PARIS/SAN FRANCISCO – Former Italian Prime Minister and European Central Bank President Mario Draghi’s recent report on the European Union’s competitiveness was a sorely needed wake-up call about the need for deep reforms to revive productivity growth, drive the energy transition, and support the bloc’s defense. The report, coming as the EU prepares to usher in a new commission, could not be more timely.
But while Draghi offers sobering and impressive insights into Europe’s economic weakness, in two key areas his report still understates the problem that the new EU Commission, and EU member governments, must come to grips with: artificial intelligence and the startup ecosystem required to develop it. In the age of AI, a competitive startup ecosystem is critical to both economic and military security.
And despite having excellent universities and research institutions, the EU is falling ever further behind in the race to develop and deploy AI technology. But the underlying cause of this condition is not specific to AI; it reflects the wider inferiority of the EU’s startup ecosystem. With partial exceptions – France and Scandinavia – the EU’s startup ecosystem is vastly inferior to those of the United States, China, Israel, Taiwan, and even the United Kingdom and wartime Ukraine. And this problem will soon worsen, because the rise of AI threatens to sharply increase the economic and geopolitical price that the EU will pay for its inferior system.
A New Epoch
This price was already high before the commercial AI revolution exploded into public view in late 2022. Since the advent of the commercial internet in the mid-1990s, information technologies have become more critical to economic, governmental, and military activity, and recently created companies have accounted for a larger share of economic growth. Amazon, Google, Uber, Facebook, Airbnb, Netflix, and Tesla, all established in the 1990s or later, have combined revenues of more than $1 trillion and nearly two million employees. The AI chipmaker Nvidia is 31 years old, and dozens of other less famous US startups and venture capital funds have become major enterprises.
For several related reasons, the AI revolution dramatically increases the cost of having an inferior startup ecosystem. First, even if narrowly defined, the AI industry will undoubtedly become large and enormously important. Nvidia is already one of the world’s most valuable companies, and newer AI startups such as OpenAI, Anthropic, and Perplexity are experiencing revenue growth of 20-40% per month. Several recent AI startups (and others yet to be created) will soon be major enterprises, providing critical services. Any country or region with a globally competitive AI sector will derive greater wealth and influence as a result.
Moreover, AI is accelerating the creation and growth of new firms not just in AI but nearly everywhere. AI not only drives the development of novel products and services; it also reduces barriers for any entrepreneur in any industry when it comes to acquiring knowledge and performing administrative tasks.
This is no small matter. The evidence increasingly suggests that AI represents much more than just another technical “revolution,” in the vein of personal computers or cloud services. Instead, AI is poised to become an engine of profound economic change, possibly comparable in importance to the first Industrial Revolution more than two centuries ago. If so, the outcome of the AI race will determine the global distribution of wealth and power for generations to come.
And lastly, AI’s potential is not confined to new industries. Within this decade, using AI effectively will become critical for most older industries, too. For example, many technologists believe that the “ChatGPT moment” has arrived or is imminent in pharmaceuticals, biotechnology, robotics, legal and financial services, software engineering, autonomous vehicles, education, and weapons. AI is already fueling a profound revolution in military power, particularly (but not solely) in conjunction with drone technology, making the military implications of lagging in AI particularly disturbing.
Bad for Business
Startups – and their absence – are at the heart of all these issues. The difficulty of starting and expanding a business in the EU both prevents new industries from rising and also protects incumbents in older industries from competition. The result is that many EU industries are dominated by oligopolies of stagnant firms that can no longer compete globally.
Consider automobiles. There is no doubt that the rise of Tesla, and then of Chinese startups, accelerated the global transition to electric vehicles. While these companies were achieving rapid growth, some of Germany’s largest automobile manufacturers were falsifying smog tests in an effort to prolong the lifespan of diesel technology. Beyond the appalling breach of corporate ethics, this episode revealed a lack of concern about competition from new entrants, and a resistance to new technology. Sustaining economic dynamism requires exactly the opposite mindset.
