What is your occupation?

I never know how to answer this common question in admin forms. I don’t have an official occupation. The closest to it is an involvement in a French non-profit called Institut Présaje. Présaje is an acronym where the "P" stands for prospective, the "J" for justice and the "E" for economy. Over the years, Présaje has done an amazing job at tackling prospective issues. As early as 2019, we started organizing conferences on artificial intelligence applied to legal systems. Présaje backed research on AI regulations long before ChatGPT was even released.

The annual conferences on AI and Law attract more people each year which may be a sign that AI is no longer a prospective issue. So, in early 2025, I was tasked to think about topics for a new cycle on economic issues. Without sharing the detailed roadmap, these thoughts constituted good material for an essay.

This type of exercise implies looking 10, 20, or 30 years into the future. Perhaps several generations. Finding long-term dynamics that are already well underway is not hard. They are everywhere, albeit they are often ignored - more or less knowingly - as they will unfold way beyond the typical horizon of mainstream media (few weeks) and politicians (the next election). Climate change is on everyone's radar, but there is more than one elephant in the room.

Aging Population

The aging of the population is the archetype of an unpleasant truth for mainstream media readers and for voters, which does not encourage anyone to put it on the table. Age pyramids, birth and fertility indicators, and celibacy rates in all developed countries indicate that the situation is worrying. Among these countries, some managed to stay on-track until about 10 years ago but they eventually fell into line.

It is unclear if people have simply looked away during the past 15 years or fooled themselves into believing that it was just temporary and illusory ; that couples were simply postponing births and that children would eventually come. Not only has the situation never normalized, but it continues to worsen. Apart from sub-Saharan Africa, all countries in the world are affected, to varying degrees. And no political solution seems to bear fruit. We now have enough data to infer the entirely predictable consequences in one or two generations: a significant portion of humanity appears destined to age and then decline.

The economic issues associated with aging populations go beyond the mere question of pension financing. This undoubtedly raises systemic questions, in particular in countries with a pay-as-you-go pension system like France but, in the spirit of prospective research, we ought to look beyond the active/retired dichotomy and consider instead what happens when an increasing share of the population loses its autonomy and becomes dependent.

As manpower decreases, societies will have to come up with innovative solutions to maximize the economic value they can extract from their retired-but-still-autonomous members. There is no doubt that active populations will self-organize to leverage the energy and experience of seniors. It seems in everyone’s best interest that seniors continue to consume and to contribute to -or take part in- projects led by active people. Even more so if stimulation turns out to have a positive impact for their mental and physical health.

Will this process need guidance and require oversight? It is not merely a macroeconomic question but it is also a human rights question: dependence often impairs dignity and dignity is the cornerstone of human rights.

All human beings are born free and equal in dignity and rights. They are endowed with reason and conscience and should act towards one another in a spirit of brotherhood.

Universal Declaration of Human Rights, Article 1

If the objective is to maximize human dignity, the demographic situation is alarming. In Western countries, the proportion of childless women is trending towards 25%. Men are obviously also affected by this trend. Who will take care of these people once they become dependent? Today, this very question typically arises in the context of family disputes. In a generation or two, it will haunt many of those who have neither fed nor changed the diapers of children who could return the favor.

And it is not only about child-free families.

Is it reasonable to assume that an ‘only child’ can bear the mental and financial burden of the dependence of two parents? To maximize human dignity, we must assume that they will not. The burden of dependence will therefore fall on the community, on the State. However, demographic pyramids show that many nations will not be able to rely solely on their working-age population to directly provide care for -or indirectly finance the support of- this dependent demographic.

Techno-optimists will argue that robotics will reach such levels of sophistication by then that there will be no shortage of manpower. It is possible. But it does not alleviate the problem of State financing in a context of a significant decrease in the active population.

Sovereign Debt

There are two main financing routes for a State: taxation and inflation.

An active population that consumes, invests and, above all, borrows, has an inflationary effect. An aging population may temporarily consume more as retirees have more free time to spend their money, but they borrow less. A dependent population does not borrow at all and does not consume more than its bare necessities.

A decline in active population exerts a deflationary pressure on the economy, and this effect is amplified if the inactive population becomes dependent. However, the deflation scenario is rarely considered by economists. “Deflation” is the economics-equivalent of “demographics” in politics: a d-word that must never be spoken.

People prefer talking about inflation, mostly to point out its negative effects: if you have cash-savings or any assets that are not indexed on inflation, the purchasing power of your assets decreases over time. And your income may only be adjusted for inflation with a lag. High inflation is bad for that reason.

