Try the Grand Continent. The ideas and analyses you won’t find anywhere else. But you’ll see them everywhere soon. In your inbox starting from
8 euros per month
You have sought to quantify the cost of Europe’s vassalization and you arrive at four billion euros per day. How do you reach such an exorbitant figure?
It is enough to group four key components of our dependencies, each representing about one billion per day: one billion tied to our energy deficit, one billion that quantifies our industrial deficit with China, and one billion for our digital bill with American players. The last billion is the one European investors place outside Europe in what are called “real” assets.
Altogether, this amounts to nearly 1.5 trillion euros per year: the equivalent of 90% of Spain’s GDP, or half of what France produces in a year, leaving Europe and roughly going to our “partners,” a term that perhaps should be reconsidered.
Our deficit in rare earths is also often mentioned. It is very small in value, amounting to a few hundred million euros per year, and above all it is already integrated into the digital and Chinese deficits: the little that isn’t attributed there is basically “acceptable,” since it is largely tied to the rise of our non-fossil infrastructures, notably the production of decarbonized energy or the electrification of uses (permanent magnets for wind turbines, for electric vehicle motors).
What justifies adding these four items together? They refer to very different dependencies: on hydrocarbon suppliers, on China, on the United States, on capital markets…
The attractiveness of an economy, or for Europe, of an integrated economic market, is measured by the relative competitiveness of its production factors. It is, first, the strength of its workforce; second, its energy; third, its access to critical raw materials; fourth, its technological lead (entry barriers); and fifth, its cost of capital.
These four “deficits” are the expression of the disaggregation of these factors.
The energy deficit is well known; it epitomizes our dependence on fossil empires. Today we pay for natural gas between 80 and 100 €/MWh. It costs barely 8 €/MWh in the United States. Not only are we dependent, but the Russian crisis pulled us out of a competitive continental dependency because it is connected (our gas is now around 45% purchased as LNG, versus about 20% before the war, which further increases its cost even just to pay for liquefaction and regasification). Thus we buy too much and at too high a price.
Regarding raw materials and critical inputs, we are like most countries, lacking or possessing little of them, thus aligned with a commodities market and in imbalance particularly with the two giants (the United States and China), a vulnerability reinforced by the five-decade weakening of our heavy industrial tools that transform raw materials.
On labor costs, the Chinese deficit—which can be linked to the deindustrialization of our economies—clearly reflects the weak competitiveness of our labor costs. We can return to this, because this gap is narrowing, but as a result, what emerges is not the competitiveness of Chinese labor, but the competitiveness of the entire model: speed, integration, internal competition, and of course the very high stability of industrial decisions and the priorities that the Chinese regime imposes on society. This is reflected in industrial developments.
In what sense?
One can decide to buy the base of a hydrogen production plant in China. It will probably cost three times less, without being truly less efficient. But the installation of the plant, the entire array of auxiliary equipment to bring in, the entire permitting process, the conditions imposed on its implementation, the management of construction sites, will cause the entire European-installed plant to cost two to three times more than the same plant installed in China. Through our trade deficits with China, we trace the fatigue of the continent’s industrial dynamism, which we see in the difficulties of all major recent European industrial projects to meet their deadlines and budgets.
In technology, we obviously have a few strong pockets (aerospace, health, agri-food, nuclear partially because China, South Korea and the United States are very competitive), but our digital dependency is terrible.
This is relatively known and analyzed, which does not mean we address it.
The most troubling today is the fifth factor, less known, which is the cost of capital, in other words the return expectations of investors in Europe. If at first glance it remains fairly aligned (in stock market return analyses) with what is seen in the United States and somewhat lower than in China (due to country risk), it is rising.
Why is that?
Because of the rise in interest rates, which themselves result from higher energy costs more than from an overheating of demand relative to supply on the continent. And above all from higher profitability expectations that result from slower growth (so investors tilt toward more margin now rather than more growth later) and from uncertainty about the stability of Europe’s prioritized trajectories, and thus about the consistency of public policies that steer growth investments that are “lost in translation.” This is especially true regarding the energy transition.
This latter part is central because it calls into question what had kept Europe’s role as a central market: regulatory and institutional stability, the quality of infrastructure, and social balances.
We stand at the edge of a deflationary vicious circle, even as Europe’s ambition to redefine the rules of the international game seemed to put us back in the race by reformulating the notion of production and by compelling our main suppliers—which due to our deficits, we remain a continent and a market for energy, digital services and many other sectors—to align with us.
