AWS is buying DuckDB's maker, not DuckDB, and the mid-market gap now has an owner

What moves in this deal is payroll, not copyright. AWS is buying DuckLabs, the Amsterdam company whose staff write the code ^[1][3]. The code itself, along with DuckLake and Quack, stays where it already sits, under the non-profit DuckDB Foundation on the MIT license ^[4]. Hannes Muehleisen and Mark Raasveldt, who created DuckDB and co-founded the company, keep running the team and setting technical direction ^[5], now from inside a Seattle-based buyer ^[3].

So the practical asset transferring is people and commercial work, not a repository ^[14]. That matters for the thing a permissive license is supposed to guarantee. Anyone may fork DuckDB, and the license terms do not change ^[4]. A fork also needs engineers who can maintain a vectorised analytical engine, and the shortest list of those engineers is now employed by AWS ^[5][14]. Independence held in a foundation charter is cheap. Independence backed by maintainers who draw a salary from somewhere else is the version that survives a roadmap dispute, and the source does not tell us that such maintainers exist.

The demand side explains why AWS wants the team rather than the trademark. Justin Copie, CEO of AWS premier partner Innovative Solutions, says customers reached for DuckDB to build rapid data warehouse prototypes and small-scale BI without standing up heavy infrastructure ^[9]. He is blunt about the AI framing too: everybody wants intelligent agents, almost nobody wants to pay to clean their data, and the honest first step is rarely a big platform buy ^[10]. That is a description of customers who deliberately avoided buying a warehouse.

Copie says the move he is watching is whether AWS turns this into a managed service, which he expects would become an on-ramp for customers to start small, prove value and grow into the bigger analytics stack ^[11]. Read that from the buyer's side. The free tool's value to AWS is measured in what the graduating workloads eventually pay, which gives AWS an interest in where the local, laptop-scale version stops being sufficient. Garman's promise is about license and cost of ownership: DuckDB stays free and open source, and AWS analytics gets faster, simpler and more cost-effective ^[6][7]. Nothing in the material commits AWS to keeping the open project at feature parity with anything it sells ^[15].

Copie's own summary is that the smartest part of the deal is what AWS did not buy, and that AWS is buying the talent while keeping the trust ^[12]. The founders describe resources and reach they could not have reached alone ^[13]. Both can be true and still leave a mid-market buyer with a narrower set of options than it had last week, because the tool it chose specifically to stay off a hyperscaler's meter is now maintained by one.