German Research Center for Artificial Intelligence (DFKI) Gmbh

Germany's largest AI research institute is usually described by its aggregate statistics — departments, staff, spin-offs — most of which are self-reported and, as this article documents, internally inconsistent. A different picture emerges when the question is narrowed to something checkable: what does the German Research Center for Artificial Intelligence (DFKI) GmbH actually operate? The answer, assembled from DFKI's own documentation and third-party grant records, is a set of named, dated systems — a documented HPC cluster with a public partition table, a dated GPU procurement timeline, one of Europe's larger underwater robotics test halls, a factory testbed selected for the Gaia-X data-infrastructure project, and a sequence of funded projects with identifiable hardware. It is also a picture with clear boundaries: several headline numbers the institute publishes about itself do not agree across its own pages.

German Research Center for Artificial Intelligence (DFKI) GmbH describes itself as Germany's largest AI research institute. Unlike many research brands, its governance leaves a dated paper trail: a supervisory board drawn from ministries, universities and industrial shareholders, a management board whose appointments are announced with names and dates, and a chair succession on 2025-11-20 that changed leadership at the top of the institute while its scale figures remain self-reported. This briefing reads that record rather than the reputation.

The German Research Center for Artificial Intelligence is usually introduced as Germany's largest AI research institute and as a non-profit public-private partnership. Both labels imply a mechanism: that public money and industrial money together push research out of the laboratory. This article tests that mechanism against dated documents rather than reputation. It keeps three things apart that institutional profiles tend to merge — the money the institute is promised, the money it reports under several different definitions, and the systems that can actually be identified in a market with a date attached.