When the cost of capital rises, the levels at which investors decide to invest rise and decision-making on new projects becomes more difficult. When regulations also seem uncertain or revisable, investors wait, but that raises project costs and, in turn, return expectations.
The spiral is underway and it feeds the pursuit of our dependencies at billions of euros per day.
And it is a vicious circle, because these are as many resources we do not reinvest domestically and which concretely prevent us from rebuilding our autonomy.
How many of these four billions are backed by official statistics and how many rely on your own estimates?
Three of them are perfectly objectifiable.
According to Eurostat, the deficit with China reached 359.8 billion euros in 2025, up about 15% year on year. It is the item deteriorating the fastest. The trade deficit with China is now almost universal, which weakens China in its leverage, paradoxically: there is almost no sector where it can threaten us with tariffs if we impose them on our side.
Also according to Eurostat, the energy bill stood at 336.7 billion euros in 2025, down 11% year on year. But the war triggered by the United States against Iran has since increased it by more than 125 billion, according to the Commission, pushing it well beyond 400 billion.
The 300 billion invested outside Europe is our estimate, which aligns with the rough figure cited by Enrico Letta and Mario Draghi; it also emerges from the euro area’s current account surplus.
The services item retained with the United States is deliberately limited to digital services (cloud, software and digital services), for which European companies pay more than 250 billion euros per year to American companies, thereby gradually reducing Europe’s trade surplus with the United States which had remained in surplus until now.
The current AI dynamics obviously accentuate this issue.
These four items do not fall under the same accounting logic. Can they really be added together?
The idea was more to simplify a complex subject with a few rounding-offs on the central themes of our dependencies: energy, digital, China and the decline of our attractiveness (which remains relative, as we continue to attract many investors; on real assets, we’re talking about 150 to 200 billion euros per year, but these figures have been falling since 2023).
What does this figure change in the debate on the cost of the energy transition?
We prefer to talk about the cost of the transition rather than about the bill of the status quo.
What is in these figures is the bill of the status quo. Not only is it substantial, but it is fluctuating and sensitive to all geopolitical variations. Moreover, it does not circulate within the European economy but fuels other economies.
The cost and stakes of the transition have been widely presented: around 800 billion euros per year would be needed to build a resilient Europe, independent from other major economic blocs for its energy and industrial fundamentals.
What is less well understood is that the incremental cost of relocating our energy sources for our primary industries is small relative to the final product.
Thus, if producing fertilizers (which are derivatives of ammonia, itself a combination of hydrogen and nitrogen, the first usually derived from natural gas, except when produced via water electrolysis) “green,” i.e., based on decarbonized resources, costs 30-40% more than Russian fertilizers, which currently flood the European fertilizer market (50% of our imports), while 30% of our European factories are idle, the impact of this rise on the price of a loaf of bread is reduced to 1%. We must choose our dependencies, especially when their downstream cost is negligible.
The same applies to steel. Producing high-quality steel for European cars, avoiding coke-based production which emits 2 tonnes of CO₂ per tonne of steel, will cost 30-40% more for a hot-rolled coil. But this premium, 200 or 300 euros per tonne of steel, again represents only 1-2% of the cost of a mid-range sedan.
Let’s run through the big numbers on a concrete case.
The Stegra project (green steel in Sweden, 2 Mt/year capacity, the largest industrial project currently under construction in Europe, the first new steel plant from ore to finished product in over 50 years, about 10 billion euros) relocates three of the four cited parameters.
80% of the investors in this project are European, and we have even attracted Asian investors.
Had we wanted to do “as usual,” we would have imported pre-reduced ore from China; the import bill would have been, depending on market costs, 700 to 900 million euros.
If we had nevertheless wanted to valorize Swedish mineral resources but produced hydrogen from natural gas to reduce that iron ore, we would have increased Europe’s natural gas imports by 300 million euros, probably on a site more suitable for gas importation…
Instead, we will use hydrogen produced with the hydroelectricity of northern Sweden, which allows us to localize the entire value chain, and in addition to cut 90% of the production’s emissions.
The transition is that, across all sectors.
It is a virtuous path on each of the cited parameters, provided it is not disrupted along the way by the fourth, which we have temporarily abandoned, the digital question.
What is blocking then?
I hear more and more often that regulation is to blame. That is a poor way to pose a real problem.
In what sense?
