Warehouse Compute Tax Drives Vendor Consolidation
Charles Chretien, co-founder of Prequel, on the modern data stack’s ROI problem
The strategic point is that modern data tools win budget by sitting on top of the warehouse, but they also create a hidden tax by making the warehouse do the heavy lifting. Fivetran keeps raw data flowing in, dbt turns that raw data into clean tables with lots of SQL jobs, and every sync and transform burns Snowflake or Databricks compute. That is why vendors next to the warehouse keep expanding into adjacent layers, because the workflow that spends the compute often captures the roadmap and the budget.
-
Fivetran started as the E and L in ELT, then moved toward T with dbt. The logic is simple, once a team already trusts one vendor to land data in the warehouse, adding transformation lets that vendor own more of the daily pipeline and more of the spend tied to it.
-
The warehouse centric stack works by pushing almost everything into SQL on Snowflake, BigQuery, or Databricks. Census described the whole ecosystem as tools speaking SQL to the warehouse, which means ingestion, modeling, and activation all trigger warehouse jobs instead of running on a separate application server.
-
This compute burden is one reason the market shifted from horizontal plumbing to more packaged products like CDPs. In marketing use cases, teams can trace warehouse work to better ad targeting, conversion, and campaign performance, which is easier to defend than paying for large volumes of background data processing alone.
The next step is for pipeline vendors and warehouses to collapse further into each other. The company that controls both the workflow and the compute bill will own a larger share of the data stack, while standalone point tools will keep getting folded into broader platforms built around concrete business outcomes.