Europe has never managed to establish a single technological sovereign. For five centuries, its states competed with each other, copied their neighbors, and welcomed the exiles of some and the inventions of others. This institutional disorder produced the Industrial Revolution, modern chemistry, quantum physics — and today, in competition with China and the United States, this same disorder is often presented as Europe’s principal handicap. A recent synthesis spanning a thousand years of economic history suggests that this diagnosis is at least incomplete, and possibly inverted.

The work reviewed by Foreign Affairs in 2026, published by Princeton University Press in September 2025, reconstructs the relationship between institutional structures and technological trajectories over the long term. Its central thesis is uncomfortable for proponents of unified governance: decentralized systems excel at exploring new technologies, centralized systems excel at deploying them at scale. Both capacities are necessary. No structure sustainably maintains both. And history knows no civilization that has managed this transition indefinitely.

The Essentials

  • Over a thousand years of economic history, decentralized systems generate more breakthrough innovations; centralized systems deploy already-established technologies faster and at greater scale.
  • China and the United States each embody a different institutional vice: concentration of power in Beijing, concentration of dominant enterprises in Washington and San Francisco.
  • Europe, institutionally fragmented, presents the historical profile of great exploration spaces — at the cost of structural difficulty in scaling up.
  • No power has succeeded in maintaining both postures simultaneously over the long term; the question for 2050 is whether AI and biotechnologies reward exploration or deployment.
  • Europe’s challenge is not to unify further, but to organize coordination without destroying the diversity that produced its historical advantage.

What a Thousand Years of Economic History Says That G7 Summits Forget

Long-duration economic historiography is not an academic exercise. It produces diagnoses that electoral cycles are unable to see. When one reconstructs technological trajectories since the Middle Ages — printing, steam engines, chemical processes, electricity, semiconductors — a pattern repeats with uncomfortable regularity.

Politically fragmented spaces generate breakthroughs. The Europe of Italian city-states invented modern accounting and financed maritime exploration. The Europe of German principalities produced Gutenberg, then two centuries later Liebig, Bayer, and BASF. England, isolated from the continent and relatively decentralized compared to continental monarchies, harbored Watt and Arkwright. Competition between states creates selective pressure on ideas: what fails in one territory can succeed in a neighbor. Inventors, heretics, dissidents migrate toward the most permeable spaces. Fragmentation is a machine for producing experimental diversity.

Politically unified spaces deploy at scale what the former invented. Han China standardized writing, weights and measures, roads — and accelerated the diffusion of iron across an immense territory. Jacobin France imposed the metric system, the Civil Code, the grandes écoles that industrialized engineer training. New Deal America electrified rural areas in a decade. Centralization is a machine for scaling, not for inventing.

This divide is not a natural law. It is a robust empirical result over the long term. It has exceptions — AT&T’s Bell Labs produced the transistor within a quasi-monopolistic structure — but the exceptions confirm the general pattern. Decentralization favors exploration, centralization favors exploitation.

China Deploys Better Than It Invents, and It Knows It

Beijing has drawn the lesson from this history without formulating it in the same terms. China’s model of technological development has long operated on a simple principle: import, absorb, improve, deploy at scale. China became the world’s leading manufacturer of solar panels through massive state investments and development of its own supply chain, successively surpassing the United States, Japan, and Germany. It built the world’s largest high-speed rail network from shinkansen technologies purchased under license. In renewable energy, the country installs roughly four to five times more new capacity annually than Europe does, at an unparalleled pace worldwide.

The limitation of the model is symmetrical to its strength. Where exploration of virgin territory is necessary — frontier generative AI, breakthrough biotechnologies, advanced semiconductor design — concentration of power creates blind spots. Chinese researchers publish prolifically, but in fields where orientations are defined by the Party. Self-censorship on dual-use applications, data constraints, and mistrust of organic, unplanned structures slow precisely the kind of disorderly exploration that produces breakthroughs.

The case of semiconductors is instructive. Despite massive public investments — over 150 billion dollars mobilized since 2014 according to the Semiconductor Industry Association and other American sources — China’s real limitation in this field is not 7 nanometers: SMIC achieved that thanks to ASML’s DUV equipment already imported. The true ceiling lies below 5 nanometers, where EUV machines, now inaccessible to Beijing, become essential — and where existing DUV equipment can neither be replaced nor repaired. Centralized money is insufficient when what needs to be produced is distributed ingenuity.

