Transcription
On May 6, 2026, in Texas, Larry Fink, the head of BlackRock, calmly explained something that should nevertheless make us prick up our ears. The financing of AI infrastructure should involve the absorption of funds from insurance, pension funds, and savings accounts of Americans in general. And if not, what happens? Well, if not, he explains that China will do it, which will result in the fall of American primacy in technology. And honestly, well, he's not necessarily wrong. China has the capacity to deploy its population's capital thanks to its political system closely intertwined with its economy. Where Beijing can direct the savings of 1.4 billion inhabitants through a single state banking channel, Washington does not have this leverage. In the United States, savings are private and dispersed, and they are largely stored in bank accounts and products considered risk-free, thus ultimately financing a good part of the American debt. To mobilize it, there is no decree, there is a price, and that price is the market interest rate. Draining savings, insurance, and pension funds towards AI infrastructure mechanically causes rates to rise. Now, higher rates mean a state pays more for its debts, and it means a real economy, typically that of SMEs, is suffocated. In other words, the plan that Fink describes with serenity looks on paper like a large, forced deleveraging movement for the American state and many small and medium-sized businesses. So, why does the man who manages more money than there are stars in the sky defend an idea that seems to turn against the one implementing it? As always, the right method can be summed up in one sentence: "Show me the incentive." I will show you the result. We should not believe that Larry Fink is improvising this statement to create a sensation, because it is part of a perfectly coherent sequence. A few days earlier, he stated that computing power would become a new asset class that would be valued in trillions of dollars. So, the financing of AI infrastructure and everything that follows financially. How to create a new asset class? That is exactly his current topic. So, okay, let's start from this principle. I know that somewhere right now, "trillion is the new black" in the USA. Fink himself estimates the investment needs for infrastructure in the United States alone for the decade at over 10,000 billion, and up to 68,000 billion globally by 2040. But it remains true that to inject trillions of dollars into infrastructure, they will have to be taken from elsewhere, because the American economy, or even the global economy, will not create trillions of dollars in surpluses out of thin air in 6 months. Now, short-term capital is a fixed quantity. It is not created by decree; it is moved, and any movement of capital has a cost for the one it is moved from. So, what exactly is he saying? Let's take his words again, because he is a much better diplomat than I am. Fink is not saying that savings will be confiscated. He says the money will come from savings accounts, retirement accounts, and insurance companies. By explaining that keeping your money in a bank account means not growing with the American economy. Translation: Americans must be convinced or incentivized to take their money out of risk-free assets and direct it towards infrastructure. He speaks of reorienting funds from insurance, pension funds, and American savings towards infrastructure financing. So, okay, but let's look at what this money is used for right now. Firstly, financing sovereign debts. Yes, when you entrust the management of your money to an insurer or a pension fund, a good portion ends up financing the debt of various states deemed risk-free. Secondly, for the portion of these funds deposited in short-term treasury products, it is, roughly speaking, the short-term liquidity base upon which the entire credit system relies. It finances treasury bills, the repo market, commercial paper. In other words, the very short-term base from which all other rates are built. Concretely, if you siphon off these funds with higher interest rates elsewhere, this base becomes more expensive, and all credit becomes more expensive with it. And it is not possible to maintain cheap loans if the short-term liquidity base goes to earn returns in another sector that pays better with the same estimated level of risk. So, this means in such a scenario that interest rates will structurally increase for anyone who wants access to credit, whatever it may be. This also means a higher cost of access to real estate, and therefore downward pressure on real estate prices. But that's not the most bothersome part. The problem will mainly come from the fabric of SMEs, often called the real economy, which often struggles to borrow beyond 5%. What I have just described, economists have named it before me: the crowding-out effect. Massive private investment in AI, by monopolizing available capital, crowds out the state and less profitable companies from access to credit. Nothing new in principle. The novelty is the scale and the identity of the one who crowds out everyone else. So, what is the plan? Because concretely, doing this means both putting the American state in difficulty regarding its deficits and imposing a recession on all companies unable to beat the new benchmark return. And this benchmark return is precisely what AI infrastructure will set. When money can easily seek around 8% to 10% in AI infrastructure, no activity remunerating capital below this threshold will be favored by lenders. The cost of capital for the entire economy shifts upwards based on what is paid, and this profitability of 8% to 10% in AI infrastructure will have to be supported by AI's ability to generate economic growth on its own. Now, if Fink is talking about it, know that the White House is aware, and all these important people already have a plan in mind. Here, in my opinion, is how the USA intends to emerge as the big winner. First of all, keep in mind that neither China nor the USA has the luxury of giving up this AI race, because the risk of being downgraded is far too great. Keep in mind that AI is not just a technological innovation; it is the emergence of a new intelligence on Earth. To give up AI is to accept being cut down in terms of international competition. On that basis, and still in the spirit of American leaders, political and economic, I think that for them, there is therefore no alternative. And that is exactly what Fink said in Texas, almost word for word. If we don't invest these trillions of dollars, China will become the world leader. It's not an option on the table among others; it's an obligation in their minds. So let's follow Fink's thinking and siphon the trillions of dollars needed for AI development for the benefit of the USA from pension