But if you are in debt with a long term horizon, e.g. young professionals with student debt or countries that issue bonds, you must also consider inflation from the perspective of your liabilities instead of your assets. If a country borrows in a currency and the relative value of that currency decreases due to inflation, then the relative value of its debt decreases. However, in the meantime, tax revenues increase with inflation. For that reason, inflation is the miracle solution to budget deficits. That's why the base scenario of any economist is low inflation but with an absolute floor at 0.

Now let's reverse this reasoning with deflation: on the one hand, the relative value of the debt increases, and on the other hand, tax revenues decrease. The economic disaster is not indebtedness on its own; it is the combination of indebtedness and lack of inflation.

Tax revenues are an important variable in this complex equation. Let’s break down the main sources to analyze the impact of the demographic trends: income tax, VAT, corporate tax, inheritance tax.

Is there a scenario where the decrease in the active population, coupled with an increase in the dependent population, would not lead to deflation and a decrease in tax revenues?

It is conceivable in the context of a technological revolution. We touched on the idea when we mentioned that robots could attend to dependent people’s needs. We may indeed be at the dawn of a new industrial revolution with the emergence of artificial intelligence, from which robotics will necessarily follow¹. So let's explore the techno-optimist angle.

The inflationary pressure that generally accompanies any technological revolution is due to increased productivity. Productivity gains spread through the economy pursuant to the Baumol effect. While this effect is well documented, a decline of the working-age population is unprecedented in the history of industrial revolutions - at least since they are studied. Inflation at least equal to zero seems to imply, at a bare minimum, the following equilibrium: technological progress must enable higher volumes and products/services of better quality, but human labor must retain a critical role, so substantially all monetary gains from increased productivity are “reinvested” into employees' compensations. Workers are just paid much more. If the increase in salaries offsets the decrease in labor volume, the implication for income tax is rather trivial.

It is much less so for VAT and corporate tax because the supply, enabled by technology, must meet a demand. However, structurally, most end-markets will contract in volume due to demographics.

The hypothesis of equilibrium can be further called into question as it largely relies on the assumption that value-add human labor remains absolutely necessary. The anguish of neo-Luddites is a scenario where robots and artificial intelligence directly replace humans or, at the very least, level all salaries downwards. Companies might make record profits, but would this be enough for the State to tax them to recoup their loss on income taxes? The techno-optimist view is that, if the cost of intelligence decreases that much, then the price of all goods and services will also decrease drastically and everything will become accessible to everyone at a marginal cost (because intelligence is the added value). This science fiction scenario² is consumers’ dream, but it is a disaster for countries’ finances.

Against this gloomy backdrop, States could find comfort in the prospect of higher inheritance tax revenues. After all, the population decline phenomenon is the consequence of people’s passing and, when that happens, the State may be entitled to a share of the value transferred from one generation to another. Not everyone is concerned though; only those whose assets exceed a certain threshold per descendant are. It nevertheless constitutes a windfall for the State. A windfall that could be described as providential if we take into account the relative wealth of the baby boomers compared to their parents. And for generations X, Y and Z, wealth transfer will be spread over fewer, more remote, descendants which means that thresholds can be more easily met and the average tax rates can be higher all else being equal.

We can already predict that inheritance tax will soon become a much more important political issue than income tax. After baby boomers retire, they will remain influential voters, by their number and by their turnout rate in the polls³. So when their focus shifts to inheritance tax, so will politicians. Barring any changes in policy, it seems possible to model the value to which a State could be entitled. And to put the simulation results in perspective with sovereign debt. But again, the impact of inflation on household assets would be a critical variable in the calculation.

In all possible scenarios, there is an interesting case where the State's debt is largely held by its own citizens. Provided that these citizens’ assets exceed the minimum threshold for paying inheritance tax, the State will levy a value indexed to its own debt. Debt held by citizens is, in a way, cheaper than that held by foreign companies or individuals⁴.

Before posting this, the news-flow reminded me of a potential source of revenue that I had omitted and that is coming back into fashion: tariffs. Donald Trump floated it numerous times in 2024 and 2025 as an alternative to other sources. I have reservations on that specific claim but I can't just discard the idea. In fact, tariffs barely fit into the above demonstration which justifies spending more than a minute thinking about their implications. At first, tariffs look like an intriguing silver-bullet: not only do they drive inflation but they finance the State in the process. This can’t be a long-term fix, in particular when the countries imposing tariffs lack manpower, as it could undermine the overall competitiveness of domestic industries. But it certainly constitutes a quirky tool at the bottom of the fiscal toolbox.

For the avoidance of doubt, fighting deflation is not the reason why tariffs are back in the spotlight in 2025: they are primarily an effective economic weapon against countries that sell products and services to Americans. It is one of many weapons currently used in a protracted war.

