“You can rent a feature, but you can’t rent a foundation”: why MotherDuck bought the startup already powering its data pipelines

MotherDuck has bought Tower, the small infrastructure startup whose technology was already doing the heavy lifting behind MotherDuck’s AI-built data pipelines. The deal is MotherDuck’s first acquisition, and it folds both Tower’s software and its team into the warehouse company just as MotherDuck pushes harder into AI agents that can build and run pipelines on their own.

Tower itself is a young German startup founded in late 2024 by former Snowflake engineers Serhii Sokolenko and Brad Heller. Its pitch is simple enough to matter: AI can write pipeline code, but somebody still has to package it, deploy it, wire up credentials and keep it alive. Tower handles that unglamorous middle part. It acts as a managed runtime for Python pipelines, and its browser-based Tower Control tool can take a plain-language request, generate code, deploy it and run it.

MotherDuck had been heading in this direction already. The company, founded in 2022 by former Google BigQuery engineering lead Jordan Tigani, is built around DuckDB and a hybrid model that lets queries run locally, in MotherDuck’s cloud, or both. It has raised about $100 million. More recently, it has moved from storage and querying into agent execution, first with MCP access and then with Flights, a June feature that exposes a general-purpose Python runtime so agents can create, run and schedule data pipelines. Tower was the infrastructure underneath that feature.

That made the buyout feel less like a surprise than a confession. Tigani says MotherDuck initially hoped to recommend a third-party tool to customers, but AI changed the problem. Claude, he argues, can help write connectors; it can’t provide sandboxing and scheduling. Tower could, and it let MotherDuck ship Flights in weeks. Once jobs created inside MotherDuck were actually being executed by Tower, the company stopped seeing the runtime as a rentable feature and started seeing it as part of the foundation.

There’s a product story buried in the acquisition too. MotherDuck wants to combine Flights with Dives, its AI-generated visualization feature, so a Flight can expose stable URLs that act like data APIs while a Dive builds the front end around them. That would let agents and users build richer applications without handing broad write access to the underlying data. For Tower customers, the next step is migration toward MotherDuck, even if the products are not identical yet. MotherDuck also plans to eventually merge Tower-backed Flights with Ducklings, its serverless DuckDB instances, into one system with faster startup and stronger sandboxing.