Europe adopts about two directives per week, which are then transposed. The essential is implemented except on climate and energy, where 45% of directives adopted in the last ten years have not been implemented. Only about fifteen are central for climate, but they are highly significant in changing industrial processes and our energy market.
It isn’t a question of not regulating; it’s a question of applying regulation correctly and not changing it constantly.
From my vantage point: the twenty billion euros we can mobilize in Europe cannot be committed if directives are not transposed, and once transposed, if they are meant to be altered again in three or five years.
Energy projects are funded for five to eight years, they then form infrastructures for 30 years or more. We are playing extremely dangerously with Europe’s future, its reputation, and its independence.
And this is utterly calamitous in a context where AI is turning the transition into a black hole.
Into a black hole?
A black hole consumes everything around it and everything heads toward it, with nothing emerging except the increasing mass of the black hole. That is what is happening.
The current dynamic wants all data centers to settle where energy is cheap, near transport-network nodes, and to capture the scarce resources you need to build an industrial project: water, skilled labor, equipment, capital, etc.
We are contending for the same megawatts, the same connection points, the same capital, and the same workforce to build the projects.
Do the numbers confirm this capture?
According to the International Energy Agency, if all announced projects come to fruition—and not all will—data centers will account for 5-10% of the peak consumption in Germany and France, and almost 10% in Spain and the Netherlands. The project portfolio equals 130% of the installed capacity of data centers in Europe. Yet, it takes two to ten years to obtain a connection in the Union, and seven to ten years in the major hubs (Frankfurt, London, Amsterdam, Paris, Dublin), whereas a data center is built in one to two years. In the United States, forecasts are far more impressive. You will easily understand that the entity arriving with an immediate check goes ahead of the one building a plant in ten years.
Are data centers not a convenient scapegoat, while Europe’s energy-intensive industry seems to be closing capacities mainly due to energy prices and Asian competition?
That is indeed the issue: as long as we do not change the game rules about what energy should be and what a sustainable productive model means, and as long as we do not set strict rules on how to produce and on imports that are compatible with that, we will continue to see jobs and factories relocate for the reasons mentioned above.
We no longer have a competitive edge on the main terms of attractiveness unless we change the terms, the game rules.
When we remain the world’s leading collective energy import power and one of the top players in major commodities on the planet, we still have the means to decide what we buy and what we do ourselves. And this especially when the state of dependencies makes the risk of retaliation in the form of tariffs from our partners increasingly tolerable—without mentioning those who did not wait to unilaterally impose them, like the United States.
Mechanisms such as border carbon adjustments and the set of rules that assign a price to carbon, which acts as a substitute for protection against what is not produced locally (since we have little or no local fossil energy—European), are moving in this direction. The Industrial Accelerator Act (IAA) as well, with lead markets, provided we take value chains from upstream to downstream to avoid eroding competitiveness on semi-finished or finished products.
Why would that be more serious than any other industrial arbitrage?
The AI issue is far more consequential. First because electricity needs are enormous, and these needs do not have immediate local outlets in Europe, since data also travels so quickly and today further away.
Thus data has effectively become the vector for a massive capture of a currently scarce resource, low-carbon and competitive electricity. Exactly what we need to transform our primary industries into resilient, local industries.
Secondly, because job creation per MW in the data-center economy is very low and then distributed across all downstream global uses of data and what it enables, and furthermore everything suggests that AI deployment in the real economy will entail a lot of job destruction, not necessarily a creator’s destruction.
Finally, because we have not yet found the right instrument to ensure that what is built locally can, if necessary, be reserved and safeguarded to store, compute and serve the needs of our territories, given the very high interconnectivity of all global servers.
One MWh in a steel mill, a fertilizer plant or a methanol unit can create a tradable good, a value chain and thousands of direct and indirect jobs. The same MWh in a data center yields a service whose value is captured far away and only a few tens of on-site jobs. If the arbitration is made at the expense of the jobs and value added generated by an MWh, then we are creating the conditions for further deindustrialization.
AI, as we can see, has the intrinsic characteristics of becoming a “monopolizing” industry, because the strongest, with the most advanced models, enjoy, as in the digital economy, immediate access to volumes.
The ongoing energy transition is an approach that changes the rules of the game but maintains the market’s competitive dynamics, virtuous for end customers. For AI, that is not guaranteed.
This is not a theoretical discussion.
Several projects have already suffered from this sudden competition.
Some will no longer be built.
