Startups and Knowledge

We are pleased to announce the publication of the book “Startups and Knowledge: How access to knowledge affects the startup performance of European cities in the age of rapid technological change.” This volume brings together studies written by our staff and students, offering valuable insights into the relationship between knowledge ecosystems, innovation, and urban startup success. The book reflects the ongoing research and intellectual work of the Centre for Next Technological Futures at MCC till 2024.

The complete book can be purchased on the MCC Press website: Startups and Knowledge.

COVER TEXT

“In our time, venture-backed startups are generating radically new technological and business ideas; they are the ones who are discovering the future. And as with any discovery process based on trial and error, the vast majority of startups fail within a relatively short period of time. But those who survive and push the boundaries of the unknown unknowns usually do so because they possess a special kind of knowledge that their competitors lack. 

Therefore, the four papers of this new volume of Studies on Innovation, Technologies and Regions from the MCC Centre for Next Technological Futures not only summarise its research activities in the academic years 2024 and 2025, but they also provide analyses of the relationship between startups and knowledge in European cities from different perspectives and on various topics. Be it the successful startups that use artificial intelligence, the startups that not only survived the period from 2021 to 2024 but were actually able to grow, the countries that apply clean and green energy solutions, or the European cities that can rely on universities, which are a breeding ground for startup founders—they all demonstrate the increasing importance of knowledge. This is why this volume is titled Startups and Knowledge.”

FOREWORD

Since Friedrich August von Hayek, we have known that behind every economic success and advantage lies a special kind of knowledge—knowledge that competitors do not have. Therefore, while it may sound very strange, it is worth knowing only the things that others do not know. This also applies to competition between national economies, and it is all the more difficult to accept as we usually tend to point to natural resources, geographical location, historical luck, or some political factor. However, more than half a century ago, the Nobel-Prize-winning economist Robert Solow famously concluded that labour productivity in the United States doubled and GDP tripled between 1909 and 1949, but only 12.5% of this increase was due to higher capital and labour inputs; 87.5% of the increase was due to technological change. In other words, America became rich and prosperous because of American innovation and know-how.

But the question remains as to what kind of knowledge can bring these returns. As a kind of response, Donald Rumsfeld, the former US Secretary of Defense, made a puzzling distinction between the known knowns, the things we know we know; the known unknowns, the things we know we don’t know; and the unknown unknowns, the things we don’t even know we don’t know. Applying this distinction, innovation is the world of unknown unknown, the terrain of novelties that we did not even know that we did not know. For example, there are currently over nine million applications waiting to be downloaded onto our smartphones. However, before 2008, when Apple released its first 500 applications, we wouldn’t have even considered becoming application developers or application users. And as this example shows, innovation is the unknown unknown that meets consumers’ needs—in many cases, needs they didn’t know they had. Without consumers who accept it, even the unknown unknown is only a novelty, not an innovation.

And that’s precisely why it’s very difficult, if not impossible, to predict the future, especially in the technology industry. We don’t know who will develop a new technology or business idea that people will use and even pay for when it is turned into a product or service, nor when this will happen. We do not know in advance, even if we do not always find something radically big and drastically new behind most innovations. As the “father” of innovation theory Joseph Schumpeter recognised, innovation is, in many cases, the “new combination” of things that already exist. On the one hand, the emphasis is on the word “new” because the real novelty is the new method of combining things, which did not exist before. On the other hand, the word “combination” points to the future, as we will be able to combine more and more technological and business ideas—as building blocks—over time.

The future of technology and innovation is hard to predict, but as Sebastian Mallaby writes in his latest book, it “can be discovered by means of iterative, venture-backed experiments.” In our time, venture-backed startups are generating radically new technological and business ideas; they are the ones who are discovering the future. And as with any discovery process based on trial and error, the vast majority of startups fail within a relatively short period of time. But those who survive and push the boundaries of the unknown unknowns usually do so because they possess a special kind of knowledge that their competitors lack.

Therefore, the four papers of this new volume of Studies on Innovation, Technologies and Regions from the MCC Centre for Next Technological Futures not only summarise its research activities in the academic years 2024 and 2025, but they also provide analyses of the relationship between startups and knowledge in European cities from different perspectives and on various topics. Be it the successful startups that use artificial intelligence, the startups that not only survived the period from 2021 to 2024 but were actually able to grow, the countries that apply clean and green energy solutions, or the European cities that can rely on universities, which are a breeding ground for startup founders—they all demonstrate the increasing importance of knowledge. This is why this volume is titled “Startups and Knowledge”.

