Can we learn from past success in times of technological acceleration?
If you zoom into the history of the German chemical dye industry between 1856 and 1914, you do not just see a handful of clever chemists and lucky patents. You see a machine. It is a dense web of feedback loops in which universities, investors, engineers, trade lawyers, sales offices and even customs officials keep nudging each other forward.
Some time ago I tried to capture this machine as a “success-engine network”, using the material from Knowledge and Competitive Advantage of Peter Murmann and the methodology of Probst & Gomez’ Networked Thinking. The screenshot above is the visual trace of that exercise, built in Kumu. It tells the story of how Germany, although not the birthplace of synthetic dyes, managed to outpace England and take global leadership in a new industry that depended on chemistry, capital and coordinated learning.
At the heart of the map sits a thick black loop. This is the inner engine, the reinforcing cycle that, once running, pushed German dye firms into a steep trajectory of cumulative advantage.
The inner loop: how success feeds on itself
Imagine starting at “Scientific and Technical Education”. Nineteenth-century Germany invested heavily in polytechnics and research universities. As scientific education improved, “Workforce Skills” in chemistry and engineering rose. Companies could hire people who did not just operate equipment but actually understood organic chemistry, thermodynamics and analytical methods.
Those skilled chemists did not stay in the lab reading journals. They fuelled “Advanced Research and Development” inside the firms. Corporate labs at BASF, Bayer, Hoechst and others became institutionalised engines of discovery. R&D did two things at once: it pushed the frontier of what was chemically possible, and it allowed companies to experiment systematically with processes, yields and new applications.
From R&D you move to “Understanding customer demands”. German firms did not simply ship whatever color their chemists could produce. They talked to textile mills, printers and tanners in different countries. They learned which shades faded in sunlight, which dyes bled in washing, which pigments worked on wool versus cotton. That learning loop fed back into R&D and process design.
Better understanding of customer needs, together with strong R&D, created “Technological Superiority in Dye Production”. This was not an abstract bragging right, but a concrete edge: more stable colors, broader palettes, higher yields, more reliable batches. That superiority, in turn, translated into “Increased Global Market Share”. Textile manufacturers from Manchester to Mumbai learned that if they wanted reliable dyes, they turned to German suppliers.
Higher global market share improved profitability and free cash flow, which the map simplifies as “Investments”. With more money at hand, firms could expand production, build new plants, send agents abroad, and crucially, invest again into “Scientific and Technical Education”, for example by funding university chairs, doctoral positions and industry-academic collaboration.
It is a classic reinforcing feedback loop: once established, it makes success self-amplifying and increasingly difficult for latecomers to catch up.
The wider web: constraints, buffers and amplifiers
Around this dark inner circle the map shows a series of lighter blue and red links. These represent the conditions and frictions that made the engine either easier or harder to run.
Take “Production Costs” and “Scaling Effects of Production Lines”. As plants grew larger and more specialised, unit costs fell. The map shows how scaling effects reduce production costs, which in turn affect “Price”. Lower prices, combined with higher quality, made German dyes more attractive globally and raised adoption in new markets. But there is a trade-off: very aggressive price reductions might squeeze margins and reduce “Free Cash Flow (FCF)”, which is why some of the links in this sector are negative. Managers had to balance volume growth against financial robustness.
“Adoption costs” play a similar moderating role. For a textile mill, switching to new dyes involved learning, possible redesign of processes and some risk. As German firms improved the products, they reduced adoption costs and accelerated the spread of their products.
Legal and political structures appear in the lower arc of the network. “Patent Laws” protect the returns from R&D, making investments in research more attractive. “Effectiveness of Trade Policies” shapes how easily firms can export, set up subsidiaries or protect themselves from hostile tariffs. In the German case, relatively strong patent protection and a national interest in export-oriented industry formed a supportive backdrop.
On the right side of the map, the human and relational side of the system becomes visible. An “Academic Industrial Network” links universities, firms and research institutes. Knowledge flows in both directions: academic discoveries inform industrial R&D, while practical problems from factories inspire new fundamental research. Above that sits “Entrepreneurial Mindset & Culture”, the shared belief that building new businesses and technologies is both possible and desirable. Without such a mindset, the same educational system might have produced civil servants rather than industrial pioneers.
Closer to the customer we find “Local Offices” and “Proactive shaping of trade agreements”. Local commercial and technical offices in foreign markets lowered informational distance. They understood local conditions, maintained relationships with buyers and governments, and fed insights back into the core loop of customer understanding and R&D. Meanwhile, companies and the state worked together to shape trade agreements that opened markets or at least prevented disadvantageous barriers.
When you follow all these arrows, what emerges is not a single cause of success, but an ecosystem. The German dye industry did not win because of one heroic inventor. It won because education, culture, finance, law, research, sales and diplomacy interacted in a way that kept amplifying technological lead and market share.
From synthetic dyes to synthetic intelligence
Fast-forward to today and change the object of analysis. Instead of coal-tar dyes, we are talking about large language models, foundation models and generative AI platforms. The contours of the map look strangely familiar.
