Fivetran + dbt: The Myth of 'Open Data Infrastructure'

Fivetran and dbt Labs just announced an all-stock merger, calling it a milestone for the future of open data infrastructure.

Both companies say this isn’t about lock-in — it’s about “uniting to set the standard for open data infrastructure.” It’s a familiar story dressed in new packaging. Every consolidation starts with talk of openness, interoperability, and standards. But in practice, acquisitions make systems tighter, not freer.

Five years ago, the modern data stack — Fivetran, dbt, Snowflake, and Looker — was the gold standard. Every data team wanted it. Clean layers. Clear ownership. Each was imperfect in their own way but it fulfilled the promise of modular freedom.

Today, that same stack represents what most teams are trying to escape: complexity, compounding cost, and slow value realization.

The Promise (and Problem) of “Open Data Infrastructure”

dbt’s post frames the merger as the start of a new category: open data infrastructure. The idea sounds great — common metadata, portable models, no more lock-in.

But openness isn’t a press release. It’s a governance choice.

Every commercial “open” ecosystem eventually faces the same tension: the need to grow revenue faster than community goodwill. Once two closed-source SaaS products merge, the gravitational pull shifts inward. The center of gravity stops being the user — it becomes the platform.

If openness isn’t protected by structure, it turns into a slogan.

The Frankenstack Wasn’t Fixed — It Was Rebranded

The Frankenstack problem was never about the logos. It was about the glue.

You can put Fivetran, dbt, Snowflake, and Looker under one roof and still have four teams shipping on different cadences, four billing models, and four sets of permissions.

Consolidation doesn’t remove friction; it just moves it behind the login screen. The seams don’t go away — they just get more expensive to cross.

Where Governance Really Matters

This is where governance really matters.

DuckDB, for example, is governed by a non-profit foundation. No investors, no Cloud upsell strategy. That’s what real openness looks like.

dbt Labs, on the other hand, has venture investors and a fast-growing commercial Cloud tier. Those two forces — openness and growth pressure — don’t coexist for long.

Open-core is a stepping stone, not a destination.

Once the business depends on the commercial layer, the open layer becomes a funnel.

The Ecosystem Is Breaking Free

The irony is that all this consolidation is happening right as the rest of the ecosystem is breaking free.

Open-source tools like DuckDB, Iceberg, and dlt have quietly become world-class. Meanwhile, AI is making it possible for people who don’t know SQL to get the answers they need — without another BI dashboard in the way.

The world’s moving toward simpler, more accessible data platforms. And the old guard is trying to solve that by merging.

This Is Exactly Why We Built Definite

Instead of bolting AI onto a legacy stack, we built an AI-native data platform from the start.

Our agent doesn’t just write queries; it operates across the entire system — monitoring pipelines, managing metadata, creating dashboards, even relabeling models. The intelligence isn’t an accessory. It’s the operator.

Under the hood, we use the best open-source tech — DuckDB, Apache projects, and modern connectors — all stitched together from day one. Everything in Definite is text: SQL and YAML.

If you ever want to leave, you take your code with you. Not screenshots of a dashboard.

That’s what open infrastructure actually looks like.