With AI on everyone’s lips, vaunted as the next general-purpose technology of the 21st century, the paper by Zoltán Cséfalvay, György Papp, and Zalán Horlik—“Competitiveness and Knowledge: Artificial Intelligence Scaleups in European Cities”—focuses on AI scaleups in Europe (startups that raised more than €1 million in funding) and the knowledge base they can rely on, as measured by patent activity. After discussing the geopolitical race for supremacy in AI, which is widely perceived as limited to just two giants, namely the USA and China, and analysing scaleups with AI technology worldwide, the authors highlight several other key players, including Israel, Singapore, Canada, India, the United Kingdom, and Switzerland. Nevertheless, the European Union is still lagging behind in this competition, and therefore, questions arise as to how the European scaleup city landscape is shaped by scaleups that use AI technology, as well as how the knowledge background of AI scaleups affects the competition between European scaleup cities.

To place the AI scaleups within the context of the European scaleup city landscape, Cséfalvay, Papp, and Horlik first retrieved data on more than 26,600 scaleups from the Dealroom database in 2024, which raised a total of €533 billion in funding, based in the European Union as well as the United Kingdom, Switzerland, Norway, and Iceland (hereinafter referred to as Europe). In the second step, they assigned these scaleups to their respective functional urban areas (FUAs) according to the location of their headquarters—these areas are defined by the EU-OECD classification and include the core city and its commuter zone. Overall, there were 123 cities (FUAs) with significant performance in terms of scaleups based on three measures, namely the number of scaleups in the city, the total funding raised by scaleups, and the number of scaleups with a market value of more than €200 million. In the third step, and based on these variables, they applied the K-means algorithm to cluster the 123 FUAs. The result was six clusters: Global scaleup cities (London, Paris, Berlin, and Stockholm), Top European scaleup cities (Amsterdam, Dublin, Munich, and Zurich), Strong European scaleup cities (Barcelona, Cambridge, Copenhagen, Helsinki, Madrid, and Oslo), and Top Emerging scaleup cities (Brussels, Edinburgh, Hamburg, Lausanne, Lyon, Manchester, Milan, Oxford, Tallinn, and Vienna), Emerging scaleup cities with 32 FUAs, and Regional scaleup cities with 67 FUAs.

Regarding scaleups with AI technology, the authors also retrieved data on over 3,200 scaleups from Dealroom in 2024 and assigned them to 123 FUAs and 6 clusters. Although these scaleups have raised almost €73 billion in funding, only a tiny fraction, about 5% of them, have achieved a market value of over €200 million. The role of inventions and knowledge beyond AI scaleups is illustrated by the fact that they have registered around 5,500 patents in total, an average of 1.7 patents per AI scaleup. However, there are fewer than 700 scaleups, meaning that only one in five AI scaleups has filed at least one patent, while only nine scaleups account for about a third of all patents filed by AI scaleups. 

Cséfalvay, Papp, and Horlik’s detailed analysis first shows that AI scaleups and their funding are extremely concentrated in literally just a handful of cities. The four Global scaleup cities (London, Paris, Berlin, and Stockholm) and the four Top European scaleup cities (Amsterdam, Dublin, Munich, and Zurich) together account for over 50% of AI scaleups, scaleup funding, scaleups with a valuation of over €200 million, and patent applications. In contrast, more than half of the scaleup cities in Europe, i.e., the 62 regional scaleup cities, are responsible for less than 10% of AI scaleups, scaleup funding, scaleups valued at over €200 million, and patents registered.

Second, this unequal distribution is embedded in Europe’s East–West divide, as even the major metropolises in Central and Eastern Europe lag far behind in this respect. The ten scaleup cities in Central and Eastern Europe—Bucharest, Budapest, Bratislava, Krakow, Ljubljana, Prague, Sofia, Warsaw, Wrocław, and Zagreb—are home to just slightly over a hundred AI scaleups, the equivalent of around a 4% share of all AI scaleups, and they raised under €1.5 billion in venture capital, which represents a mere 2% of the total funding raised by AI scaleups in Europe.