In AI, the current “increased global market share” belongs largely to US-based firms: the dominant cloud providers, major AI labs and platform companies. They benefit from their own reinforcing loop: the more users and enterprise customers they have, the more data and revenue they accumulate; the more revenue, the more they invest in GPUs, research talent and infrastructure; the better the models, the more users they attract.
Where does Europe stand in this picture, and what would it take to build its own success engine for AI instead of merely being a customer of American platforms?
If we reuse the historical map as a template, Europe would need a robust loop between advanced technical education, a deep pool of AI talent, strong research organisations, customer-oriented product development, leading-edge infrastructure and globally competitive firms. It would also need the surrounding support structure: smart patent and data regimes, effective trade and competition policies, an entrepreneurial culture that accepts failure, and a dense web of local and global relationships.
Some pieces already exist. European universities produce excellent research in machine learning. Open-source initiatives and new startups such as Mistral show that there is both technical skill and entrepreneurial ambition. Programmes like EuroHPC aim to build shared compute infrastructure. The EU’s AI Act, with all its imperfections, is at least an attempt to shape the rules instead of being a passive taker.
Yet when you look at the full network, gaps become obvious. Venture capital and growth financing are weaker than in the US. Many promising startups end up being acquired by American or Chinese players early in their life cycle, effectively exporting the upper layers of the success engine. The market itself is fragmented along language and regulatory borders, making it harder to scale B2C platforms quickly. Risk-aversion in corporate and public procurement slows down adoption, which in turn slows down the feedback loop of learning from customers and improving products.
On top of that sits a cultural pattern: Europe has strong traditions of critical reflection, precaution and regulation. These are strengths when it comes to safety and ethics, but they can easily become brakes if they are not matched by equally strong ambitions in building world-class firms.
Digital sovereignty: realistic ambition or comforting fantasy?
Can Europe realistically hope to regain digital sovereignty and overcome US dominance in AI? The honest answer is: not in the sense of replicating Silicon Valley one-to-one. The initial conditions are simply different. The US enjoys a uniquely large unified market, deep capital pools and an established concentration of tech giants who can cross-subsidise massive AI investments from their existing business lines.
However, if we interpret “sovereignty” more modestly as the ability to shape and operate critical digital infrastructure on our own terms, the story changes. Europe does not need to dominate all layers of AI to be sovereign, but it cannot be purely dependent either.
A realistic path probably looks less like a single giant flywheel and more like several interlinked success engines in specific domains. Europe’s industrial strength is (for the moment) in manufacturing, automotive, energy systems, healthcare and public infrastructure. If it manages to build AI stacks that are deeply integrated into these sectors – combining local data, domain expertise and specialized models – it can create niches of global leadership that matter economically and strategically.
In network terms, that means deliberately tightening the loops between domain-specific customer understanding, applied AI research, specialised infrastructure and regulatory clarity. For example, a European healthcare-AI engine would connect hospitals, universities, regulators and startups in a way that makes it easier to innovate responsibly on sensitive patient data, perhaps through federated learning or privacy-preserving techniques. The same logic could apply in manufacturing or climate tech.
Regulation must then be designed as part of the success network, not just as an external constraint. The AI Act and related frameworks need to create predictable rules that reduce adoption costs for businesses rather than increase them. If compliance becomes too complex, the feedback loop turns negative: firms will either avoid AI adoption or simply buy pre-packaged solutions from US providers, who spread the compliance costs over a much larger customer base.
Capital is the other hard constraint. Without substantial and patient funding, the inner loop of talent, R&D and market share cannot spin fast enough. That suggests that Europe may need new instruments: public-private funds with long time horizons, strategic procurement by governments that intentionally favour European AI solutions when feasible, and perhaps loosening some state-aid rules in targeted ways. Otherwise, ambitious projects will keep migrating to where money is more abundant and risk-tolerant.
Learning from the dyes
The success of the German dye industry was not inevitable. It was constructed step by step through overlapping feedback loops: education, research, patents, trade, culture and investment reinforcing one another. England did have the first mover advantage, but it lacked that tightly coupled network.
For AI, Europe currently looks a bit like England did back then: scientifically important, but industrially overshadowed. The lesson from the nineteenth century is not that it is hopeless, but that isolated strengths are not enough. They have to be wired together into a success engine.
That wiring is a political and cultural choice. If Europe decides that it wants to be more than just a regulatory power and consumer market, it will have to treat AI infrastructure, talent and entrepreneurship with the same strategic seriousness that previous generations reserved for coal, steel and railways. That means accepting some discomfort: faster experimentation, more tolerance for failure, and the willingness to bet billions on technologies where the outcome is uncertain.
Will this completely overturn US dominance? Probably not. But it can move Europe from dependency to interdependence, from being a peripheral customer to being a co-author of the next technological chapter. The map of the dye industry reminds us that dominance is never simply given by invention alone. It is produced by networks of actors who decide – Europe must find its own success-engine niches and achieve leadership in relevant sectors.