Startups will continue to play a vital role in pioneering new industries and renewing old ones. But the EU’s inferior startup ecosystem severely handicaps new firms. Part of the problem is that many normal government decisions are needlessly slow. To cite just one example, it took Greece four years to grant Microsoft permission to build a data center.
In some EU countries, including Germany and Spain, merely forming and investing in a new company is unnecessarily time consuming and expensive. The Spanish process is archaic to the point of absurdity, requiring expensive legal work, notarizations, and apostilles. A new law intended to streamline incorporation and investment for Spanish startups took effect in 2023; but to qualify, companies must meet many requirements, including that 60% of its employees “have an employment contract in Spain.”
The German process is not much better, and multiple anecdotes from startup founders make clear that dysfunctional regulatory processes are a major problem. For example, a Taiwanese startup founder living in Germany recently told one of us that to apply for a permit, she had to hire a German lawyer to read the entire document to her in German, and obtain her permission in person. No electronic signatures were allowed, and because her German was not yet sufficiently fluent, she also had to hire a licensed interpreter (using her own bilingual colleague was not acceptable) to repeat the full text of the document to her in English. Equally Kafkaesque, German privacy and data regulations are sometimes interpreted to prohibit transfers of medical data outside of the building in which the data was created, even when patients and doctors have agreed to it.
Another problem concerns the EU’s research and educational systems. Although the EU produces excellent research, it repeatedly fails to commercialize its results, in part because research institutions are not adequately connected to the economy. Many of the EU’s best business schools – INSEAD, for example – are standalone institutions, rather than being housed within larger universities with strong STEM (science, technology, engineering, and math) programs.
In contrast, most leading US business schools are part of larger universities that have joint degree programs combining STEM subjects with business. Many also have extremely active programs to support entrepreneurship (including clubs, grants, alumni networks, incubators, competitions, patent licensing, and academic leave policies), even for students who drop out to start companies.
Within the EU, France and Scandinavia are the exceptions that prove the rule. They have made recent policy changes and launched initiatives (“la French Tech”) to support entrepreneurship, and there are signs that their own startup ecosystems are growing – as demonstrated by the launch of the French company Mistral AI. Overall, however, the EU is far behind, and falling further behind. Together, Israel and the UK have nearly as many “unicorns” (private companies valued at $1 billion or more) as the entire EU; and both China and the US have vastly more. In AI, the gap is even wider.
While the problems with Europe’s startup ecosystem have been much discussed, the solutions being proposed are often flawed. The EU has repeatedly created large, government-run research-and-development programs, but such responses miss the point. Even though the World Wide Web was invented in Europe, at CERN (the European Organization for Nuclear Research), it was commercialized almost exclusively by US startups. Moreover, as the commercialization of AI accelerates, the locus of R&D is shifting rapidly from academia to industry, both to incumbents (Microsoft, Google, Meta) and well-funded startups (OpenAI, Anthropic, Mistral AI, Safe Superintelligence).
Similarly, many discussions of reform focus on the need to harmonize EU member states’ regulations or to unify the EU’s 27 fragmented national markets. But while this is certainly an issue, it is secondary. After all, Israel and Taiwan have developed globally competitive startup systems, and they are home to only nine million and 23 million people, respectively. Israel and Singapore each have five times more startup unicorns per capita than the EU, and, aside from France, the most successful national startup systems within the EU are in small Scandinavian countries. With one probable exception – capital markets – the fragmentation of Europe’s national markets matters far less than the substance and administration of regulations.
Imperfect, But Better
Technological progress in AI is now blindingly rapid – faster even than at the dawn of the internet revolution of the 1990s. Even the US startup system is having trouble keeping up; its institutional arrangements are quickly evolving with a profusion of new incubators, hacker houses, AI-focused “solo GP” VC funds, new VC funds housed within AI unicorns, angel networks, founder cooperatives, and so forth.
This is not to say that we endorse blindly copying the US system. A university education in America has become stratospherically expensive; immigration policy is hugely dysfunctional; and the discriminatory tax treatment of startup founders and VCs versus ordinary employees severely worsens inequality.