Third, cities with renowned universities in Western and Northern Europe perform very well in terms of AI scaleups, especially relative to their population and economic output. Given the overall highly skewed distribution of AI scaleups across scaleup cities, the density measurements indicate a slightly decentralised structure. In terms of AI scaleup density (number of AI scaleups per 100,000 inhabitants), AI scaleup funding density (€ million funding raised by AI scaleups per $1 billion of GDP), and AI scaleup patent density (number of patents filed for AI scaleups per 100,000 inhabitants), the leading FUAs are cities with world-class universities such as Cambridge, Oxford, Lausanne, Heidelberg, Basel, Munich, and Bristol. One of the reasons for these regional patterns, according to the paper by Cséfalvay, Papp, and Horlik, is that cities that have access to AI knowledge (measured by patent activity) perform better than cities that lack this background. Nevertheless, they found that the patents filed by AI scaleups were strongly concentrated in a few scaleup cities: around 2,100 patents, or 38% of all patents, were filed by scaleups from just four cities—London, Paris, Munich, and Stockholm. Added to the nearly 2,000 patents filed by AI scaleups in Cambridge, Mainz, Oxford, Copenhagen, Bristol, Basel, Berlin, Dublin, Zurich, Toulouse, and Oslo, these 15 cities together account for 75% of the patents. The authors constructed an ordinary least squares (OLS) regression model to examine the relationship between the total number of patents filed by AI scaleups in scaleup cities and the AI scaleup performance of these cities, and the model indicates a strong correlation. The authors also filtered AI-related patents for European countries from the PASTAT database using keywords, resulting in a total of almost 2,800 patent applications for the period from 2005 to 2020, of which more than 1,100 patents were granted. While there were 24 European countries where entities filed at least one AI-related patent, the vast majority of filed and granted patents were registered by entities from only a few countries: the United Kingdom, followed by Germany, France, and Spain. In this case, too, the country-level linear regression analysis shows a very strong correlation between countries’ performance in terms of AI scaleups and knowledge background, measured by AI-related patents. In other words, scaleup cities where AI scaleups file a large number of patents and countries where various entities also file a large number of AI-related patents perform significantly better in terms of AI scaleups than cities and countries where this knowledge background is weak. Using a unique database of scaleups in Europe from 2021 to 2024, built at the MCC Centre for Next Technological Futures, Szabolcs Dudás investigates the dynamics of scaleups during this period. In his paper, “Pareto Distribution in Europe: Which Scale-Up Cities Dominate and Which Struggle to Grow in Europe?”, he distinguishes four classes of scaleups with regard to growth dynamics: New Entrant, Disappeared, Survivor, and Absolute Survivor scaleups. Overall, the number of scaleups in Europe’s 123 scaleup cities increased by more than 14,500 during this period, from 12,000 in 2021 to 26,600 in 2024, while their total funding more than doubled from €205 billion to €533 billion. However, around 40% of this increase, and the interplay of new, disappeared, and surviving scaleups, was recorded in just four cities: the Global scaleup cities of London, Paris, Berlin, and Stockholm. More than 10,000 scaleups were classified as Absolute Survivors, meaning they were present in every year of the period, and the total funds they raised increased by around 68% from 2021 to 2024. Nevertheless, the type of industry still plays an important role here, as scaleups in Global and Top European scaleup cities predominantly focus on fintech, enterprise software, and healthcare, producing intangible assets that are more easily scalable, while scaleups in cities with lower-ranked clusters are strongly represented in energy and transportation and deal with less scalable tangible assets.

The fact that 40% of scaleups and 50% of their funding are concentrated in four of the 123 scaleup cities in Europe and that the majority of changes in scaleup classes, according to their growth dynamics (New Entrants, Disappeared, and Survivors), occurred in these cities, suggests that there is a very uneven distribution among scaleup cities in Europe. After a detailed theoretical discussion of unequal distributions—Pareto principle, power law, and lognormal distribution—Szabolcs Dudás analyses whether the distribution of scaleups in Europe, particularly with regard to their funding, follows these distribution patterns. He concludes that although the distribution does not meet the requirements of exact mathematical forms, the reality of scaleups in Europe follows the Pareto-like pattern.