The New American and Chinese Economic Weapons

Regardless of demographic dynamics, we are in for a new Cold War where two superpowers, the United States and China, will clash - hopefully without direct military confrontations. Foreign aid and tariffs are used for leverage in negotiations, and the economic arsenal is growing. On the one hand, the emergence of new artificial intelligence and robotics technologies and, on the other hand, a new space race. The new levers of power are both a matter of influence and a necessary condition for supremacy.

Both sides have started to convert their technological leadership into an economic weapon. The United States has the best private AI systems - which are made available to the Government^5,6 - but the leadership in software cannot be defended because each innovation is replicated by other countries before the bureaucracy even has had time to digest it. To turn Artificial intelligence into an economic weapon, the US has to use two strategies: (i) keeping the most advanced models secret for as long as possible and (ii) restricting access to hardware.

Home to the three cloud leaders, Google, Amazon, and Microsoft, the United States has, de facto, control over the computing power needed to run the best AI models. They have done everything, for several years, to prevent China from building data centers with the best graphics chips (GPUs). Now, the United States is using computing power like they use their currency: as a regulatory tool with global reach, delegating compliance control to operators, and as a vector of sanctions against States that are not commercial and political partners. This has repercussions on the domestic politics of the countries concerned. For example, one of Joe Biden's last executive orders in January 2025 contradicted the free circulation of goods within the EU - one of the most fundamental principles for the Union.

Notwithstanding their leadership in software and infrastructure, the United States do not control the entire value chain. Far from it. The designers of chips and the developers of software to use chips are American. Nvidia is the figurehead of this very hot industry where the historical players, Intel and AMD, have had difficulties catching the wave. They will eventually do it as the Government pushes them. Google develops their own chips (TPUs). In the vast majority of cases, these chips are manufactured in Taiwan - constantly under the threat of China. Also, the chips are manufactured with machines of Dutch origin, and require raw materials that are very often controlled by… China. Going further, China controls most of the resources to provide the non-fossil energy that data centers will need so much if AI establishes itself as an essential technology.

The US is therefore engaged in a frantic race to patch the holes in their coverage of the AI value chain and, trying to slow down China as much as possible, while China is trying to design and build their own chips⁷. The US is still undeniably ahead, but China is progressing at an impressive pace considering the hardware constraints to which it is subjected. China does not just replicate, it innovates to circumvent constraints. China is far ahead of Europe. Fortunately for Europeans, the Chinese publicly release their models and their research – which the Americans do not do, apart from Meta and Google to a lesser extent. China draws enormous soft power from its technology industry.

The geopolitical dimension of AI is visible everywhere. I wrote about this in 2023, and it was more recently exemplified by the release of a model by a Chinese company called Deepseek. That model had capabilities equivalent to what American labs offered to the general public at the time. The model was made available under a permissible license (MIT), and Deepseek basically explained how they overcame the hardware constraints imposed by the US. The release happened on the day of Donald Trump's inauguration, and was an occasion for the Chinese to show the CEO of Deepseek on official television for the first time, in an interview with the number 2 of the Chinese Government. Even if Deepseek is originally a subsidiary of a Chinese hedge fund with no link to the Government, you would not be blamed for seeing a political message in it.

Deepseek went relatively unnoticed outside educated circles until January 2025 but it was an established player⁸. They had started a terrible price war on APIs several months earlier (up to 90% cheaper than OpenAI). It is a war of attrition that only the American or Chinese Tech giants seem able to win - Google is the most competitive of all. The odds that Deepseek could make it through the price war were very high, but they received Government backing after their engineering prowess. As part of their recent release, they also launched an application that was the most downloaded in the United States at one point, with the consequences that we know in terms of data security and privacy.

Putting aside the distraction of individual pieces of news on specific companies, there is a secular trend that neither the baby boomers nor the millennials seem to perceive: China is today at the forefront of technology, which Generation Z recognizes, as evidenced by its significant use of Chinese technologies (gadgets, social networks, marketplaces…) despite legitimate concerns about potential dangers.

There are certain segments of AI where China is far ahead of the US – at least on what is shared with the general public. This is the case for embedding models (essential for semantic research) or certain modalities like vision (input) and video (output). Only Google seems able to compete with Chinese labs. And all this has repercussions for downstream use cases of AI models, such as robotics – civil or military – where China seems impossible to catch up with.

Space

The new Cold War will have strikingly similar dynamics to the previous one. This is particularly true for the space race, which has many economic implications.

The general awareness of Chinese capabilities⁹ is very low.

• Since 2020, China has completed its Beidou navigation satellite network, making it totally independent of foreign solutions, notably the American GPS network
• China landed a rover on Mars in 2021
• China is the only country that has a "sovereign" space station. Tiangong 3 only has 3 modules (compared to 16 for the ISS station operated by Europe, Japan, Russia, the United States, and Canada), but no other nation can maintain even a 3-module station on its own. The station should be in operations until 2037, 7 years after ISS.