Others have seen their construction costs rise to the point of threatening viability. Others have had to cede access to planned sites because bids from data-center players were several times higher. This happens in Europe, in South America, in Canada, wherever abundant and competitive hydroelectric or nuclear potential exists.
When we are temporarily in a deficit of electricity demand (as is the case in France), it is a bit of a mirage and demand quickly inflates with projects, soon building a demand that would replace the one that came from implementing our decarbonized industries, sources of downstream value and direct and indirect jobs that are not easily relocated.
France and other European countries have understood and are progressively defining swimming lanes, so to speak, to safeguard what must go to industry and to other critical services of our societies.
It is urgent to write down what we want to do with AI, before this wave finishes destroying Europe’s initial ambition.
You lament a billion per day of digital dependency on the United States. Would restraining data centers in Europe risk worsening this problem?
There are several questions.
The first is what we want to do with this technology. We cannot stop a technology that disrupts our access to knowledge so profoundly, even if it contributes to this acceleration of the world’s times that is a disaster for ecological, biological and climatic balances.
So we will have trouble preventing it. We must still eventually find a way to direct its use toward topics that best serve the common good and thus reduce its grip on our societies.
The second is to ensure that what is built and developed on European soil serves Europeans first. We have already built, through GDPR and all rules on digital services and the protection of European citizens’ data, the foundations of a legal framework to protect our data and their uses. We will also need to ensure that data processing and storage assets located in Europe have priority rules for the use and preservation of European data.
Finally, we must find our place in the value chain of this wave, in the infrastructures but also at the end of the chain. It’s not only a matter of competitiveness; it is also because if the value rents of this technology remain in the hands of the four or five main American or Chinese leaders (we have not yet seen everything this will cost each of us), then the gains it enables cannot be redistributed and notably cannot support the other central movement affecting our developed economies, the personal services economy from which we expect growth of several million jobs, critical for fragile and dependent people, to maintain relationships in our increasingly digital societies, and whose productivity levers, by nature weak, will be largely captured by digital. This is part of our societal sovereignty in the medium term.
Does this argument apply to capital as well?
It is even more pronounced. In the United States, the construction of data centers and the associated equipment accounted for about 0.8% of GDP in the first quarter of 2026, and all IT infrastructure about 1.5%, versus 0.7% on average from 2015 to 2022. Several estimates attribute between one third and half of American growth to these expenditures. It has become the leading engine of private investment.
In Europe, we do not have the same amount of capital or the same amount of available electricity. What we direct toward data centers, we simply do not direct elsewhere.
You describe the first engine of American growth. Why should Europe stay on the sidelines? Do you think it’s a bubble?
The future will tell the size of the bubble, because there is inevitably, as with any large systemic technological wave, a bubble effect, the bursting of which would bring asset values back to a level that makes them competitive. So it is likely and we will see.
Europe cannot stay on the sidelines, but sacrificing its overarching ambition for its future economic model without properly characterizing what serves our needs and what serves others’ needs would be a very grave mistake.
So you’re asking to choose?
I am asking that we choose explicitly rather than implicitly. The Industrial Accelerator Act is a very good way to place industry at the center, and I would not want us to sabotage it again with artificial intelligence.
Let’s stop trading the prize for the shadow. We have been doing this for 20 years, yearly chasing the next bubble, instead of staying the course.
Saying that we need some artificial intelligence is fair, but first we must settle the issue of how we treat our data, which is not yet secured.
Then, we must precisely define what is destined for AI and what is destined for the rest of industry. Otherwise, industry will not restart.
Concretely, what mechanism do you propose? Priority criteria for grid connection? Quotas? A differentiated price? And who will arbitrate? Brussels, the member states, or the grid operators?
We need swimming lanes, we need default sovereignty rules over data centers built on our soil so that no foreign power can abruptly cut off access to data and the functioning of our systems when they are located on our territories.
We must, in a sense, re-anchor the digital realm in the reality of our territories.
And if this economy can pay part of the cost of our decarbonization by financing its share of strengthening our production, transport and distribution capacities for our electrons, that will have been virtuous rather than a black hole of our hopes.
Then, regarding governance, states should be large and responsible enough to mandate network operators and national regulatory authorities to find the right balance.
But at some point Europe will have to call an end to the game.
That said, Spain, Denmark, Norway and other countries have already begun processes to slow the dynamics. France has started very recently.
Finally, why European indicators? Because on all these matters, if we could reason first at the European level before reasoning nationally, we would make a real giant leap forward.