Compete or Disappear (III): A Market or a State

Draghi stops diagnosing and starts prescribing. Common debt, the race for artificial intelligence, a GDPR frozen in 2016: the bill is clear, what's missing is someone willing to pay it.

In the first two parts of this reading of "Competere o sparire" I dwelt on the diagnosis: the productivity gap, the fragmentation of the internal market, the pragmatic federalism that Draghi proposes as a method before it is a programme. In the second half of the book, from chapter 8 onwards, the tone shifts. Draghi gradually stops describing the patient and begins writing the case notes for those who ought to be treating him, that is, us, and above all our governments.

The Tariff We Gave Ourselves

The chapter on dismantling the barriers within the internal market is, to my mind, the most technically solid section of the entire book. The International Monetary Fund, Draghi notes, estimates that internal barriers to the single market amount to an ad valorem tariff of 45 per cent on manufacturing and 110 per cent on services. Let me repeat that figure, because it deserves repeating: 110 per cent. I am not talking about Trump's tariffs, which have filled the front pages for months. I am talking about the tariffs Europe imposes on itself, every day, through twenty-seven different regulatory regimes, precisely in financial, digital and professional services, the segments where the most innovative and highest-productivity part of the European economy is concentrated. Draghi had already made the point, with a certain wryness, in a Financial Times article in February 2025 with an eloquent title: forget the US, Europe has already put tariffs on itself.

From Taboo to Infrastructure

This is where the proposal originates that many, myself included when I had only heard of it second-hand, took to be a technocratic gambit fit for Brussels conferences: the issuance of common European debt, up to 15 per cent of GDP according to ECB estimates, to fund not current spending but the investments that no single member state can afford at the scale required. Draghi's reasoning is not ideological. It is a matter of financial arithmetic. A European safe asset, he argues, would benefit from a scarcity premium, a convenience yield, which would lower the cost of debt relative to expected growth rates. It is the same mechanism, I would add, that has made US Treasuries the instrument through which Washington has financed decades of deficits without paying their full price on the markets. Europe, for the first time, could have a comparable asset. But only if it stops treating the idea as a taboo to be rationed out in times of crisis and starts treating it as infrastructure.

Zero Out of Ten

Chapter 9, titled with an almost desperate imperative, "Do Something!", captures how far reality had outpaced the original report, published barely a year earlier. Here appears the figure of 750-800 billion euro a year, "to be conservative", needed for energy, defence, digitalisation and research and development. But the figure that struck me most concerns artificial intelligence itself, the area Draghi says has moved faster than any other since September 2024. Large language models, the systems underpinning the current AI revolution, trained on vast quantities of text to respond, reason and generate content, are today the near-exclusive preserve of two countries. Eight of the world's ten most advanced models were developed in the United States, the other two in China. Zero, I repeat zero, are European. It is a figure that on its own tells us more than any political speech about where the technological centre of gravity of the planet currently sits, and how much Europe risks remaining, in this contest, a mere regulatory spectator rather than a protagonist. And the costs of building these models have fallen sharply since the original report was published: tenfold for training costs, that is, the cost of "teaching" the system by having it process vast quantities of data before it is ready for use, and twentyfold for inference costs, that is, the cost incurred every time the model is actually used, for instance to answer a question. In practical terms, building and running an advanced artificial intelligence system now costs a fraction of what it did two years ago. This is only good news in theory, because it means the economic barrier to entering this race is falling precisely at the moment when Europe, rather than seizing the opportunity to catch up, continues to debate whether and how to regulate a technology its competitors have already learned to build at ever lower cost.

Frozen in 2016

Chapter 14, devoted entirely to governing artificial intelligence, is probably the one where Draghi gets the most off his chest, with an understatement that is very institutional in form but unmistakable in substance. Take the GDPR, the General Data Protection Regulation that the European Union enacted in 2016 to safeguard citizens' privacy against the mass collection of personal data by companies. Draghi does not dispute its objective, which he considers legitimate, but the way it was built: a law designed for a world that, ten years on, simply no longer exists. Training modern AI systems requires vast quantities of data, and the legal uncertainty over what may lawfully be collected and used has, Draghi argues citing academic research, increased the cost of data by 20 per cent for European firms and cut venture capital investment in the continent's technology sector by roughly a quarter. In practice, a rule designed to protect the European citizen from big platforms has ended up hitting small European startups hardest, since they lack the legal resources of American multinationals to navigate the regulatory complexity, while those same multinationals continue to operate more or less unimpeded in markets across the world. It is an analysis I share on the diagnosis, while remaining convinced, from the political science training I carry with me from my years at Cesare Alfieri, that regulatory precaution stems from a legitimate concern, that of protecting the citizen from the often opaque power of whoever controls the data. The problem, for Draghi, is not that Europe regulated. The problem is that it froze, in 2016, a balance between privacy and innovation that technology has since overtaken ten times over, without ever pausing to update it.

A Market or a State

Chapter 15, the closing one, is where Draghi finally drops the posture of the technician and takes up that of the politician, in the highest sense of the word. Its title, "A Market or a State?", poses the question I myself have tried to phrase in sharper terms in a recent conversation: can Europe go on being merely a large regulated market, or must it choose to become an autonomous geopolitical actor, with all that this implies in terms of ceding national sovereignty? Draghi uses an example I find illuminating: Greenland. Faced with the threat of annexation, he argues, Europe was for the first time forced to make a genuine strategic assessment of its own levers of power, and the resolve it showed generated a solidarity that no treaty could ever have imposed from above. It is, however indirectly, confirmation of the thesis I was discussing a few days ago: political courage rarely stems from a pre-existing consensus, it almost always stems from a shock that makes visible what was previously only theory.

The phrase Draghi coins to describe the method he proposes, "pragmatic federalism", strikes me as the most honest summary of the whole book. Pragmatic because limited to what is currently feasible, with the partners currently available, in the areas where concrete progress is currently possible. Federalist because the goal, stated without euphemism, remains political union. It is not the rhetorical shortcut of those who want a Europe united in words and divided in fact. It is, rather, the acknowledgement that European integration has never rested on domination or subordination, unlike the Chinese model or the one the United States now seems intent on pursuing, but on shared will. Which, Draghi adds in one of the few lines in the book where genuine optimism shows through, makes it harder to build, but also the only form of power Europe can aspire to exercise without betraying itself.

Closing this three-part reading, I am left with a conviction the book confirms more than it disproves: Europe today has all three ingredients that, according to Draghi himself, have historically preceded every leap forward in integration, namely a manifest crisis, an external political shock and an action plan already on the table. All that is missing, as ever, is the political will to use them before the window closes. And here the book, in fairness, offers no easy answers: it offers a pitiless diagnosis and an implicit deadline that none of the twenty-seven governments has yet truly agreed to face.