In the US, the new space race does not emanate directly from an institution like NASA or the Government, but from an individual - American by adoption - : Elon Musk. Jeff Bezos, another billionaire, may try to challenge Musk in the medium term with his own company, Blue Origin, but Musk’s SpaceX is for now unrivaled in the satellite launch market. His reusable rockets, which made people smile in Europe when Arianespace dominated, have established themselves as the new standard.

Musk accumulates commercial and government contracts. And he also launches his own load to orbit. Starlink, his network of about 7,000 telecommunication satellites, has become a major geopolitical stake in complete indifference. Why is it so important?

• The honeymoon period with Donald Trump may not last but Elon Musk is now close to the US Government. So close that he even has an official role in the administration.
• Starlink satellites allow communication virtually anywhere in the world without going through official communication infrastructures. They therefore allow one to bypass potential national or regional regulations.
• The satellites also allow communication where there is no infrastructure, or where infrastructure is out of order. As in the case of armed conflicts. This geopolitical influence materialized in 2023 in Ukraine, where Elon Musk had direct control over the Ukrainian army's access to Starlink. There is little doubt that some people do not particularly welcome the idea that an individual – close to the US Government or not – has so much power.

Starlink represents nearly two-thirds of the satellites in orbit around the Earth. Any incident, could snowball as debris can in turn destroy other satellites - Starlink or not -, and therefore affect a large number of services on Earth. Given the geopolitical stakes, an incident would not necessarily mean an accident: in 2007, China deliberately destroyed one of its meteorological satellites, two-meter-wide, with a missile launched at 29,000 km/h, which Wired⁹ described as a military achievement equivalent to destroying a revolver bullet with another bullet. And in early January, Chinese scientists announced that they had simulated a hunt of Starlink satellites. By the end of the 21^st century, what people refer to when they mention "Star Wars" could be very different from today's meaning.

Chinese Commercial Dominance Makes Americans Nervous and Brings Europeans to Their Knees

Future technological developments and their economic and geopolitical impacts can only be speculative. But we can already see the polarization of the economic world, with a significant decline in Europe in terms of trade.

The Chinese trade surplus reached one trillion dollars in 2024 - a figure never reached by any other nation before. It reveals the extent of Chinese dominance. This dominance is no longer limited to low-end consumer products but now extends to high-end equipment. Countries that had bet everything on a cutting-edge industries, like Germany, are suffering the full force of competition from China which, not only curbed imports, but now exports products at prices so low that they are competitive everywhere despite tariffs. Imports of German products into China thus fell by 11% in 2024 (while exports of Chinese products to Germany increased by 6.5%).

Industry-heavy Germany has an inherent inertia which means that the country is entering a very difficult cycle. But it is all of old Europe that has to reinvent itself because the prospects are very bleak. Italy, due to its demographic situation, the United Kingdom, due to its isolationism. France's burden seems the lightest, despite its volatile political situation.

For Europe, the convergence of these long-term trends - demographic decline, sovereign debt challenges, and the US-China technological Cold War - paints a stark picture of the decades ahead. Europe finds itself caught between these tectonic forces but is no longer the strategic priority it once was during the first Cold War. While China and the United States vie for technological and economic supremacy through AI, space technology, and trade dominance, the Old Continent must confront not only its diminishing geopolitical relevance but also its rapidly aging population. These challenges are not insurmountable, but they require a level of long-term thinking and continental unity that has proven elusive thus far. For China, a halving of the population in the next 50 years will necessarily lead to internal unrest irrespective of the geopolitical situation.

---

1 See Domestic Machines for a deep dive on technology and human lives

2 my personal opinion, for what it’s worth, is that the perfect strom under the science fiction scenario has a low-probability

3 this is a key reason why the topic of the aging population is a taboo

4 somewhat also true for debt held by domestic institutional investors if corporate taxes catch interests earned by the investors

5 https://investors.palantir.com/news-details/2024/Anthropic-and-Palantir-Partner-to-Bring-Claude-AI-Models-to-AWS-for-U.S.-Government-Intelligence-and-Defense-Operations

6 https://openai.com/global-affairs/introducing-chatgpt-gov/

7 information about Chinese breakthroughs in semi-conductors should generally be taken with a pinch of salt but there is evidence that they will catch-up in this decade. Nvidia GPUs remain the go-to solution to train AI models but local chips, notably by Huawei's, are now used for inference.

8 Deepseek had state of the art models for math and coding in 2024. They were mentioned in all Not_too_fast blog posts of the second half of the year

9 https://www.wired.com/story/china-space-race/