Level up the engine

One dbt, one engine, built to move as fast as you do

dbt has always evolved with growing data workloads. Last year, we introduced the dbt Fusion engine, a full rewrite of dbt in Rust with native SQL comprehension and dramatically faster performance than the original Python-based standard. But maintaining two engines created real friction, both for us and for the thousands of teams trying to figure out which one to build on.

dbt v2 ends that. Now GA, dbt v2 is one modern, Rust-based engine powering all of dbt, whether you work locally or in the dbt platform. It's the same workflow, on a faster foundation—thanks to all the innovation in Fusion over the past 18 months.

We publish two distributions of v2:

• The superset dbt includes all parts of the framework, super-charged with SQL comprehension features. It’s free to use, with additional optional paid features (for example, dbt State).
• The subset distribution dbt-oss includes only the components that have an Apache 2 license

dbt Core isn't going anywhere either. It’s still Apache 2.0, still open source, simply renamed as the previous version, dbt v1.

Adapter availability:

• BigQuery, Databricks, DuckDB, Redshift, and Snowflake are GA
• ClickHouse and Spark are in beta and available to install and test locally
• Athena, Fabric, and Postgres are coming soon

Get started on v2 today.

Build what's changed, skip what hasn't with dbt State, now GA

Now officially GA, dbt State checks your warehouse metadata and model SQL for what's changed, then builds, skips, clones, or defers each run accordingly. That intelligence results in an average 15-30%+ reduction in warehouse compute and removes the manual syntax and workarounds teams built to avoid running more than they needed to.

With dbt State, freshness moves from the job to the model. Every model carries its own freshness requirement in code, a lag_tolerance that says how stale it's allowed to be, simplifying orchestration. And because codified freshness rules and automatic reuse constrain what any run can cost, the guardrails now sit in the infrastructure instead of with whoever, or whichever agent, issues the command. The result is felt daily in development. dbt State removes complex dev setup rituals and risks of accidental builds for faster, safer dev work.

dbt State runs on dbt v1.7 through v2, wherever you run dbt: locally, in the dbt platform, with your own orchestrator, and across Snowflake, BigQuery, Databricks, and Redshift.

Get started on dbt State today so you can optimize costs, save time, and level up wherever you run dbt.

Own your data. Align costs to value. Introducing Lake Compute, now in Beta

Most companies run every dbt model, from massive joins to simple staging tables, on the same warehouse compute. With the rise of Apache Iceberg, you can store your data in open-table formats, and have the flexibility to choose the right compute engine for each workload.

We now offer that choice with Lake Compute (Private Beta), a single-node SQL engine built on DuckDB for dbt that runs transformations directly against Apache Iceberg tables. Tag one model or a hundred to run on Lake Compute. The rest keep running on your warehouse, and refs keep working across both. Engine choice becomes a per-model decision instead of a replatforming program. With Lake Compute, we’re working towards the promise of an open data lakehouse: one where you can choose the right compute for the job.

Learn more about Lake Compute here. Want to see it in action? Join our upcoming webinar on cost-optimized transformations with dbt, Apache Iceberg, and multi-engine compute. We'll walk through a live, dual-engine dbt project running partial Snowflake, partial Lake Compute, on the same Iceberg tables, so you can see exactly what "engine choice as a per-model decision" looks like in practice. Save your seat here.

Level up for AI: context, engineered

AI agents don't necessarily need more of your data. They need an understanding of what the data means, where it came from, whether it's fresh, and who owns it. That foundation is already in place: dbt Semantic Layer governs metrics, Agents Schema centralizes context, and dbt MCP Server exposes your models, metrics, lineage, and test results to any AI agent.

Delivering structured dbt context to your favorite AI tools just got easier with new out-of-the-box integrations with Anthropic and a plugin in ChatGPT (GA). No more multiple MCP servers to manage. Just one click, and your team can securely access structured context from your dbt project instantly.

Structured context gets you closer to reliable AI answers. But most of what a business runs on doesn't live in structured tables at all. It lives in call recordings, support tickets, Slack threads, and it changes constantly. That's what Fivetran Context Layer is built for: turning every source your business runs on—both structured sources like dbt and unstructured ones that BI tools never touched—into context for the AI tools you're already using, from Claude to Slack. Because Fivetran and dbt have visibility into your entire data estate, this context gets built and maintained as data moves and transforms, not stitched together after the fact.

Sign up for the early access program. Want to go deeper on this? Join our session, From Analytics Engineer to Context Engineer: A dbt Playbook, where we make the case that context engineering is analytics engineering with a new last mile. We'll walk through the exact patterns dbt Labs uses internally to turn unstructured sources into governed, versioned context, and introduce the dbt_context_engineering package that puts those patterns into practice. Save your seat here.

Level up with AI

AI is raising the bar on what it takes to ship trustworthy data: more models, more context, less room to guess. Most of that time gets lost relearning what should already be known: agents rediscovering a project's lineage and tests before they can start, visualizations living outside the codebase entirely.

dbt Wizard is an agent built specifically for analytics engineering. It’s grounded natively in your project, so it already knows which tools to call and which context to pull without any setup. It validates proactively: checking upstream and downstream impact, compiling and building the change before anyone sees the diff. And because data work is visual, you can review all of that against the full DAG. Wizard in the dbt platform is now in Public Preview.

dbt Wizard Explore Mode (Public Preview) brings conversational analytics right where the data work already happens. It lets business users and data teams ask questions in plain language and get answers grounded in the same dbt project your data team already maintains. When an answer falls short, you're already one step from the model that needs fixing.

Wizard CLI (Public Beta) puts the same project-grounded agent in the terminal you're already running dbt from. No new app, no new tab, and no platform account required. And Wizard Desktop (Private Beta) picks up where the terminal runs out of room: a dedicated local workspace for longer, more complex work. You can run several tasks side by side instead of juggling windows with a visual preview of the code, data, and lineage before anything ships.

Learn more here.

Closing the last gap: dashboards as code

Asking questions is only one way people need to work with data. Sometimes what they need is a dashboard—one they'll come back to every day. And that's where the story gets uncomfortable: everywhere else in the stack, teams have brought in real engineering discipline—version control, code review, CI.

dbt was the reason SQL went from copy-pasted queries to models you can trust. But the dashboard at the end of that pipeline—the thing an executive looks at—was still locked in, trapped inside whatever proprietary tool you bought, stored in someone else's format, disconnected from the code that produces the numbers.

dbt Charts (Public Beta) closes that gap. It brings discipline to dashboards: built as YAML, version-controlled right next to the models they depend on, living in the same repo, the same pull request, the same CI as the SQL underneath it. Because it's declarative, it becomes a shared language, one humans and agents can both read, write, and review, just as reliably as they do your SQL.

Check out dbt Charts today.

We’re all leveling up

These announcements represent more than a set of new products. Together, they represent the forward direction of Fivetran + dbt Labs. AI is changing the game, and we are building the data foundation to help you succeed in the agentic AI era.

Here’s a recap of how to get started with any of these new products:

• Upgrade to dbt v2 for faster development, SQL comprehension, and more.
• Turn on dbt State and start simplifying your development and saving on compute.
• Use Lake Compute to run the right engine for each model, without a replatforming project.
• Start developing agentically with dbt Wizard—it’s available in the dbt platform, in a brand new Desktop app, and via the CLI.
• Explore mode in Wizard makes it easy for business users to chat with your data. Add a read-only user at no cost, they'll learn about the data, and you'll learn a lot from what they ask.
• Get on the waitlist for Fivetran Context Layer.
• Try dbt Charts, the first language-first BI layer.