← GermanyThe Exclusive FeatureShare your journey
The people who built Germany’s companies · November 2026

Generative AI / large language models

Aleph Alpha

Andrulis built a German-sovereign LLM, Luminous, for companies and administrations that did not want their data sent to US models.

Jonas Andrulis has said that being called 'Germany's answer to OpenAI' has cost him the odd sleepless night. It is an honest line from a man whose company was handed a national burden before it had much revenue to carry it.

Andrulis trained as an industrial engineer at the Karlsruhe Institute of Technology, a short drive from Heidelberg. Then he went to Silicon Valley and spent years at Apple, where one profile places him managing AI research in a special projects group from 2016. A German profile suggests Apple is where his world-changing ambition took shape. He returned in 2019 and registered Aleph Alpha in Heidelberg with Samuel Weinbach on 27 February. 'Let's change the world!' is a phrase he reaches for in keynotes. He also says, 'I believe we have a chance that's worth fighting for.'

Its statutory accounts for the year showed turnover under €1 million and a loss of €18.9 million.

The pitch was specific. European governments and companies were nervous about sending their data to American models, and nobody on the continent was building a serious alternative. Aleph Alpha aimed to make transparent models, Luminous among them, aimed at businesses and administrations rather than consumers. A €5.3 million seed came in 2020 and €23 million followed in 2021. In 2021 the team added models that read both text and images.

The moment that made the company a national story came in November 2023, when a round was announced at $500 million. It was read as a sign that Europe could field a rival. What almost nobody noticed at first was how the number was built. Trade press later reported that only about €110 million was genuine equity, with around €300 million going to research funding for a subsidiary and €60 million in order commitments.

The books were harder still. Aleph Alpha missed an internal 2023 sales target of $6 million. Its statutory accounts for the year showed turnover under €1 million and a loss of €18.9 million. Meanwhile US labs were spending at a scale no European startup could match.

The strategy bent. The company moved away from racing on the largest models and toward helping clients run AI in-house. In January 2026 Andrulis ceased to be a managing director, and the company moved to the supervisory board. In April 2026 Cohere announced it would acquire Aleph Alpha, with a definitive merger agreement in September, and the Heidelberg site is to stay as a research centre. In October 2026 the company still released Kolibri, an open-weight model under the Apache 2.0 licence.

For first-time founders, the lesson is uncomfortable. A big headline number can become a promise you have to live up to in public. Andrulis got the attention, the sleepless nights and a chance worth fighting for, and he learned that carrying a country's hopes is not the same thing as having a business that pays for them.

The Chronology year by year

  1. 2019
    • founded in Heidelberg
  2. 2021
    • multimodal prompting
  3. 2023
    • Multifusion at NeurIPS, $500M round announced
  4. 2026
    • Andrulis leaves as managing director
    • Cohere announces acquisition
    • merger agreement signed
    • Kolibri open-weight model released

The Money Trail how it was funded

  1. 2020 Seed €5.3M
  2. 2021 €23M
  3. Nov 2023 round announced at $500M (reportedly only ~€110M equity)

Principal backers: Schwarz Group, Bosch Ventures, Burda and others

The Close Calls what nearly broke it

  • 2023 sales target of $6M missed
  • 2023 turnover below €1M with €18.9M loss; controversy over announced vs real funding (2023-24)

In Their Own Words from the record

Spotted something wrong? Send a correction.

Startup Talk · Preview edition · Figures are from public sources and may be out of date
About · Contact · Terms · Privacy · Badge · Share your journey