In light of increasing environmental concerns, the growing role of corporate environmental, social, and governance (ESG) performance, and especially the transition to clean and green energy sources, Dávid T. Nagy analyses a new method of providing long-term secure energy solutions for companies in European countries: power purchase agreements (PPAs). A PPA is essentially an agreement between an electricity producer (often a renewable energy developer) and a buyer (such as a utility company, government agency, or corporation), in which the buyer agrees to purchase electricity at a pre-agreed price for a set period of time. In his paper, “Catalysts of Change: How Corporate R&D Spending, Green Energy-Related Patents, and Green Technology Scaleups Impact the PPA Performance of European Countries”, he provides an introduction to the two main types of PPAs. One is a physical PPA, in which the electricity generated by a third-party-owned renewable energy facility is delivered to an energy service provider, which then delivers the electricity to the corporate customer. The second is a financial or virtual PPA, in which the producer sells clean and green electricity to the grid, while the corporate customer purchases the electricity separately from the grid.

The fundamental question for Nagy, however, is which factors influence the PPA performance of different countries in Europe. To answer this question, he includes 14 European countries with the largest PPA capacities in his analysis: Belgium, Denmark, Finland, France, Germany, Greece, Ireland, the Netherlands, Norway, Poland, Slovenia, Spain, Sweden, and the United Kingdom. As the primary source for power purchase agreement market data, he uses the RE-source platform, which provides insights into the contracted capacity of PPAs in Europe by country, sector, company, and annual number of deals. To analyse the factors influencing PPA performance, he built a unique database. First, he used keyword analysis to retrieve nearly 3,800 granted patents for clean and green energy technologies from the PATSTAT database. Second, applying the industry classification of the Dealroom database, he was able to identify and retrieve data on more than 300 clean and green energy scaleups in the analysed European countries. Third, he filtered out almost 350 green companies from the list of the 1,000 largest European R&D investors that also have an ESG rating. To name just the top three countries for each metric: Spain, Germany, and Sweden lead in terms of total contracted PPA capacity by country; Germany, Spain, and the United Kingdom lead in terms of patents granted for clean and green energy technologies; the United Kingdom, Germany, and the Netherlands lead with regard to the number of clean and green energy scaleups; and Germany, France, and the Netherlands lead in terms of R&D investment by green companies.To accurately assess the role of these factors on the PPA performance of the 14 European countries studied, Nagy created a regression model. Specifically, he calculated the Pearson correlation coefficient between each data series. As important evidence for the prominent role of knowledge background, there is a very strong relationship between contracted PPA capacity and the number of granted clean and green energy patents across the 14 European countries, with a very high correlation coefficient. However, the relationship between contracted PPA capacity and the number of clean and green energy scaleups is rather moderate, and the correlation is similar for the R&D expenditure of green companies.

Finally, the paper by Zoltán Cséfalvay and Zalán Horlik, “Startups and Talent: How Access to Locally Available Talent at Universities Affects the Startup Performance of European Cities”, raises the crucial question of the relationship between the talent available at universities and the startup performance of European cities and, more mportantly, which universities offer the best breeding ground for startup founders. Therefore, to measure access to locally available talent, the authors retrieved data from Dealroom.co in 2024 on about 1,400 universities and colleges that had at least ten students who became startup founders during their studies or later. They then assigned these universities to their respective scaleup cities (FUAs) and analysed three indicators to assess them: the number of startup founders who attended a university in the scaleup city, the number of companies founded by startup founders who studied at a university in the city, and the number of startup founders who studied at a university in the scaleup city and have already raised more than €10 million in funding for their respective startups. In total, there were more than 150,000 startup founders in Europe’s scaleup cities who attended their universities and who founded a total of around 175,000 companies.