This last point is underappreciated. The founders and earliest employees of US startups can purchase their stock at negligible cost and are not taxed until they sell; the same is true of VCs and their investors (“limited partners”). As a result, they pay only lower capital-gains taxes when they do sell, and many avoid taxes completely by borrowing against their stock without selling it.
In contrast, average employees are taxed immediately when an acquisition or public offering establishes a clear value for their stock. They often are forced to sell stock immediately to pay those taxes, and therefore face a higher tax rate because the gains are categorized as “ordinary income.” This differential treatment is now a significant driver of rising inequality in the US.
Still, for all its imperfections, the US system is highly productive, dynamic, and responsive; even in the age of remote work, people come to the US from around the world, including the EU, to study and work in AI.
What Europe Must Do
To its credit, Europe has demonstrated leadership in antitrust/competition policy and several other areas that could aid in developing a strong startup ecosystem. Its robust social safety nets could help cushion the dislocations that AI may cause, as well as assist would-be entrepreneurs in taking the risk of leaving a stable job.
But the EU must also support – but not govern or control – technology ecosystems’ institutions and remove sources of drag, such as overly restrictive administration of data and privacy regulations (while not overlooking valid concerns). It also must promote universal education in English and basic computer science; revise employment regulations; reduce the complexity and expense of incorporation, financing, and hiring; reduce immigration barriers for startup employees; foster R&D; promote a startup culture in universities; and establish a capital-market union. The lack of a unified capital market remains a key disadvantage relative to the US, with consequences for valuations and access to finance.
The EU startup ecosystem will need major reform if Europe is going to prosper in the age of AI. But reforms must also be accompanied by improved, and more balanced, transatlantic cooperation in AI, technology, trade, competition, and security policy. Neither European decline nor a bitter US-EU rivalry would benefit either party, particularly when both face common threats ranging from climate change to Russia under Vladimir Putin and China under Xi Jinping. The US-EU relationship requires leadership, vision, and a sense of common interest, not only for the benefit of the EU and America, but also to ensure that AI becomes a key tool in addressing the biggest global challenges of our time.
PARIS/SAN FRANCISCO – Former Italian Prime Minister and European Central Bank President Mario Draghi’s recent report on the European Union’s competitiveness was a sorely needed wake-up call about the need for deep reforms to revive productivity growth, drive the energy transition, and support the bloc’s defense. The report, coming as the EU prepares to usher in a new commission, could not be more timely.
But while Draghi offers sobering and impressive insights into Europe’s economic weakness, in two key areas his report still understates the problem that the new EU Commission, and EU member governments, must come to grips with: artificial intelligence and the startup ecosystem required to develop it. In the age of AI, a competitive startup ecosystem is critical to both economic and military security.
And despite having excellent universities and research institutions, the EU is falling ever further behind in the race to develop and deploy AI technology. But the underlying cause of this condition is not specific to AI; it reflects the wider inferiority of the EU’s startup ecosystem. With partial exceptions – France and Scandinavia – the EU’s startup ecosystem is vastly inferior to those of the United States, China, Israel, Taiwan, and even the United Kingdom and wartime Ukraine. And this problem will soon worsen, because the rise of AI threatens to sharply increase the economic and geopolitical price that the EU will pay for its inferior system.
A New Epoch
This price was already high before the commercial AI revolution exploded into public view in late 2022. Since the advent of the commercial internet in the mid-1990s, information technologies have become more critical to economic, governmental, and military activity, and recently created companies have accounted for a larger share of economic growth. Amazon, Google, Uber, Facebook, Airbnb, Netflix, and Tesla, all established in the 1990s or later, have combined revenues of more than $1 trillion and nearly two million employees. The AI chipmaker Nvidia is 31 years old, and dozens of other less famous US startups and venture capital funds have become major enterprises.
For several related reasons, the AI revolution dramatically increases the cost of having an inferior startup ecosystem. First, even if narrowly defined, the AI industry will undoubtedly become large and enormously important. Nvidia is already one of the world’s most valuable companies, and newer AI startups such as OpenAI, Anthropic, and Perplexity are experiencing revenue growth of 20-40% per month. Several recent AI startups (and others yet to be created) will soon be major enterprises, providing critical services. Any country or region with a globally competitive AI sector will derive greater wealth and influence as a result.
