--- title: "kausable closes €12 million seed round in Heidelberg; Seed funds to scale European causal AI" sdDatePublished: "2026-08-31T10:07:00Z" source: "https://www.htgf.de/en/kausable-interview/" topics: - name: "artificial intelligence" identifier: "medtop:20001298" - name: "start-up and entrepreneurial business" identifier: "medtop:20001158" - name: "business financing" identifier: "medtop:20000183" - name: "scientific innovation" identifier: "medtop:20000736" - name: "research and development" identifier: "medtop:20000208" - name: "university" identifier: "medtop:20000407" locations: - "Heidelberg" - "Germany" --- kausable closes €12 million seed round in Heidelberg; Seed funds to scale European causal AI Causal AI from Europe: How kausable wants to build the next generation of AI Causal AI from Europe: How kausable wants to build the next generation of AI Heidelberg-based AI startup kausable recently closed a €12 million seed financing round. The round was led by UVC Partners and Entourage, with HTGF and Mätch VC also participating. Founded in 2025, the company is developing causal AI that adapts to changing situations using only a small amount of new information. In this interview, CEO and co-founder Johannes Haux explains how kausable came to be founded, what lies behind the “reasoning-first” approach, and what concrete benefits the technology can offer industrial companies. Hendryk Hosemann, Senior Investment Manager at HTGF, talks about the investment thesis, the founding team, and the open questions on the road to scaling. How did kausable come to be founded? Johannes: My co-founders Benjamin, Gregor and I know each other from our time together as researchers at Heidelberg University. But we didn’t start the company straight out of research. By that point we had all already been away from university for a while: Benjamin, our CTO, worked at secunet after his PhD. Gregor, our COO, was at Atruvia, among other places, and I had already founded a company before and then worked at AskUI. Ultimately, kausable took shape as an EXIST project at SRH Heidelberg. An important reason for us was that we wanted to create something fundamentally new — and measure it against the one metric that really matters: are we building something that delivers real value, something people buy? We were often asked why we didn’t simply turn it into a research group. That’s exactly why: we want to build something that genuinely benefits people. On top of that, we saw the potential early on. We know the founders of Black Forest Labs very well — we worked with them at university and sometimes even shared an office. So, we have some sense of what it looks like when a still-raw idea is on the verge of a breakthrough. When we were thinking about founding the company in 2024, we had exactly that feeling: this could become something big. We should try it as a company. Hendryk, what were the decisive reasons for you to invest in kausable? Hendryk: For me, two things are decisive in an early-stage investment: the team and the market dynamics. With kausable, both are spot on. We believe in a new wave of AI focused on Europe’s core industries. There, high quality and regulatory demands meet very complex processes and often limited amounts of available data. Classical, purely correlation-based AI models and applications are not sufficient for that. What’s needed is AI that works deterministically and can be deployed with only a small amount of existing data. kausable hits that thesis right on the mark. From the very first conversation, it was clear to us that we had found a unique team capable of both driving and shaping this strong market dynamic — technologically and commercially. Johannes, you talk about causal AI, a “reasoning-first” approach, and a world model. What does that mean in concrete terms — and what does your approach promise compared to large LLM providers? Johannes: The term “world model” has become quite overused by now, so I try to define it as precisely as possible. When we as humans learn to open a door, we don’t need to run through the door a million times or watch hours of videos. We see two or three demonstrations, try it ourselves once or twice, and then we’ve got it. By doing this, we update our understanding of how the world works — in this case, how a door operates. For me, that is a kind of world model — something I carry within me that enables me to make sense of the world. We’re building a model that very efficiently constructs the right world model for the specific situation at hand, without needing to be explicitly retrained for it. This adaptation doesn’t require classical gradient updates. You show the model the data, it processes it, and makes a prediction based on the best understanding it can derive from that. This is a fundamentally different approach from what is mostly done today. A direct comparison with LLMs doesn’t necessarily make sense. They perform reasoning based on language, while our model is very data-agnostic. We don’t yet work with text, but we can already handle a range of different dynamic systems. A language model brings a lot of general knowledge with it and is a wonderful interface to humans. Our model, by contrast, brings virtually no general knowledge — apart from the assumption that the world operates through cause-and-effect relationships — but it can adapt to new situations and data. We therefore see both approaches as complementary technologies: the language model forms the natural interface to humans, underneath which lies a more flexible layer that can natively work with different types of data. Combining the two is likely to produce very powerful technology. What results already show that your approach transfers across different fields of application? Johannes: We already have solid results in predicting critical points in complex systems. We’ve submitted a paper on this that is currently under review and examines around 15 domains. After the paper was finished, we added even more domains. For example, we can predict blackouts in power grids with several minutes of lead time. For epileptic seizures, we achieve at least one minute of lead time, based on EEG data. These are time windows in which you can meaningfully react and that’s exactly where the value becomes