Over the weekend, the data community woke up to surprising news: Fivetran is planning to buy dbt Labs. This comes just three weeks after the announcement that sqlmesh, a dbt competitor, is also being acquired by Fivetran.

Fivetran is the bigger fish here: roughly $300M ARR compared to dbt Labs’ $100M. Together with sqlmesh, the ETL tool triumvirate forms what you could call a transformations superhouse. Throw a stone into a crowd of data folks and chances are you’ll hit someone using one of these three tools - whether to move data from A to B, or to shape the warehouse once the raw data has landed.

Fivetran’s strength has always been extraction and loading, the E and L of ETL. It became the go-to way to pull from dozens of data sources into Snowflake, BigQuery, Redshift, and others. Once the raw or bronze layer is there, dbt and sqlmesh step in: define models, build dependencies, and generate the silver and gold layers (if you follow the Medallion view) or facts and dimensions (if you don’t).

Why the sudden appetite for T?

Nobody knows for sure, but two explanations sound reasonable:

More compute = more dollars

Extraction and loading only capture maybe 25% of a company’s analytical compute. Transformations eat up another big slice, perhaps another 25%. By owning T, and becoming a full ETL tool, Fivetran has doubled the market it can address.

Future survival of ETL tools

Once open table formats like Iceberg, Hudi, or Delta Lake really spread, the E/L business weakens. Customers won’t need to shuttle data between multiple data silos - they’ll own their tables and have just one “silo”. The money saved from E/L will partly flow into transformations instead, and partly move into other IT areas.

What could happen next?

A new data warehouse?

With E and L locked in, plus the leading T layer and warehouse-update features, Fivetran is inching toward offering a full warehouse of its own. Combine that with Iceberg, and why couldn’t they challenge Snowflake or Databricks SQL? Imagine: Fivetran SQL.

Room for new ETL tool vendors

If Fivetran offers a data warehouse, other warehouse vendors will need a neutral ETL partner. At events like Snowflake Summit or Databricks Summit, Fivetran was always the friendly third-party mover of data. That balance could break. Who steps in? Kafka-based options, Debezium, or libraries like dltHub are all in the wings.

Another gorilla in the SQL zoo

The SQL analytics arena just got a new heavyweight alongside Snowflake, BigQuery, Databricks SQL, Redshift, and Fabric. But history shows: every time a challenger is bought, five new challengers appear (and four of them later die). Customers dislike monopoly. dbt itself was born as an open-source response to the closed ETL tools of the 2000s (Informatica, SSIS etc). sqlmesh appeared as a counterweight to dbt’s proprietary drift. So - who builds the next dbt or sqlmesh?

What it means for Python data engineering

And then there are transformations in Python pipelines. Transformation libraries like Polars and Apache DataFusion are getting stronger, but none yet has the connector ecosystem that drives adoption. Connectors are the unsung heroes of ETL tools: they handle schema detection, change capture, backpressure propagation, and make or break pipeline reliability. Without them, even the fastest engine stalls.

The open question: will the Python ecosystem grow into a true rival for Fivetran? Tools like dltHub already provide hundreds of connectors. Combine that with Polars, and you have a working Python ETL stack - runnable on platforms like Tower.dev. It’s not the same scale as Fivetran yet, but the pieces are coming together.

Closing thought on ETL tool consolidation

The consolidation we see today doesn’t end the story. It just sets up the next cycle: new warehouses, new ETL tools, new Python libraries. The gorilla fight is far from over.