Moreover, AI is accelerating the creation and growth of new firms not just in AI but nearly everywhere. AI not only drives the development of novel products and services; it also reduces barriers for any entrepreneur in any industry when it comes to acquiring knowledge and performing administrative tasks.
This is no small matter. The evidence increasingly suggests that AI represents much more than just another technical “revolution,” in the vein of personal computers or cloud services. Instead, AI is poised to become an engine of profound economic change, possibly comparable in importance to the first Industrial Revolution more than two centuries ago. If so, the outcome of the AI race will determine the global distribution of wealth and power for generations to come.
And lastly, AI’s potential is not confined to new industries. Within this decade, using AI effectively will become critical for most older industries, too. For example, many technologists believe that the “ChatGPT moment” has arrived or is imminent in pharmaceuticals, biotechnology, robotics, legal and financial services, software engineering, autonomous vehicles, education, and weapons. AI is already fueling a profound revolution in military power, particularly (but not solely) in conjunction with drone technology, making the military implications of lagging in AI particularly disturbing.
Bad for Business
Startups – and their absence – are at the heart of all these issues. The difficulty of starting and expanding a business in the EU both prevents new industries from rising and also protects incumbents in older industries from competition. The result is that many EU industries are dominated by oligopolies of stagnant firms that can no longer compete globally.
Consider automobiles. There is no doubt that the rise of Tesla, and then of Chinese startups, accelerated the global transition to electric vehicles. While these companies were achieving rapid growth, some of Germany’s largest automobile manufacturers were falsifying smog tests in an effort to prolong the lifespan of diesel technology. Beyond the appalling breach of corporate ethics, this episode revealed a lack of concern about competition from new entrants, and a resistance to new technology. Sustaining economic dynamism requires exactly the opposite mindset.
Startups will continue to play a vital role in pioneering new industries and renewing old ones. But the EU’s inferior startup ecosystem severely handicaps new firms. Part of the problem is that many normal government decisions are needlessly slow. To cite just one example, it took Greece four years to grant Microsoft permission to build a data center.
BLACK FRIDAY SALE: Subscribe for as little as $34.99
Subscribe now to gain access to insights and analyses from the world’s leading thinkers – starting at just $34.99 for your first year.
Subscribe Now
In some EU countries, including Germany and Spain, merely forming and investing in a new company is unnecessarily time consuming and expensive. The Spanish process is archaic to the point of absurdity, requiring expensive legal work, notarizations, and apostilles. A new law intended to streamline incorporation and investment for Spanish startups took effect in 2023; but to qualify, companies must meet many requirements, including that 60% of its employees “have an employment contract in Spain.”
The German process is not much better, and multiple anecdotes from startup founders make clear that dysfunctional regulatory processes are a major problem. For example, a Taiwanese startup founder living in Germany recently told one of us that to apply for a permit, she had to hire a German lawyer to read the entire document to her in German, and obtain her permission in person. No electronic signatures were allowed, and because her German was not yet sufficiently fluent, she also had to hire a licensed interpreter (using her own bilingual colleague was not acceptable) to repeat the full text of the document to her in English. Equally Kafkaesque, German privacy and data regulations are sometimes interpreted to prohibit transfers of medical data outside of the building in which the data was created, even when patients and doctors have agreed to it.
Another problem concerns the EU’s research and educational systems. Although the EU produces excellent research, it repeatedly fails to commercialize its results, in part because research institutions are not adequately connected to the economy. Many of the EU’s best business schools – INSEAD, for example – are standalone institutions, rather than being housed within larger universities with strong STEM (science, technology, engineering, and math) programs.
In contrast, most leading US business schools are part of larger universities that have joint degree programs combining STEM subjects with business. Many also have extremely active programs to support entrepreneurship (including clubs, grants, alumni networks, incubators, competitions, patent licensing, and academic leave policies), even for students who drop out to start companies.
