MotherDuck Acquires Tower, the Startup Already Powering Its AI-Generated Data Pipelines
MotherDuck bought Tower Computing Inc., the runtime provider already powering its Flights feature for AI-built data pipelines.
Overview
Data warehouse company MotherDuck has acquired Tower Computing Inc., a startup that was already running the infrastructure behind MotherDuck’s AI-generated data pipelines, according to MotherDuck’s announcement. MotherDuck, which is built on the open-source DuckDB database and was founded in 2022 by Jordan Tigani, a former Google/BigQuery engineering lead, has raised approximately $100 million since inception, according to The New Stack.
What We Know
Tower was founded in late 2024 by former Snowflake engineers Serhii Sokolenko and Brad Heller, and it provides a managed runtime for Python data pipelines, handling packaging, deployment, and production maintenance, according to The New Stack. Tower was already powering MotherDuck’s Flights feature — a mechanism, launched recently, that lets AI agents create and schedule data pipelines through a general-purpose Python runtime, according to The New Stack and MotherDuck.
“We became their largest customer almost overnight, and our teams have been shipping together ever since,” Tigani told The New Stack. Explaining the rationale for buying rather than continuing to license Tower’s technology, Tigani said, “There’s a rule I’ve relearned at every infrastructure company I’ve worked at: you can rent a feature, but you can’t rent a foundation,” adding that “when an agent inside MotherDuck builds a job and schedules it, the thing executing that job is our product — whatever logo is on it,” according to The New Stack.
In MotherDuck’s own announcement, the company said the parts of the pipeline that its AI coding wasn’t handling — “the sandboxing, scheduling, and observability” — were exactly what Tower provided, and that building on Tower let MotherDuck launch Flights “in only a matter of weeks without compromising on features like observability and reliability.” MotherDuck also said the acquisition will let it turn Flights into callable “Data APIs”: a Tower job can expose a stable URL, so a Flight can become a custom data API, which can be paired with MotherDuck’s hosted visualization tool, Dives, to build applications from a single prompt, according to MotherDuck.
Sokolenko, for his part, contrasted joining MotherDuck with being acquired by a larger cloud provider, telling The New Stack, “Joining a hyperscaler usually means adapting to its legacy architecture.” Tower’s technology also underpins a feature called Tower Control, an AI agent that converts plain-language descriptions into deployable pipelines without manual infrastructure setup, according to The New Stack. MotherDuck plans to eventually merge Flights with Ducklings, its serverless DuckDB instances, according to The New Stack.
The deal follows a broader industry pattern of data-platform vendors building agentic tooling that lets AI generate, deploy, and operate data infrastructure on its own: The New Stack notes that Databricks offers a comparable capability called Genie Code, and Snowflake offers CoCo and CoCo Automations, according to The New Stack.
What We Don’t Know
Neither MotherDuck’s announcement nor The New Stack’s report discloses the financial terms of the acquisition. The exact timeline for merging Flights with Ducklings has not been specified.
Analysis
The acquisition is narrower than the DuckDB ecosystem’s other recent deal: AWS’s agreement to acquire DuckLabs, the company behind the open-source DuckDB engine itself, is a separate transaction with a different acquirer and a different target. MotherDuck buying Tower is instead a vertical move by a single vendor to own the execution layer underneath its own AI-agent pipeline product, rather than continuing to depend on an outside runtime provider for a feature it says has become central to its roadmap.