The United States Explores, But Its Center of Gravity Is Shifting

The American case is more subtle and, in certain respects, more concerning for those who value long-term dynamics. The United States functioned for most of the twentieth century as a remarkable decentralized system: autonomous universities, competing federal agencies, atomized venture capital, talent mobility. The country’s institutional diversity — fifty states, thousands of universities, a startup-garage culture — produced personal computing, the internet, and messenger RNA.

But over the past twenty years, a silent concentration has taken place. Five companies — Apple, Microsoft, Nvidia, Alphabet, Amazon — represent approximately 27 to 30 percent of total S&P 500 capitalization. In AI, two to three actors control frontier models, training data, and computing infrastructure. What economists call the “superstar economy” produces globally competitive champions extraordinarily efficient at exploiting existing technologies. It simultaneously reduces the selective pressure that produces subsequent breakthroughs.

The concentration of dominant American tech companies creates a different institutional vice than Beijing’s — less coercive, more subtle — but potentially equally limiting over a horizon of several decades. A private oligopoly can stifle exploration nearly as effectively as a planning state, simply by setting standards, acquiring emerging competitors, and capturing the best engineers. The accelerating wage concentration in American tech companies — which pay juniors less and less and seniors more and more — is a symptom of this phenomenon: the market structure tends to concentrate gains on a few profiles, which is exactly the behavior of a system that exploits rather than explores.

Europe Does Not Suffer From a Lack of Unity, But From a Lack of Coordination

Here, the lesson of a thousand years disrupts the dominant narrative. The standard story about Europe unfolds in three acts: institutional fragmentation, absence of global champions, technological decline. The logical conclusion is accelerated integration: a true capital markets union, a European AI champion backed by sovereign funds, a continental industrial policy in the fashion of the American Chips Act or Chinese subsidies.

This narrative is not wrong. But it is incomplete, and long-term history suggests it can be dangerous if pushed too far. European fragmentation is precisely what produced its historical exploration capacity. Member states with different education systems, different regulatory cultures, different labor markets generate a diversity of approaches that homogeneous unification would erase. France and Germany do not have the same relationship to personal data, financial risk, or intellectual property. This divergence is a friction — and an insurance policy.

Recent figures on European fundamental research illustrate the paradox. Europe produces markedly more scientific publications than the United States in volume — approximately 28 percent of global publications versus 19 percent for the United States — even if the latter maintain their edge in terms of most-cited publications. Its universities — Oxford, Cambridge, ETH Zurich, EPFL, the major Dutch universities — rank among the world’s best for fundamental research. The European problem is not the generation of knowledge. It is their transformation into enterprises, into jobs, into deployment at continental scale.

The challenge is therefore not to reproduce the centralized model to rival Beijing or Washington. It is to build coordination mechanisms that allow European diversity to scale without suffocating it. The distinction is fine but decisive: coordinating is not unifying. Common standards for medical data allow twenty different health systems to contribute to the same research database without Brussels deciding each country’s public health policy. This model — informational commons, institutional competition — is precisely what long-term history suggests as optimum for a large decentralized space that also wants to deploy.

The dynamics of open science in biotechnologies illustrate what can result from pooling resources without centralizing decisions: when AlphaFold made a billion protein structures accessible to all, the most diverse European teams — not the largest ones — were the first to draw unprecedented applications from them.

The Impossible Transition: Why No Power Has Sustained Both Postures

Perhaps the most disturbing lesson from long-term economic history is this: no civilization has durably combined excellence in exploration and excellence in deployment. Eighteenth-century Britain explored; nineteenth-century Britain deployed across an empire; twentieth-century Britain managed decline. The United States explored from 1945 to 1990, deployed in the 1990s-2000s with the internet, and began to concentrate. Song China was perhaps the world’s most technologically advanced system in the tenth century; progressive centralization reduced its experimental capacity, and Europe caught up and surpassed it between the fourteenth and seventeenth centuries.