Within the EU, France and Scandinavia are the exceptions that prove the rule. They have made recent policy changes and launched initiatives (“la French Tech”) to support entrepreneurship, and there are signs that their own startup ecosystems are growing – as demonstrated by the launch of the French company Mistral AI. Overall, however, the EU is far behind, and falling further behind. Together, Israel and the UK have nearly as many “unicorns” (private companies valued at $1 billion or more) as the entire EU; and both China and the US have vastly more. In AI, the gap is even wider.
While the problems with Europe’s startup ecosystem have been much discussed, the solutions being proposed are often flawed. The EU has repeatedly created large, government-run research-and-development programs, but such responses miss the point. Even though the World Wide Web was invented in Europe, at CERN (the European Organization for Nuclear Research), it was commercialized almost exclusively by US startups. Moreover, as the commercialization of AI accelerates, the locus of R&D is shifting rapidly from academia to industry, both to incumbents (Microsoft, Google, Meta) and well-funded startups (OpenAI, Anthropic, Mistral AI, Safe Superintelligence).
Similarly, many discussions of reform focus on the need to harmonize EU member states’ regulations or to unify the EU’s 27 fragmented national markets. But while this is certainly an issue, it is secondary. After all, Israel and Taiwan have developed globally competitive startup systems, and they are home to only nine million and 23 million people, respectively. Israel and Singapore each have five times more startup unicorns per capita than the EU, and, aside from France, the most successful national startup systems within the EU are in small Scandinavian countries. With one probable exception – capital markets – the fragmentation of Europe’s national markets matters far less than the substance and administration of regulations.
Imperfect, But Better
Technological progress in AI is now blindingly rapid – faster even than at the dawn of the internet revolution of the 1990s. Even the US startup system is having trouble keeping up; its institutional arrangements are quickly evolving with a profusion of new incubators, hacker houses, AI-focused “solo GP” VC funds, new VC funds housed within AI unicorns, angel networks, founder cooperatives, and so forth.
This is not to say that we endorse blindly copying the US system. A university education in America has become stratospherically expensive; immigration policy is hugely dysfunctional; and the discriminatory tax treatment of startup founders and VCs versus ordinary employees severely worsens inequality.
This last point is underappreciated. The founders and earliest employees of US startups can purchase their stock at negligible cost and are not taxed until they sell; the same is true of VCs and their investors (“limited partners”). As a result, they pay only lower capital-gains taxes when they do sell, and many avoid taxes completely by borrowing against their stock without selling it.
In contrast, average employees are taxed immediately when an acquisition or public offering establishes a clear value for their stock. They often are forced to sell stock immediately to pay those taxes, and therefore face a higher tax rate because the gains are categorized as “ordinary income.” This differential treatment is now a significant driver of rising inequality in the US.
Still, for all its imperfections, the US system is highly productive, dynamic, and responsive; even in the age of remote work, people come to the US from around the world, including the EU, to study and work in AI.
What Europe Must Do
To its credit, Europe has demonstrated leadership in antitrust/competition policy and several other areas that could aid in developing a strong startup ecosystem. Its robust social safety nets could help cushion the dislocations that AI may cause, as well as assist would-be entrepreneurs in taking the risk of leaving a stable job.
But the EU must also support – but not govern or control – technology ecosystems’ institutions and remove sources of drag, such as overly restrictive administration of data and privacy regulations (while not overlooking valid concerns). It also must promote universal education in English and basic computer science; revise employment regulations; reduce the complexity and expense of incorporation, financing, and hiring; reduce immigration barriers for startup employees; foster R&D; promote a startup culture in universities; and establish a capital-market union. The lack of a unified capital market remains a key disadvantage relative to the US, with consequences for valuations and access to finance.
The EU startup ecosystem will need major reform if Europe is going to prosper in the age of AI. But reforms must also be accompanied by improved, and more balanced, transatlantic cooperation in AI, technology, trade, competition, and security policy. Neither European decline nor a bitter US-EU rivalry would benefit either party, particularly when both face common threats ranging from climate change to Russia under Vladimir Putin and China under Xi Jinping. The US-EU relationship requires leadership, vision, and a sense of common interest, not only for the benefit of the EU and America, but also to ensure that AI becomes a key tool in addressing the biggest global challenges of our time.