Fivetran and dbt are one company now. Here's what that means.

We announced the merger in October. Today it's official: Fivetran and dbt Labs are one company, with a shared mission to build open data infrastructure for the agentic era.

The timing couldn't be better. In the months since October, the shift toward agentic AI has gone from a curiosity to a force of nature. What started as a strategic combination of two organizations to deliver best-in-class data pipelines has become something even more urgent: an absolute imperative to prepare the world for agents.

We've spent years building complementary halves of the same stack. Now we get to build them together…at exactly the moment it matters most.

The thesis

Agents are rapidly becoming the primary consumers of your enterprise data.

Agents are no longer just in production at frontier companies, but in enterprises globally. Agents approve insurance claims, deflect support tickets, and write code. And they do it at increasingly massive volumes: up at least 1-2 orders of magnitude YoY.

Many of these agents will need access to your organizational data.

Unfortunately, most data infrastructure deployed today was not built for agents, but for humans: analysts running queries, dashboards refreshing on a cadence, data consumers clicking around in a BI tool. Humans are patient. Humans bring their own context. Humans do one task at a time.

Agents don’t.

Agents operate at machine-speed. They’re running 24/7. They can issue queries at a volume human traffic never approached. And crucially: they don't have the institutional knowledge that a human analyst carries in their head. They can't fill in the gaps. When they encounter undefined metrics, ungoverned data, or fragmented business logic, they don’t shoulder-tap their colleagues...they often just (confidently!) produce the wrong answer. And the more autonomous they are, the harder that wrong answer is to identify and troubleshoot.

80% of IT leaders believe their enterprise data is not ready for agentic AI. 70% worry about AI governance with the proliferation of agents, according to Gartner. These aren't anxieties about the future. They're descriptions of the present.

The bottleneck has shifted. For the past decade, the bottleneck in data was infrastructure, and we, collectively, largely solved it. The modern data stack worked, although sometimes it could be a bit of a pain. The new bottleneck is trust: trust in the data, the context, the governance.

That is precisely the problem we built to solve. Together.

What Fivetran and dbt each bring, and why coming together matters

Reliable, governed, and trusted data for agents. Fivetran ensures data is complete, fresh, and reliably moved. When an agent reaches for your data, it's current and it reflects the actual state of your business across many sources. Agents querying stale, siloed data don't just produce wrong answers—they take wrong actions. Freshness matters more when the consumer is autonomous.

But fresh data alone isn't enough. Agents don't bring context; they inherit it. dbt's role is making sure that context is governed, defined, and trustworthy: business logic versioned in code, metrics defined once and tested, lineage traced from raw source to downstream consumer. This matters well beyond conversational analytics. The agent approving a loan, deflecting a support ticket, or triggering a supply chain reorder needs the same thing an analyst needs: reliable data. When an agent asks, "what’s the shipment status of this order?" or “is this customer in good standing?” it has to know exactly how to get that answer, without guessing.

Flexible and portable. Any engine, cloud, or model. No lock-in. Here's the thing about that business context: it needs to travel. The AI landscape is shifting fast: new models, compute platforms and tools are coming in and out of favor so fast that organizations can’t fully adopt one framework before the tide turns and it’s onto the next. The platform-native approach to data infrastructure—where your semantic definitions and pipelines are tightly coupled with a single compute vendor—breaks down the moment you need to either migrate or go multi-engine.

Fivetran moves data across any source without owning your storage. dbt keeps your business logic in code you control, portable across any engine or cloud. Deep integrations across the AI and analytics ecosystem means this infrastructure is pluggable at every layer, even as you migrate or go multi-engine. Together, you get a flexible, portable foundation that evolves with your architecture instead of constraining it. That's not a feature: it's a core design principle.

Scalable and optimized for production AI demands. Agents will issue queries at a volume human analysts never approached, and without an architecture built for it, costs grow incredibly fast. The data supports this: Despite a 95% drop in token costs, enterprise AI spend has exploded from $1.7B in 2023 to $37B in 2025. Cheaper tokens, yet bigger bills. Agents repeatedly querying fragmented systems for missing context is expensive in compute and tokens. Managed data connections reduce operational overhead. Governed context minimizes unnecessary retrieval hops. Decoupled storage and compute means you can route workloads to the most efficient engine for each job. The goal is for per-query cost to fall as volume scales, not rise with it.

Together, Fivetran and dbt deliver open data infrastructure for agents you trust, at scale:

This is just the beginning. Today, we announced two new exciting dbt features available to users wherever they write code.

Our commitment to open source, and what we're doing to reinforce it

Our commitment to open source isn’t changing. Open collaboration and community contributions have defined dbt's evolution from the start. It’s the very thing that’s made dbt the de facto standard for transformations on structured data, and that same ethos is what’s going to propel our users to succeed in this AI wave.

Today, we announced the alpha of dbt Core v2.0: the next major version of dbt Core. As it has been since the first commit, dbt Core remains licensed under Apache 2.0 with this release. What changes in v2.0 is what's under the hood: the kernel of the dbt Fusion engine, a full Rust rewrite of the dbt runtime, is being open-sourced under Apache 2.0. The Rust foundation powering Fusion—faster parsing, more consistent execution, dramatically better performance in agentic workflows—is now available to the entire community under the same license you’ve been using for a decade. Instead of two codebases, two licenses, and ambiguity about what’s available on dbt Core versus Fusion, where to build or how to migrate, we will simply have one engine: dbt. Not only are we excited to deliver all of these benefits to every dbt user, having a single engine to invest in allows us and the dbt community to innovate faster in the AI age where speed is everything.

This is the single biggest drop of new Apache 2.0-licensed code we've shipped in years…potentially ever.

What’s next

Personally, I’m excited about the future of the data ecosystem, of the role of the data practitioner, of what Fivetran and dbt Labs can build together. After several years of fascinating progress in consumer AI but a constant refrain of it-doesn’t-quite-work-yet in the enterprise, AI and agents are truly here. Which has caused the pace of change for everything inside of the data ecosystem to absolutely skyrocket.

This is great. This is what we should all want. Data practitioners—from engineers to analysts—are critical to the AI future, but have been sitting on the sidelines for the past few years saying “put me in, coach!”

Guess what: it’s time. While none of us know exactly how the next several years will play out, things will move quickly, and data practitioners everywhere will be central to the story.

What you can see us doing—with the merger, with our product launches, with our investments in open data infrastructure—is to answer the question “How do we position ourselves to elevate and advocate for data practitioners into the next decade?” The “how” is changing, but the mission remains constant. We want to be your partners in building data infrastructure for the age of AI and agents.

If you have questions about our shared vision, Core v2.0, and new products we announced at Snowflake Summit, join me on June 25th for a live Q&A.