In this Rise of AI Virtual Chat from 2020, host Fabian Westerheide interviews Rasmus Rothe and Adrian Locher, founders of the AI venture studio Merantix AG, about how to make money with AI. They explain Merantix's model of building and funding AI companies over a ten-plus year horizon, validating business model, technology and team before startups raise external Series A funding, and describe screening entrepreneurs through interviews and a two-day on-site challenge, talking to about 300 candidates to find a handful of entrepreneurs-in-residence. They discuss where current AI opportunities lie, from mature computer vision to an emerging wave in natural language processing and longer-term bets like biology and neuroscience, and how AI creates business value through both automation and improved accuracy over human decision-making. They describe Merantix's own portfolio, including a services arm, a certified medical imaging product for breast cancer screening, and a sensor-data platform for automated driving. The conversation closes with a debate on what Europe needs to compete with the US and China, and audience questions on AI and climate change, B2C strategy, cloud choice and how they define AI.
Key takeaways
- Merantix runs a venture studio with a ten to fifteen year horizon, validating business model, technology and team before a company raises a Series A.
- Merantix talks to about 300 candidates a year to find a handful of entrepreneurs-in-residence, screened partly through a two-day on-site challenge.
- They see computer vision as largely mature, natural language processing entering its own breakthrough moment, and biology as a fast-growing frontier.
- AI creates business value both by automating complex decisions and by making fewer mistakes than humans in inconsistent conditions.
- They describe Europe as regulating new technology before exploring it, unlike the US's tolerance for failure or China's roughly 130 billion dollar AI push.
Detailed notes
Merantix's venture studio model
Rothe and Locher describe Merantix as a venture studio that sources, validates, funds and incubates AI companies from a 25 million euro fund, launching two to four companies a year. They stress a long investment horizon, ten to fifteen years, because investors care less about near-term revenue than about companies having real impact over a decade, especially in regulated fields like medical products that can take a year or two just to reach market, unlike a typical app. A company is only allowed to graduate from the studio and raise a Series A once its business model, technology and team have each been validated, and the studio provides shared research, best practices and access to industry partners for early proof-of-concept work rather than just capital.
Screening and validating entrepreneurs
Asked what they look for in founders, they say persistence matters most, since AI problems can take five to ten years to pay off, alongside leadership, charisma and enough technical depth even for a business-side CEO to hold their own with regulators, technical co-founders and later-stage investors. Merantix talks to roughly 300 people a year and currently works with three entrepreneurs-in-residence, using a two-day on-site challenge where candidates work on a potential venture case so the team can assess culture fit for a multi-year partnership.
Where the opportunities are
They note that most machine learning ideas they encounter are better suited to incumbents applying the technology internally than to new ventures, since a good venture case needs a credible path to a billion-dollar outcome. They see computer vision as largely mature for tasks like image classification and object detection given enough training data, natural language processing as entering its own 'ImageNet moment' and opening up document processing, customer service automation, translation and content generation, and biology, aided by falling DNA sequencing costs, as a fast-growing frontier for protein and chemical optimization. Longer-term, more speculative bets they mention include quantum physics and neuroscience, alongside more predictable automation opportunities like e-commerce.
How AI creates business value
They frame AI's business value in two layers: automation of complex decisions previously requiring human experts, and improved accuracy, since a well-trained machine makes fewer mistakes than humans affected by inconsistent conditions. Selling AI to enterprises has become easier over the past two to three years as expectations have matured, but pilots remain uncertain research projects rather than fixed-scope deliverables; a typical sales pipeline runs through workshops, then a proof-of-concept, then a production system, and can take up to two years to reach a recurring-revenue contract. They estimate that three years earlier, about 80% of companies approached them wanting to 'do something with AI', whereas now about 80% already have a concrete application in mind.
Merantix's own portfolio companies
Beyond the venture studio, Merantix runs a services arm that helps corporates and mid-sized companies build machine learning solutions, and owns product companies including a certified medical imaging platform used by doctors for breast cancer detection and screening across Europe, and a data platform that ingests, searches and filters the petabytes of camera and lidar sensor data produced by automated-driving vehicles.
Q&A highlights
On Europe's competitiveness against the US and China, Locher argues Europe defaults to regulating new technology before exploring it, while the US tolerates failure and China treats AI as a national priority with roughly 130 billion dollars invested; both call for more visible lighthouse companies, more entrepreneurs leaving safe jobs to found startups, and more later-stage venture capital willing to fund European deep tech at scale. Asked how AI could help with grand challenges, Rothe points to a long paper by leading AI researchers, including Yoshua Bengio, cataloguing dozens of ways AI could help tackle climate change. Asked why Merantix has no B2C product, Locher says building consumer products is more about marketing and execution than technology and simply is not their expertise. On cloud infrastructure, they recommend major providers like Amazon, Google and Microsoft over on-premise once a company scales, though a European provider can sometimes ease sales in data-sensitive industries. Asked to define AI, Rothe says internally they mostly say machine learning and reserve 'AI' for customer-facing language, while Locher ties the term to decisions complex enough that they go beyond narrow pattern recognition. Asked about the strongest natural language technology company, they name DeepL as a German example that consistently beats Google Translate. They close by saying their ten-year ambition is to build more than fifty AI-driven companies.
Questions answered in this talk
How does the Merantix venture studio decide when a startup is ready to raise external funding?
According to Rasmus Rothe and Adrian Locher at Rise of AI's 2020 virtual chat, a company graduates from the Merantix venture studio and raises a Series A only after its business model, technology and team have each been validated, following a ten to fifteen year investment horizon.
Where do the biggest current AI business opportunities lie?
According to Rasmus Rothe at Rise of AI's 2020 virtual chat, computer vision is largely mature, natural language processing is entering its own breakthrough moment similar to computer vision's ImageNet moment, and biology, aided by falling DNA sequencing costs, is a fast-growing new frontier.
What does Europe need to compete with the US and China in AI?
According to Adrian Locher at Rise of AI's 2020 virtual chat, Europe needs to stop regulating new technology before exploring it, build more visible lighthouse AI companies, encourage more entrepreneurs to start companies, and attract more later-stage venture capital for deep tech.