While billions continue to pour into ever larger foundation models, Berlin-based Merantix Capital is pursuing a different proposition. Its portfolio is built around embedding artificial intelligence deep inside established industries — from healthcare and manufacturing to logistics, energy and cybersecurity. The strategy points to an area where Europe may still possess a structural advantage over the United States. Yet Merantix’s previous exits also raise a more uncomfortable question: is Europe creating the technology champions of tomorrow, or simply increasingly attractive acquisition targets for today’s global corporations?

At first glance, the Merantix Capital portfolio appears remarkably disparate. One company develops novel biomaterials, another analyses factory production lines. Artificial intelligence is being used to detect breast cancer, investigate immune-mediated diseases, optimise logistics decisions, identify job candidates, automate energy systems and make employees more resilient against cyber attacks. Legal technology, enterprise data, ERP systems and mobile AI agents add further layers to a portfolio whose companies sometimes have little in common on the surface.

Behind that diversity, however, lies a relatively precise investment thesis. Merantix is not primarily betting on artificial intelligence as an industry in its own right. Instead, it assumes that AI will become a technological layer embedded across existing sectors. The relevant question is therefore not simply who can build the most powerful model, but which business processes can be reorganised once software moves beyond storing information and digitising workflows to analysing situations, prioritising actions and increasingly preparing decisions. While vast amounts of capital are being directed towards data centres, semiconductor infrastructure and foundation models, Merantix is looking for economic leverage one level further up the value chain: where AI meets the processes of the real economy.

Different industries, one investment logic

The breadth of the portfolio reflects this approach. Cambrium develops new biomaterials, Graph Therapeutics focuses on AI-enabled precision medicine, Vara applies artificial intelligence to breast cancer screening and Ovom Care operates in reproductive medicine. Deltia brings computer vision into industrial manufacturing, Arqh develops decision-making technology for logistics, Meteoric targets greater automation of energy supply, while Revel8 addresses the human layer of cyber defence. Other ventures span recruitment, e-commerce, enterprise data, ERP systems and emerging forms of mobile AI agents.

What connects these businesses is therefore less the sector in which they operate than the economic structure of the markets they address. Merantix tends to favour industries characterised by complex processes, large volumes of data and costly human decisions. Such markets become particularly attractive when regulation, specialist expertise or proprietary datasets create additional barriers to entry. Healthcare, life sciences, industrial production, finance, logistics and security all display several of these characteristics.

This represents a counter-thesis to one of the dominant assumptions surrounding the current AI boom: that the largest share of economic value will inevitably remain with the providers of the underlying models. Models may become increasingly interchangeable, computing capacity can be purchased and access to powerful AI systems is already available to almost any software company through standard interfaces. What remains much harder to reproduce is access to clinical or industrial data, integration into physical production environments, regulatory approval, deeply embedded workflows and customer relationships developed over many years. It is in precisely these areas that Merantix is attempting to establish a defensible position for its portfolio companies.

Building companies before there is a company

Another distinguishing feature of the model is the stage at which Merantix becomes involved. A significant proportion of its ventures originated in what the company calls the pre-idea phase — before a founding team has developed into a conventional start-up with a clearly formulated product and business model. Other investments began at pre-seed stage.

Merantix has consequently operated less like a conventional venture capital fund and more like a hybrid of investor, company builder and venture studio. Markets are analysed, founding teams assembled, technologies evaluated and initial applications tested with potential customers before a company reaches the maturity typically expected by institutional investors. The model shifts both risk and influence towards the investor. Entering before a business has been fully formed provides access at valuations that bear little resemblance to those of later financing rounds, but it also means assuming risks that traditional venture capital firms normally leave largely to founders. Merantix must not only select the right company; it first has to determine which problem is capable of becoming a commercially attractive market.

The new fund broadens that approach considerably. In June 2026, Merantix Capital closed a €103 million fund, around half of which is intended for companies developed through its established pre-idea model. The other half is, for the first time, systematically allocated to externally founded European start-ups at pre-seed and seed stage. Approximately 40 investments are planned, with ticket sizes ranging from around €500,000 to €4 million. The first fund, by comparison, amounted to approximately €30 million and focused largely on businesses originating from Merantix’s own venture-building ecosystem.

The significance of the new fund therefore extends beyond its size. Merantix is gradually moving from a start-up factory backed by its own investment capital towards a broader institutional early-stage investor in European artificial intelligence.

