AWS Buys the Team Behind DuckDB, and Says Agents Are Why
Amazon is acquiring DuckLabs, the thirty-person Amsterdam company behind the open-source database DuckDB. The project stays MIT-licensed, and AWS is unusually direct about the AI angle.
Amazon has signed a definitive agreement to acquire DuckLabs, the Amsterdam company behind DuckDB, the open-source database that has quietly become the default tool for anyone who needs to run a query over a few gigabytes of data without setting up a server. Terms were not disclosed. The deal is expected to close in early September.
The reassuring parts first, because open-source acquisitions usually come with an anxious week. DuckDB itself is not being acquired. The project stays under the independent DuckDB Foundation and remains available under the MIT licence exactly as it is today. Hannes Mühleisen and Mark Raasveldt, who created DuckDB and co-founded DuckLabs, keep leading both the team and the project’s technical direction inside AWS. DuckLabs is just over five years old, has grown past thirty employees, and notably never took outside investment: the founders and staff owned it outright and funded growth through support contracts.
DuckDB’s premise was that everyone had built databases for enormous datasets while ignoring the ordinary case. More than 90 percent of real-world queries run over a terabyte of data or less. DuckDB runs in-process, meaning inside your application rather than as a separate service you connect to, and reads files like Parquet, CSV and JSON directly from local disk or cloud storage. The Allen Institute, quoted in the AWS post, says queries over scientific data that took minutes now come back in under a second.
What’s actually going on here: AWS is unusually blunt about the AI reasoning. In its own words, agents behave a lot like people when interacting with data: “They poke. They experiment. They run exploratory analysis on small data sets before figuring out what they really want to do.” An AI agent asking twenty exploratory questions in a row is exactly the workload DuckDB was built for and exactly the workload a heavyweight cluster engine handles badly. Amazon has already been eating its own cooking here, processing over 2.5 billion queries through DuckDB integrations in its Amazon Quick product since October 2025 and cutting average query latency by 30 percent. The wider strategy is turning S3, Amazon’s storage service, into somewhere you analyse data rather than merely park it.
What this means for you: if you have never heard of DuckDB, this is a good moment to notice it exists. It installs in seconds, needs no server, and will happily run SQL over a spreadsheet export on your laptop. That makes it one of the more useful things to point an AI coding assistant at, since you can ask for a query and actually run it without infrastructure. If you already use DuckDB in production, the licence and governance guarantees look genuine, but the honest caveat is that these promises are made at announcement time by every acquirer, and the thing to watch over the next year is whether new features land upstream or only inside AWS services. For everyone else, it is a small reminder that the AI boom is now reshaping unglamorous infrastructure, not just chatbots.
Sources
Source: https://aws.amazon.com/blogs/big-data/aws-and-ducklabs-building-the-future-of-analytics-together/
Amazon Is Closing Mechanical Turk, the Human Workforce It Once Sold as AI
The platform that Jeff Bezos called artificial artificial intelligence shuts down on September 30 after 21 years, along with SageMaker Ground Truth and Augmented AI.
