Authors: Bojan Stojkovski and Bogdan Iordache
Rasmus Rothe is a visionary in artificial intelligence; founder and GP of Merantix Capital, and co-founder of the Merantix Group, a group of companies and initiatives driving AI forward in Europe. With a rich academic background from Oxford, Princeton, and ETH Zurich, and over 15 years of research in deep learning, during the past decade, Rothe has established himself as a leader in translating AI research into real-world applications.
Through the Merantix Group, he is looking to leverage AI to solve some of the world’s most pressing problems, from healthcare to climate technology. His work spans multiple sectors, from predictive maintenance and smart factories to AI-powered solutions for industries struggling to adopt next-generation technologies.
In this interview with Underline Ventures, Rothe shares his insights on the evolution of AI, the opportunities and challenges in Europe’s industrial landscape, and how Merantix is bridging the gap between academic breakthroughs and practical AI solutions.
Underline Ventures: How does Merantix turn theoretical AI research into practical applications, and how has this process impacted your own experience?
Rasmus Rothe: The Merantix Group operates with three main pillars. The first, Merantix Capital, is our investment arm that functions as a venture fund. This arm invests at the company formation stage, even before a team or idea is fully formed, as well as at the pre-seed and seed stages. This approach allows Merantix Capital to partner with founders who are still figuring out their team and business model and with startups that are ready for a seed round.
The second pillar is Merantix Momentum, an AI services business with nearly 100 employees. This team builds customized AI solutions for corporations, governments, and SMEs. It also includes a research team that publishes at leading conferences like NeurIPS and ICML.
The third pillar is a community and thought leadership ecosystem made up of different entities. The largest, Merantix AI Campus, is a co-working and events hub located in Berlin with 1500 members and around 300 events annually. Recently, the organization expanded by opening the London AI Hub, together with the Founders Forum. Additionally, we organize the AI House in Davos, a think tank conference during the World Economic Forum. Regarding the transition from research to production, the Merantix Group stays closely connected to academic advancements through its research team, the co-located researchers on campus, and its involvement in major machine learning conferences.
UV: You mentioned organizing the AI House during the Davos conference. This forum brings together industry leaders, researchers, and policymakers to discuss AI. How has this multi-stakeholder forum evolved over the years?
RR: We saw the World Economic Forum as an opportunity to connect the academic community with industry leaders, startups, and policymakers by creating a think tank environment. The AI House in Davos is a collaborative, non-profit initiative involving around 40 partner organizations, including leading technical universities like ETH Zurich and EPFL and international partners such as the University of Tokyo and G42 from the Middle East.
The goal is to bring together top researchers, like professors from ETH and AI pioneers such as Yoshua Bengio and Yann LeCun, with CEOs of major corporations, startup founders, and policymakers. This diverse group discusses emerging opportunities, regulatory directions, and geopolitical impacts on AI. The event fosters dialogue but also aims to build meaningful relationships that can lead to new companies, collaborations, and investments.
We also see a gap in Europe’s ecosystem compared to the US, where regions like Boston and the Bay Area have tightly connected networks of universities, VCs, policymakers, and big tech companies. In contrast, Europe’s ecosystem is less integrated, so Merantix aims to bridge these gaps and strengthen connectivity across stakeholders.
UV: What are you looking for in founders, and what is your thought process behind an investment decision?
RR: At Merantix Capital, we often look at megatrends, where we might not have a fully developed use case yet, but we see significant potential. From there, we talk to experts in that space to gain insights, as we can’t fully validate a business case ourselves. The founder must own the idea, so we don’t focus too much on specific details from the start.
For example, we’re excited about cybersecurity because we believe AI introduces new risks, such as phishing, threats to critical infrastructure, and AI-driven cyberattacks. We’re exploring how we could build a company in this space, but where exactly we go depends on the founders we meet and the opportunities we uncover.
Another area we’re focused on is looking for strong system-of-record companies and assessing whether we can build workflow solutions on top of them. These might even eventually replace the original systems, like CRM, ERP, or EHR systems. This approach is industry-agnostic—we’re open to any sector where an established system could be disrupted.
UV: Can you provide an example of a company in your portfolio that resulted from this process?
RR: There are many, but one of the standout examples is Deltia, where we are joint investors. We’ve always been excited about industries where we have deep domain expertise in Europe. I’ve known Max Fischer for nearly 15 years, and I’ve watched his work in the industrial sector. When Max was thinking of starting something new, I suggested exploring AI use cases in the industrial context, given the advancements in technology. Max has extensive experience, having worked in hundreds, if not thousands, of factories, so we decided to spend a few months exploring different business cases to identify a venture-scale opportunity.