This pattern has direct implications for technological competition at the horizon 2050. Generative AI, biotechnologies, and new forms of energy are general-purpose technologies — they reconfigure all sectors, as electricity did in the twentieth century. For each, the exploration phase (defining what is possible) precedes the deployment phase (spreading what works). The two phases reward different institutional structures.

On AI, the phase of fundamental exploration is probably still open — future architectures are not fixed, as recent work on autonomous agents and multimodal models shows. The acceleration of AI agents in scientific research suggests that competition for frontier applications is not closed. A decentralized structure retains an advantage in this phase. On biotechnologies, the same uncertainty prevails: cell therapies, next-generation mRNA vaccines, therapeutic genome editing are in phases of accelerated exploration. These are domains where team diversity, diverse approaches, and diverse regulatory cultures have historically produced the best surprises.

The transition to a phase of dominant deployment — which will reward centralization — is possible but not certain. It assumes that foundational technologies stabilize, that standards become established, that economies of scale dominate. If this transition occurs before 2040, China and major American platforms will enjoy a structural advantage. If it delays, Europe and its models of loose coordination could hold their rank. History does not say which of these scenarios will materialize. It only says that betting on a single institutional posture, whatever it may be, has always proved costly in the end.

What Europe Must Build Without Destroying What It Has

The operational conclusion is not to leave Europe as is. Current European fragmentation carries real defects that have nothing virtuous about them historically: twenty-seven capital markets that make financing a continent-wide startup harder than in the United States, competition rules that have sometimes protected established players against emerging competitors, decision-making slowness that lets windows of opportunity close.

What long-term analysis suggests is that useful reforms are those that build coordination mechanisms without standardizing approaches. A unified venture capital market would allow European entrepreneurs to raise funds in Berlin, Warsaw, and Madrid without recreating a single institutional model that erases specificities. Common databases for clinical trials would allow each country to conduct its own health policy while contributing to collective research efforts. Interoperability standards for industrial data would allow German, French, and Spanish companies to cooperate without merging their governance.

This is a less spectacular vision than creating an “Airbus of AI” or a European semiconductor champion backed by tens of billions in public funds. But it is more faithful to what economic history teaches about the conditions for durable technological success.

European fragmentation is not the problem to solve. It is the resource to organize. The question for European policymakers over the coming decade is simple to state, difficult to maintain: how do we build enough unity to deploy without building so much that we stop exploring?

Sources

  1. Foreign Affairs — review of How Progress Ends: Technology, Innovation, and the Fate of Nations (Princeton University Press, September 2025): https://www.foreignaffairs.com/reviews/how-progress-ends-technology-innovation-and-fate-nations
  2. European Commission — data on the European Research Area and scientific publications (ERA Progress Report, 2023-2024 editions)
  3. Mercator Institute for China Studies (MERICS) — estimates of Chinese public investment in semiconductors since 2014
  4. S&P Global Market Intelligence — S&P 500 market capitalization data by company
  5. Princeton University Press — How Progress Ends (official page): https://press.princeton.edu/books/hardcover/9780691233079/how-progress-ends
  6. Wikipedia — High-speed rail in China: https://en.wikipedia.org/wiki/High-speed_rail_in_China
  7. Semiconductor Industry Association — China Semiconductor Report: https://www.semiconductors.org/taking-stock-of-chinas-semiconductor-industry/
  8. MERICS — China Tech Observatory: Semiconductors: https://merics.org/en/china-tech-observatory/semiconductors
  9. Visual Capitalist — Every S&P 500 Company in 2026: https://www.visualcapitalist.com/the-entire-sp-500-in-2026-in-one-chart/
  10. HCÉRES — Europe produces more publications than China and the United States: https://www.hceres.fr/en/news/europe-produces-more-scientific-publications-china-and-usa-usa-still-leads-terms-excellence
  11. IEA Renewables 2025 / Carbon Credits — renewable energy growth: https://carboncredits.com/renewables-2025-how-china-the-us-europe-and-india-are-leading-the-worlds-clean-energy-growth/
  12. Nikkei Asia — Can China build its own ASML?: https://nikkei.shorthandstories.com/can-china-build-its-own-asml/
  13. ERA Monitoring Reports 2023 and 2024: https://european-research-area.ec.europa.eu/era-monitoring-reports