German AI's second wave: DeepL, Black Forest Labs, Parloa, n8n and the Aleph Alpha lesson
Germany's most successful AI companies have not tried to out-build OpenAI. They picked narrow, valuable jobs: translation, images, customer calls, workflows. The exception, Aleph Alpha, is now merging into Cohere.
The first wave of European generative AI was framed as a race to build home-grown rivals to the large American model makers. The most durable German successes have taken a different path. They chose specific, commercially valuable tasks and built products that do one thing very well. Taken together, they offer a clearer template for European AI than any single "national champion".
DeepL: translation as a product, not a feature
DeepL grew out of Linguee, the online dictionary. A team led by Jarosław Kutyłowski developed a neural translation system there, and the DeepL Translator launched in August 2017, with the company renamed DeepL. It is headquartered in Cologne. Its strategy has been to treat language as an enterprise workflow rather than a consumer gimmick: DeepL Write for improving text, DeepL Voice for real-time spoken translation and, from November 2025, DeepL Agent, an autonomous assistant for knowledge workers.
According to the company, DeepL has more than 1,000 employees and over 200,000 business customers. In November 2025 its chief executive joined the Federal Chancellor's Strategy Group for Technology and Innovation, a sign of how seriously Berlin now takes applied AI.
Black Forest Labs: frontier models from Freiburg
Black Forest Labs, based in Freiburg and co-founded by Robin Rombach and Andreas Blattmann, develops the FLUX family of image generation and editing models. Its FLUX Kontext model was designed to edit images while keeping characters consistent, a practical requirement for commercial design work. The company's business model mixes open-weight releases under non-commercial licences with paid commercial licences and an API for production use. Its models are used by large platforms, including design tools, which shows that a small team in the Black Forest can compete at the frontier in one modality.
Parloa: AI agents on the phone
Parloa, founded in Berlin by Malte Kosub and Stefan Ostwald, builds a platform for companies to design, test and run AI agents for customer service conversations, including voice calls. It has since opened a US headquarters in New York alongside its offices in Berlin and Munich. In January 2026 the company announced a 350 million US dollar Series D round led by General Catalyst, at a valuation of three billion dollars, according to Parloa. Customer service is an attractive niche: calls are expensive, measurable and highly repetitive, which makes the return on automation easy to demonstrate.
n8n: automation for technical teams
n8n, founded in Berlin by Jan Oberhauser in 2019, offers a workflow automation platform that connects applications, APIs and AI models. Users can build flows visually or write their own JavaScript or Python, and can run the software on their own servers or in the cloud. It is distributed under what the company calls a fair-code licence, which keeps the source visible while restricting certain commercial uses. Self-hosting matters for European companies that want to keep data under their own control, and n8n's large community of template and node contributors has become a distribution engine in its own right.
Aleph Alpha: the sovereign model that changed course
Aleph Alpha, from Heidelberg, was for several years the standard-bearer for German "sovereign AI": large language models built and hosted in Europe for public sector and regulated customers. The economics of competing on general-purpose foundation models proved difficult. Its founder, Jonas Andrulis, gave up the chief executive role in October 2025. In April 2026 the Canadian AI company Cohere announced plans to combine with Aleph Alpha, and the two signed a definitive agreement in September 2026. The combined business will operate as Cohere, with headquarters in Toronto and Berlin, and the Heidelberg office is to continue as a research centre. At the time of signing the transaction still required final regulatory approval and was expected to close later in 2026.
The lesson is not that European foundation models are impossible. It is that scale matters enormously in general-purpose models, and that a niche, data or distribution advantage is a more reliable foundation for an AI company of German size.
What the second wave has in common
- A defined job: translate, generate images, answer calls, automate workflows.
- Enterprise focus: revenue from businesses that value reliability, privacy and integration.
- European data sensitivity as an asset: self-hosting options and EU operations are selling points, not just compliance costs.
- Global ambitions from day one: several have US offices and customers worldwide.
For investors and corporate buyers, the message is that German AI is no longer a single bet on one sovereign model. It is a portfolio of specialised products, some of which already lead their categories.