Another example is in the fertility space. We’ve been excited about this area for years, as declining fertility rates present a significant social issue, especially in Europe, where fewer children are being born. We believe data could make a difference in addressing this challenge. After years of thinking about the space, we met Felicia (Ovom Care) and realized they were the right team to tackle the problem. This came two years after we first had the idea.
In both cases, we didn’t bring an exact startup idea to the table but focused on broad themes, which is how our process works. We are very thesis-driven, looking for opportunities in areas we believe have potential but waiting for the right team and circumstances to align.
UV: You’ve been building and investing in this field since 2016. When did you notice the biggest shifts in turning AI research into practical solutions, and when did this start happening in Europe?
RR: There have been two major shifts. The first occurred around 2016-2017 when things started to work in the computer vision sector. I remember doing my PhD in computer vision between 2013 and 2016, and when I started, I was the first person in my lab to train neural networks. At the time, computer vision was still largely separate from neural networks. Then, in 2012, Alex Krizhevsky’s breakthrough with ImageNet showed that AI and computer vision could work with neural networks. Between 2013 and 2016, the field shifted from being purely academic to involving more industry players. This was when we started Merantix Capital, realizing that computer vision was now ready to be applied in the industry.
The second major shift happened on November 30, 2022—an unforgettable day for me as it was also the day I got engaged. On that day, people began to realize that AI was not only working for vision but also for text. Even before that, we were experimenting with language models and saw significant improvements. From a technical perspective, this shift was marked by the public’s perception of AI, which made it easier to convince people that AI was here to stay. The business case also became clearer as AI started working at a level that made sense commercially.
Looking ahead, I think the next big shift will be in robotics, particularly with the development of smart humanoid robots that are robust and flexible across a broad set of tasks.
UV: What are the biggest challenges industrial companies face when trying to keep up with the newest technologies?
RR: There are a few macro trends to consider. For example, certain sectors like automotive in Europe are struggling, which isn’t ideal. Another challenge is that much of the data in industries like customer support or legal is non-digital or stored in different systems. In these industries, data is often in documents, such as contracts or customer support logs. AI solutions can easily be built on top of that data to improve tasks, but in the industrial context, we’re dealing with the physical world. A lot of machine data isn’t stored in centralized systems, and some data isn’t available at all. The integration process becomes cumbersome, and the lack of standardization in factories complicates the situation.
However, there is progress being made, such as the ability to use cameras to gather data instead of connecting directly to machines. This leapfrog innovation is already helping. Despite the challenges, there is tremendous potential in the industrial sector, especially in Europe. But it’s more difficult compared to other industries, where tasks can be done remotely, like law or content creation. In those fields, all data is typically digital and easier to manage.
The key issue with industries like manufacturing is that data isn’t always digital, which limits the effectiveness of AI models. In comparison, industries like law or content creation can more easily digitize data, such as emails, voice, and documents. For industries that require physical presence—like manufacturing—the gap between human and AI capabilities can be hard to close if the data isn’t available. The exciting part is figuring out how to bring AI into these industries, bridging that data gap, and unlocking the power of AI. This challenge also extends to sectors like defense, logistics, and space.
UV: Amid budget constraints and productivity demands, how will European manufacturing adopt AI and next-gen factory solutions?
RR: Some industrial sectors are actually doing quite well, like intralogistics or those providing solutions for data centers. In parts of the industrial space outside of automotive, there’s growth as well. However, the European manufacturing industry must become more efficient to compete internationally, especially with countries like China that are heavily subsidizing industries, and the US, which is aggressively investing. European industries have no choice but to adopt more advanced technologies, including AI, to stay competitive.
To achieve this, clear ROI is essential. Many industrial companies currently lack the budget for projects without tangible returns. However, if AI solutions can be pitched with a clear business case—showing how they can save costs, increase efficiency, or reduce labor costs—there is potential for investment.
UV: Will future technology help incumbents improve or favor new players as leaders?
RR: Scaling quickly is an advantage, but often the real value lies in the process of producing the product itself. For example, if you create a highly specific or standardized machine, you could potentially build factories at scale, requiring more capital expenditure. However, only a small subset of VCs fund such ventures. That said, sometimes it’s worthwhile, as seen with companies like Hadron or those in the defense space that build their drones to capture more value because that’s what customers demand. Customers want more integrated solutions.
This approach also applies to other industries. For instance, in legal tech, software can enhance lawyers’ capabilities. However, a more vertically integrated model could involve selling a tool to corporate legal departments that allows them to handle legal questions internally, reducing the need to hire external lawyers. This model captures more value and democratizes knowledge within the company. Alternatively, you could build a fully AI-enabled law firm that offers services more efficiently and at a lower cost.
Both models—helping existing businesses or completely replacing them—work, and many startups pursue them. In the legal sector, some firms are becoming full-stack, while others work within corporate legal departments. Similarly, in manufacturing, you could buy a traditional business, AI-enable it, and transform it into something more efficient. Private equity investors are particularly interested in this approach, as it offers an opportunity to merge AI with traditional businesses for a significant impact.