Capital is not necessarily the scarce resource

The composition of the investor base is equally revealing. Limited partners include institutional investors, corporates and foundations, among them Union Investment, Jungheinrich, KPMG Germany, the Robert Wood Johnson Foundation and the W.K. Kellogg Foundation. Jungheinrich committed €5 million while simultaneously entering into a strategic AI partnership.

That combination points to an increasingly important reality for young B2B AI businesses: capital is no longer necessarily the only, or even the most serious, bottleneck. Building a technically impressive prototype has become dramatically easier. Obtaining access to a production line, hospital, logistics network or critical infrastructure environment — and then proving that the technology works reliably under operational conditions — remains considerably more difficult.

Merantix attempts to institutionalise precisely this kind of access. Its Berlin AI Campus brings together dozens of companies and AI organisations, while Merantix Momentum works with established businesses on practical AI projects. This creates an ecosystem in which investors may simultaneously become pilot customers, development partners or door-openers for portfolio companies. For an early-stage investor, such a network can ultimately represent a more durable competitive advantage than simply having a larger pool of capital available.

The exits reveal what strategic buyers value

The companies that Merantix has already sold offer some of the clearest evidence of how the investment thesis works in practice. SiaSearch developed infrastructure for structuring large quantities of sensor and image data for autonomous driving before being acquired by US company Scale AI in 2021. Kausa, specialising in AI-supported causal analysis of enterprise data, transferred its technology and team to SAP in 2023, where its capabilities were used to strengthen analytical functions around SAP Signavio.

The most financially visible exit to date followed in 2025, when Wolters Kluwer acquired Berlin-based legal AI company Libra Technology for a purchase price of up to €90 million. An initial €30 million was agreed, with a further amount of up to €60 million linked to future business performance. What made the transaction particularly striking was the speed of the company’s development. Libra had been founded only in 2023, launched its AI assistant at the end of 2024 and, according to the buyer, was expected to reach approximately €5 million in annual recurring revenue by the end of 2025.

Based on that forecast, the initial purchase price alone represented roughly six times expected ARR, while the maximum consideration could theoretically have reached 18 times that figure. Such a comparison is necessarily limited because a substantial part of the consideration was performance-related, but it nonetheless illustrates how much strategic value established information and software groups are beginning to attach to highly specialised AI applications.

Even more revealing than the valuations is what the transactions have in common. Scale AI did not acquire another foundation model when it bought SiaSearch; it acquired infrastructure for complex data. SAP obtained technology that could be incorporated directly into established enterprise workflows. Wolters Kluwer acquired an application combining generative AI with highly specialised legal processes. In each case, the strategic value was situated closer to the customer process than to the underlying AI model. That may represent the clearest validation so far of the Merantix investment thesis.

Europe may not need to win the race for the largest model

There is also a distinctly European dimension to this strategy. European companies enter the competition for the world’s largest models, most expensive computing infrastructure and dominant consumer platforms from a difficult starting position. The sums that US technology groups can mobilise for data centres, semiconductors and model development are difficult to replicate within Europe.

Europe does, however, possess a different set of assets: highly developed industrial companies, manufacturing facilities, hospitals, research organisations, pharmaceutical businesses, logistics networks, energy infrastructure and a large number of heavily regulated markets. The very complexity that is frequently regarded as a disadvantage of the European economy could therefore become an asset in the age of vertical AI.

The more complex a process, the stricter the regulatory environment and the more valuable the underlying data, the less sufficient it becomes to place a user interface on top of a generally available language model. Sustainable differentiation must then emerge elsewhere — through integration into processes, proprietary datasets, regulatory expertise and domain-specific decision logic.

Several companies in the Merantix environment illustrate this logic. Vara has integrated artificial intelligence into established breast cancer screening processes. Deltia uses computer vision and AI to analyse manual manufacturing operations. Graph Therapeutics combines artificial intelligence with precision medicine, while Outpost Bio develops models around the human microbiome. External financing rounds secured by several companies originating from the Merantix ecosystem do not yet prove that they will become large and profitable businesses, but they indicate that the combination of vertical specialisation, data access and early industrial partnerships is also considered attractive by investors outside the Merantix network.

The uncomfortable question of Europe’s AI champions

Successful exits nevertheless create a fundamental tension. Merantix has articulated the ambition of helping to build global AI champions in Europe. Its exits, however, demonstrate for the moment something slightly different: European specialist AI companies are proving highly attractive acquisition targets for larger technology and information groups.

