At Rise of AI 2025 in Berlin, Adrian Locher, Founder and General Partner of Merantix Capital, argues that early-stage AI investing succeeds through platform support rather than capital alone. He describes Merantix as a mix of venture studio and VC fund, backed by a EUR100 million fund, a roughly 1,500-person AI community on its Berlin campus, a 100-person consulting arm called Merantix Momentum, and the AI House event at the World Economic Forum in Davos. The firm invests in the application layer rather than infrastructure or tooling, and looks for products deeply embedded in client workflows, which Locher sees as more defensible than thin wrappers around existing AI models. He shares learnings from ten years of investing, including how Merantix uses its trust with European corporates to get early-stage startups signed as design partners before a product even exists, and describes five recurring founder personas in its portfolio. He closes by urging Europe to act and accept failure rather than keep discussing plans.
Key takeaways
- Locher argues that Merantix's edge is its platform (campus, consulting, AI House) rather than its capital alone.
- He says most AI wrapper products are not venture-backable because a venture case needs a 100x to 1,000x return.
- Merantix's manufacturing startup Delta signed four paying customers, including ABB, on a PowerPoint before building a product.
- Four of five companies in the studio's latest batch reached 500k ARR within six months, and three then raised seed rounds within ten months.
- Merantix has added KPMG and Jungheinrich as corporate limited partners in its fund.
Detailed notes
Merantix as a platform, not just a fund
Locher describes Merantix as a mix of venture studio and classical VC, built over roughly ten years, now running a EUR100 million fund with a team of fewer than ten people. Around it sits a wider platform: a Berlin campus with about 1,500 people working in AI and machine learning across roughly 100 teams (with a second campus being built in London), a consulting arm called Merantix Momentum with about 100 people serving corporates and governments, co-founding of the German AI association Ki-Bundesverband, and the AI House event, which ran for the second time at the World Economic Forum in Davos in January, hosting about 50 events over five days with roughly 3,000 attendees.
Where Merantix invests
Merantix focuses on the application layer of AI rather than infrastructure, foundation models or tooling, looking for companies that grow vertically deep along a value chain instead of staying point solutions. Locher stresses building multi-disciplinary teams that would not normally meet, such as chemists, biologists and machine-learning engineers on the same team, and says Merantix is deliberately European, aiming to build European companies that can become global champions.
Learning one: complexity is the moat
Locher argues most AI wrapper products will not survive, since building a working product is now cheap; he describes using the research tool Manus and the builder Lovable to produce a working SaaS-style tool in twenty minutes for about $50 in credits. Because a venture case needs a 100x to 1,000x return, he says most such companies are not venture-backable. He cites Inflection, which raised $1.2 billion before being absorbed into Microsoft, as an example of a well-funded company that did not deliver venture-scale returns. The business models Merantix favors are deeply integrated into client workflows and hold client data, comparable to Salesforce, so they cannot easily be ripped out.
Learning two: bridging trust with European corporates
Locher says European startups struggle to sell into corporates the way US startups can, so Merantix uses its campus, Momentum and AI House relationships to build corporate trust that it then extends to its portfolio companies. He cites Delta, a manufacturing analytics and robotics company in the portfolio now working with about 20 large industrial players, which had four paying customers, including ABB and Eastman, signed on the basis of a PowerPoint before any product existed.
Founder personas and studio results
Locher describes founders who join purely to make money as lacking the stamina for the long road to success, and outlines five recurring personas in the portfolio: the technical academic, the corporate escapee, the serial entrepreneur (Felicia, who built the fertility company AOM and recently opened its first clinic in Portugal), the 'mafia entrepreneur' (Julius, building a cybersecurity company after early experience at a company that scaled past a billion in revenue), and the solo AI developer (Sylvio, formerly leading Volkswagen's autonomous-driving research, now building Delta with a paired commercial co-founder). He reports that four of five companies in the studio's latest batch reached 500k ARR within six months, and three of those raised seed rounds within ten months of their pre-seed.
Q&A highlights
Asked by Dell's AI CTO how to collaborate, Locher points to tiers ranging from campus partnerships, which hosted about 300 events last year, to a corporate LP program that has recently added KPMG and the forklift manufacturer Jungheinrich, a EUR6 billion revenue company, as fund investors. Asked how many founder personas a startup needs, he says Merantix will back incomplete or solo founding teams and figures out needed profiles after validating go-to-market. Asked how talent is sourced, he points to the campus as the main magnet plus second-degree network referrals. Closing on Europe's competitiveness, Locher says the region needs to stop discussing and start acting, and must be willing to fail often and fail fast.
Quotes
Early stage is not about capital. It's about a lot of other things. Yes, it's also about capital, but capital is just to keep the lights on. The rest matters much, much more.
Adrian Locher
Founders who come in to make money, they usually don't have the stamina, they usually don't have the energy to pull it through over this long period of time.
Adrian Locher
Questions answered in this talk
Why does Merantix Capital focus on the application layer of AI rather than foundation models?
According to Adrian Locher at Rise of AI 2025, Merantix believes the strongest business models come from applying AI vertically deep into a value chain and embedding it into client workflows, rather than competing in infrastructure or foundation models, which he leaves to larger, better-capitalized players.
How did the startup Delta get its first customers before building a product?
Locher explains that Delta, a manufacturing analytics and robotics company in the Merantix portfolio, signed four paying customers such as ABB and Eastman on the strength of a PowerPoint presentation at the company's founding, before any working product existed.
What return does a startup need to be a good venture capital investment, according to Merantix?
Adrian Locher says Merantix needs to see a clear path to a company being 100 to 1,000 times as valuable to justify venture funding, which is why he considers most AI wrapper products not venture-backable even if they are profitable businesses.