UV: How do you see the role of a strong ecosystem in scaling AI applications across industries?
RR: I think ecosystems are crucial because they allow you to focus on the right use cases and scale quickly – which is why after establishing Merantix Capital, we quickly set out to grow the much larger Merantix Group. You need a mix of investors, customers, founders, researchers, and policymakers all in one space. Founders need capital, access to customers for design partnerships, and an environment to quickly iterate and learn. Policy support is essential, and exchanging ideas with others disrupting industries is key. Peer-to-peer learning is also incredibly valuable—our CTOs and CEOs often share insights on common challenges like outbound sales, tech stacks, human-in-the-loop systems, regulation, expansion, and hiring. Many of these challenges are industry-agnostic, and there’s overlap across AI-first companies.
We prioritize connecting our portfolio companies and supporting the broader ecosystem, particularly in Europe. We believe in the value of physical ecosystems. The concept of hubs, where companies can meet, collaborate, and learn from one another, is crucial. We’re focusing on how to connect these hubs—ensuring people can travel between them and potentially open offices in different locations. In the future of work, we see hubs as more than just workspaces—they are places for collaboration and innovation, providing a club-like atmosphere where teams work together and with other companies.
UV: What do you think are the key factors needed to support AI innovation in Europe?
RR: First, I think the public debate in Germany is misfocused. When we talk about the economy, topics like migration often take precedence, while digital transformation, AI, and startups should be at the heart of the discussion. We shouldn’t view the economy as separate from technology; in fact, digital innovation is the playground that enables the economy to remain relevant in the future. So, when discussing the economy, we need to integrate the conversation about AI, digital technologies, and the like, rather than treating them as distinct subjects. This separation is a big problem in policy.
Secondly, we need to stop trying to imitate the US. The US is focused on foundation models, and now European companies are following suit. But this is a race we’re unlikely to win, as these models are getting commoditized, and the costs of switching are low. Even the business case for foundation models is challenging. While it’s essential for Europe to support initiatives like Mistral, simply trying to replicate OpenAI will not transform the European economy. Instead, we should focus on integrating AI into verticals like industrial sectors, life sciences, healthcare, and FinTech, where Europe still has a competitive edge with its deep domain expertise.
European companies also tend to prefer European solutions. It’s not the hardware or infrastructure that makes the biggest impact—it’s the applications. Just as the internet’s value came from companies like Amazon and Google that built on top of the existing infrastructure, the value in AI will come from the applications built in sectors such as healthcare and manufacturing, which contain proprietary data that isn’t available to U.S. companies. We need to make sure these companies are founded in Europe and focused on these verticals.
UV: Do you think there will be further pushback from incumbents, syndicates, and other groups that are focused on preserving the existing social order as we move toward a future where AI increases productivity and transforms industries?
RR: I think it’s a huge problem, especially in Germany. There are many government subsidies for industries like energy-intensive sectors and automotive, which are clearly struggling. These subsidies ultimately just prolong their existence or reduce the pressure to make disruptive changes, essentially keeping them alive without addressing the core issues. This not only wastes taxpayers’ money but also has a negative impact.
Another big issue is the role of unions. While they are proud of their efforts to secure jobs, the reality is that preserving these jobs through union intervention can simply prolong the inevitable decline of companies, rather than securing jobs in the long term. I’m strongly in favor of protecting jobs and strengthening our economy, but we need to be strategic about which industries will survive and which will phase out. Some industries, like chemicals, life sciences, and med tech, have a lot of expertise and potential for the future, and we should focus on nurturing them.
Moreover, reducing the power of unions could help accelerate the transformation of these industries. It’s disheartening to see what’s happening in Germany right now, where this resistance is hindering progress.
UV: As an investor in AI-powered solutions in spaces like manufacturing, healthcare, etc., what is your view on what is the effective way to prepare for the changes AI will bring?
RR: I think we need to fully embrace AI and use it from the bottom up. Everyone in an industrial company should experiment with AI models and think about how they can make their jobs more efficient. At the same time, it needs to be driven top-down, with executives defining key use cases and implementing them either internally or through partnerships or market solutions. The drive from leadership is essential. It’s similar to when people resisted computers or the internet 20 years ago—it didn’t work out for those who didn’t adapt.
There’s no choice here. It’s not about potential—it’s about moving forward. You can either move slowly or quickly, but the gap will only grow larger if you don’t keep pace. In places like China, where people work 12-hour days, six days a week, and are well-educated and smart, it’s hard to compete if we’re working fewer hours, especially if a significant portion of that time isn’t productive due to things like home office distractions. With higher living costs, how can we stay competitive? It’s a hard reality to face.