For founders and investors, a transaction involving SAP or Wolters Kluwer is clearly a success. From an industrial-policy perspective, however, the calculation is more complicated. Europe needs successful exits because they recycle capital and entrepreneurial expertise into the technology ecosystem. A functioning innovation market cannot depend on every promising start-up remaining independent until an IPO. At the same time, Europe will struggle to create a new generation of major software companies if its most promising businesses are repeatedly absorbed into established groups before reaching global scale.

The sale of SiaSearch to US-based Scale AI makes this tension particularly visible because it touches one of the central questions of European technology policy: is it sufficient for Europe to create innovative technology companies, or must the continent also develop the capital markets required to finance them through much longer periods of independent growth?

The €103 million Merantix fund can be interpreted as one part of an answer. Within the European early-stage market, it represents meaningful capital. Measured against the scale of the international AI investment market, however, the amount remains comparatively modest. Once a company moves into later growth phases, building a globally competitive software business can require financing rounds several times larger.

Not every AI business has a defensible moat

There is a second structural risk. Technical barriers to entry are falling rapidly in some areas of generative AI. Functions that only a few years ago required specialist development teams and proprietary models are increasingly becoming standard capabilities offered by major model providers. For investors, this creates a permanent race against commoditisation: a product that can be sold today as standalone software may become part of tomorrow’s standard release from Microsoft, Google, OpenAI, Anthropic or an established sector-specific software provider.

Merantix appears to be responding by concentrating increasingly on markets in which additional barriers to entry remain. Its areas of interest include tech-bio and healthcare automation, industrial AI, new ERP architectures, cybersecurity, defence, energy, digital trust systems and highly regulated security applications. The portfolio therefore looks less like an attempt to accumulate as many AI start-ups as possible and more like an effort to embed artificial intelligence deeply within the infrastructure of the real economy.

The more closely a technology becomes intertwined with proprietary data, regulation, business processes and physical operations, the less likely it is to become obsolete overnight because of the next model release. The durable moat, in other words, may not be the algorithm itself. It may be the integration.

Security is becoming an investment field

This development is particularly relevant to the security industry. Merantix already has Revel8 in its portfolio, a company using artificial intelligence to address the human dimension of cyber defence. Cybersecurity, AI-supported threat detection and defence and security are also among the fields in which the investor sees further potential.

The wider Merantix group’s recently announced investment, together with IBEX Wachstumspartner, in German video-security specialist Geutebrück fits this pattern, even though the company is not part of Merantix Capital’s conventional start-up portfolio. Geutebrück begins from a fundamentally different position than an early-stage AI venture: it already possesses decades of customer relationships, established video-management technology and experience in security-critical environments.

Artificial intelligence therefore does not first have to create a new market. Instead, it can potentially increase the value of an existing data infrastructure and expand the use of video beyond conventional security applications. Economically, this could represent an important next step in the evolution of the Merantix model. Until now, the group has primarily been associated with building young companies around new AI applications. An established medium-sized technology company already has customers, products, data structures and international sales channels. The question is therefore no longer how artificial intelligence can become the basis for a new company, but how it can transform the value creation of an existing one.

If this approach proves successful, the Merantix ecosystem could develop a third field alongside venture building and early-stage investing: the AI transformation of established European technology companies.

From AI incubator to industrial investor

Ten years after its creation, Merantix itself is undergoing a transformation. What began as a Berlin venture studio has developed into a hybrid investment model that builds companies from scratch, invests in externally founded start-ups, connects young technology businesses with large industrial groups and, through the broader Merantix organisation, is increasingly becoming involved with established companies.

Its portfolio is therefore more interesting as an expression of a particular theory of capital allocation than as a collection of individual company names. Merantix is not betting on financing Europe’s largest foundation model. Its more consequential bet is that a substantial share of artificial intelligence’s economic value will emerge only when those models become deeply integrated into specific industrial processes.

There are good reasons to take that thesis seriously. Europe’s greatest strengths do not lie in hyperscale cloud infrastructure or globally dominant consumer platforms. They lie in companies that build machines, develop medicines, distribute energy, move goods, manage risk and operate critical infrastructure. All of these sectors possess data, processes and specialist knowledge that cannot easily be replicated elsewhere.

The decisive question is therefore no longer whether these industries will use artificial intelligence. They already do. The more consequential question is who will control the software layer through which their data is analysed, their processes are optimised and, increasingly, their decisions are prepared.

That is the layer for which Merantix Capital is competing. And ultimately, this is also where its strategy will have to prove itself: whether the portfolio produces more than a succession of notable exits and instead contributes to creating the independent European AI companies that the continent has discussed for years, but has so far produced only in limited numbers. 

[ML]