concrete. What’s especially exciting for us is that the model can handle a very large and diverse portfolio of datasets. That’s a strong early sign that our approach transfers across different dynamic systems. Of course, these are research results for now. The next step is to validate them together with partners in concrete applications. What concrete problem can kausable solve for industrial companies? Johannes: One project we’re currently looking at with a potential design partner revolves around what-if questions. If, for example, I want to adjust a price and understand the effects of that. To do this, I not only need to understand a complex dynamic system and make predictions — I also need to recognize that I am actively intervening, that I am nudging a variable, and that this is something different from simply letting the system run. We can do that, and it’s exactly what we’re exploring right now with potential design partners. Another exciting field is robotics — for example, grasping soft objects. A grape may be firm one day and soft the next. If the robot crushes the first grape, it needs to immediately recognize that something is different and adjust its grip strength using only a small amount of data. That kind of learning-in-the-moment is what we find exciting. The benefit becomes measurable through better predictions, faster adaptation, less data required, and — depending on the application — for example less scrap, fewer failures or fewer wrong decisions. We’re still in the design-partner phase here. The task now is to translate the technological advantage, together with customers, into clearly measurable economic benefit. Hendryk: What impressed us about kausable was the strong interest from within our LP network. There is a wide range of needs and already very concrete ideas about where deterministic AI will create value. Although the possible fields of application are very broad, there are already exactly these kinds of concrete examples, as Johannes described. Interestingly, kausable doesn’t just address classic engineering problems across a wide range of industries, but also very strategic and commercial challenges. After the pre-seed round in 2025, you have now raised €12 million in seed capital. What goals have you set for yourselves? Johannes: The biggest proof point for us is the proof of value. In the pre-seed phase, we built our research team and wrote our first publications. The next step now is to move into application. There’s still a lot of research work to do, but we no longer want to do it in a vacuum. We want to do it together with design partners and real users. This will provide us with concrete feedback and demonstrate that our technology genuinely works and creates value, rather than just being a pipe dream. The great thing about our technology is its scaling behavior. A language model trained on very little data typically doesn’t yet produce usable results. With our approach, however, even smaller, cost-efficient models can solve valuable problems. As we scale up, we can tackle correspondingly more complex problems, enabling us to collaborate with partners right now. Our goal is to deliver, by the end of the seed phase, very compelling proof of two things: that we can continue doing research at the cutting edge, and that at the same time we’re building something with really great potential. Hendryk, which milestones are particularly important from an investor’s perspective? Hendryk: For us, kausable works in many ways like a classic deep tech case, just without hardware. We were especially impressed that Johannes and his team thought commercially very quickly and started early on developing the technology together with real customer feedback. For a top-tier scientific team, that is absolutely not a given, and perhaps not even a core requirement at the seed stage. For us, it’s a very strong indicator and the right path to setting the right course for the long term. Interest in foundation labs in Europe is very high. But if we’re honest, there are only very few teams that are even capable of developing something this fundamental. That’s why we assume, just as Johannes mentioned, that the next financing round will also be strongly driven by research results. The design partnerships will be helpful here and will show in which fields of application the technology can translate into concrete customer value. What have you learned since founding the company, Johannes? And what would you pass on to other founders? Johannes: One important lesson is: it is possible to finance difficult, deep tech topics with VC capital in Europe and Germany. In the beginning, we were often told it would be really, really difficult. It wasn’t easy — but it is possible. With Entourage and UVC Partners, we’ve won over two very strong European VCs who understand what a deep tech bet means. We already experienced that in the pre-seed phase with Mätch VC as lead, and we’re experiencing it again now with UVC, Entourage and, of course, HTGF. A special shoutout goes to Hendryk, who really got behind this. As a founder, you’re allowed to say boldly: we’re doing something really difficult, we have a long road ahead of us, and in the first few years we may not have significant revenue yet. That’s a good filter: you get a “no” faster, but you end up having the right conversations. A second lesson is: doing helps. I’m very good at mulling things over and developing ideas. But what mattered was getting started and having Benjamin and Gregor as the right co-founders. We push each other, keep the pressure up together, and stay consistently on it. Thirdly, to be honest, it’s really exhausting. During fundraising, you get a lot of rejections and very often hear, ‘Interesting. Get back in touch once you’ve proven more.’ That’s frustrating. During the financing phase, I took a holiday that wasn’t really a holiday. You shouldn’t underestimate that. When it does work out, it feels especially good precisely because of that. What excites me anew every day is the chance to work with people who are smarter than me in many ways. Behind kausable stands a genuinely strong team. The opportunity to work with these people is the real driving force behind all this work and what motivates me to go to the office every day. Hendryk, what