It came as no surprise that the distribution of scaleup cities in terms of access to talent also follows a pattern familiar from analyses of almost every aspect of the European startup world. First, the scaleup cities show a very uneven distribution. More than a quarter of startup founders (over 45,000) studied at universities in the four Global scaleup cities of London, Paris, Berlin, and Stockholm. Combined with the four Top European cities (Amsterdam, Dublin, Munich, and Zurich) and the six Strong European scaleup cities (Barcelona, Cambridge, Copenhagen, Helsinki, Madrid, and Oslo), well over half of startup founders attended a university in these 14 scaleup cities. Second, the East–West divide is visible here too, as fewer than 6,000 startup founders studied in the ten scaleup cities of Central and Eastern Europe—Bucharest, Budapest, Bratislava, Ljubljana, Krakow, Prague, Sofia, Warsaw, Wrocław, and Zagreb—which corresponds to less than 4% of all founders. Third, scaleup cities with world-class universities, such as Cambridge, Oxford, and St. Gallen, perform exceptionally well in terms of startup density— measured as the number of startupfounders who attended a university in the scaleup city per 100,000 inhabitants. But other smaller cities with top universities, such as Ghent, Uppsala, Gothenburg, Malmö, Aarhus, Leiden, Lausanne, Bristol, and Edinburgh, also have high density values.

To analyse in detail how access to talent impacts the overall performance of scaleup cities, Zoltán Cséfalvay and Zalán Horlik calculated two composite indices: one that summarises the scaleup performance of cities in a single number, and one that summarises in the same way access to locally available talent in the scaleup cities. The result of the linear regression model indicates a relatively strong relationship between the two indices. However, the outliers suggest that the relationship is not always that clear-cut. While Paris, for example, performs significantly better in terms of access to talent than in overall performance of scaleups, the situation in London is exactly the opposite: the city’s share in access to talent is lower than its share in the overall performance of scaleups. The Top European scaleup cities—Munich, Amsterdam, Zurich, and Dublin—all perform slightly better in scaleups than in access to talent, while Barcelona, Cambridge, and Madrid, which are among the Strong European scaleup cities, perform very well in access to talent, scoring almost twice as high as their scaleup performance index values.

According to the authors, the reason for this relatively strong relationship between access to talent and the overall scaleup performance of cities, on the one hand, and the various outliers that seem to break this relationship, on the other hand, requires an in-depth analysis of those universities that are breeding grounds for startup founders. In the first step of this analysis, they created a consolidated ranking for a total of 657 European universities based on the four most recognised global university rankings—QS World University Rankings, Times Higher Education World University Rankings (THE), Center for World University Rankings (CWUR), and SCImago Institutions Rankings (SIR)—which primarily assess the academic performance and reputation of these institutions. In the next step, the authors created another ranking of universities based on the Dealroom data on startup founders who studied there. Since the overwhelming majority of startups fail in a relatively short period of time, the authors calculated a ranking of universities in terms of the value creation potential of startup founders.

The two rankings of European universities—the academic performance ranking based on global university rankings (QS, THE, CWUR, SIR) and the ranking of universities in terms of the value creation potential of startup founders based on Dealroom data—differ significantly; between them, there is only a moderate relationship with weak explanatory power. Here, too, the outliers are the ones that make the picture clearer. On the one hand, the very high rankings of universities in terms of their academic performance are associated with the relatively good value creation potential of startup founders. For example, if one looks at the top 100 universities both in the ranking according to academic performance and in the ranking according to the value creation potential of startup founders, the result is still a cross-section of 55 universities. On the other hand, many universities, especially business schools and technical universities, are highly ranked in terms of their value creation potential for startup founders but are in the middle or bottom half of the European university rankings in terms of academic performance.

The authors therefore conclude that innovation—the development of new technological or business solutions, products, and services that meet customer needs—is closer to business and technology than to scientific research and elite education; for this reason, the true breeding grounds for startup founders are usually found at the best business schools and technical universities, while, of course, world-class universities often provide their students good opportunities to establish a startup during their studies or later in their professional lives.

The research for these four papers was conducted at the Mathias Corvinus Collegium, which, as a talent development institution, is perfectly suited to the fundamental question addressed by all of them: the role of knowledge for startups in an age of rapid technological change. The Mathias Corvinus Collegium also supported the presentation and discussion of research results at various international conferences, including the conferences of the Eurasia Business and Economics Society (EBES) in Lisbon and Rome, and the IEEE 16th International Conference on Cognitive Infocommunications (CogInfoCom) in Vienna. The Centre for Next Technological Futures thanks the Mathias Corvinus Collegium for its outstanding support and facilities, which enabled the research showcased in this book.

Prof. Zoltán Cséfalvay

Head of the Centre for Next Technological Futures

Mathias Corvinus Collegium, Budapest