funds, insurance, and the banking system in general. Remember, as I told you, AI remunerates cash at around 10% for projects carried by large groups and around 14% for smaller, inherently riskier companies. Finally, small companies, let's say SMEs. It is therefore through an increase in proposed interest rates that capital will be attracted, to the great delight of savers. So, in this case, the idea, and this is the core of the plan, is that the American state will finance its deficits not through new debt, since that is too expensive, but through accelerated economic growth. Fink did not let this sentence slip by chance. He explicitly linked the 3% growth he calls for to the deficit that the federal government must finance. And it is perfectly logical. If the taxable base grows faster, the state can then reduce its debt without borrowing more. Not by reducing the numerator, the debt, but by increasing the denominator, the GDP. So, I hear you asking why economic growth would accelerate. Well, honestly, we have to be honest. Do we observe this boom today, this decoupling between wealth production and employment, for example? Because that's what the whole reasoning is based on. So, what do we observe today? Well, that American growth is holding at levels that are difficult to explain by classical economic drivers alone. The third quarter of 2025 showed an annualized growth of around 4.5%, driven by a surge in business investment massively oriented towards AI. At the same time, Goldman Sachs documents that over 12 months, AI has already destroyed around 16,000 net jobs per month in the United States, focusing on junior white-collar positions. So, what can we deduce from these facts? Because yes, these are very real facts, but still, 16,000 jobs per month in the United States is not an avalanche. Well, AI seems to have begun to produce a decoupling between economic growth and employment as early as the end of 2025. In other words, the country would produce more wealth with less human labor. More and more signals, including studies from Stanford on the exposure of professions to automation, point in this direction. It would therefore not be a projection for the future, but the reality of 2026. I say "seems" and "signals" because correlation is not proof, and one or two quarters of strong growth do not make an economic law. But if the trend is confirmed, it still means, within the scenario of American elites, that the increase in tax revenues based on a more productive economy will take over from financing. We shall see. But in any case, this is the plan, and it's not for 2040; it's happening today. And note the almost ironic beauty of it. It is the same AI that drains capital from the real economy and is supposed to save state accounts through productivity gains. The poison and the antidote come from the same vial, somewhere. Consequently, in this scenario, it is absolutely not the time to buy long-term American bonds, because I remind you that when rates rise, the value of old bonds falls, as there are better ones to buy on the market. But wait, you haven't seen everything. There remains the thorny problem of financing the overwhelming majority of SMEs that will no longer have access to credit at rates around 10%. The principle is that the growth of the AI sector will occur either by absorbing the most successful companies from those that don't keep up, or by adapting certain companies that will adapt, that will transform themselves to benefit from the multiplication of their productivity thanks to AI. Well, in a grand Schumpeterian cycle of creative destruction, these new companies that will transform will have to compensate more broadly for the decline of the economic fabric that will be suffering. Yes, all of this is still a gamble, and in a gamble, you have two possible worlds. The optimistic side where you win, you have a creative destruction taking place as we have seen many times in the past, and entire sectors of the economy will disappear in favor of more productive sectors without the overall wealth collapsing. Quite the contrary. But the other possible world is the timing, the calendar. An SME that closes today does not become an AI laboratory tomorrow. And a 50-year-old accountant whose job is automated does not become a machine learning engineer overnight. Between the decline of the old world and the rise of the new, there is a gap, assuming this new world comes to pass. And it is in this gap that bankruptcies, transitional unemployment, and anger reside. The theory may be correct over 30 years and catastrophic over 5 years, because the other parameter is that the duration over which this transformation of the economic fabric could occur has no historical equivalent. Now, on a human scale, the capacity of the economic fabric to adapt for the people who constitute it relies heavily on temporality. If you have 20 years, 15 years, or 10 years ahead of you, it's not the same as if you only have one year. Well, I'm willing to accept that, but still, a big question remains. We will have to find ways to manage social order cohesion, because if you disempower the companies that provide 60% or 80% of the country's employment, there will be a lot of grumbling. And it is at this point that the big names in AI arrive at the magic answer that closes the debate. Don't worry, you will have a universal income that will allow you to abandon all your worries to enjoy life and go mushroom picking every day. Personally, I doubt it will be that simple, but in fact, I am convinced of it. And yes, fundamentally, work is what occupies us most in our lives, and above all, it is the primary key to wealth distribution in our modern societies. Work also gives us our social position, a sense of purpose, a structure, a reason to get up in the morning. So, sweeping all that aside with a universal income seems entirely speculative and dangerous to me. Giving money to people whose function has been taken away is not the same as giving them back a place in society. We can feed a people, but leaving them idle for a long time has always been a mistake harshly punished in history, suggesting that all societies of pure spectators maintained rarely end well. So, where does this lead us? If we put all the pieces back on the table, the plan that Fink described with his disconcerting calm is not an investment plan. It is a project of reallocation of national capital on a continental scale. We drain savings from risk-free assets towards AI. In doing so, we raise rates. We strangle the state and the real economy in the process, and we bet that the productivity gains from this same AI will grow the economy fast enough to repair public accounts and compensate for the damage from below. If we really go down this path, we can find coherence and even elegance. But at each stage, it relies on a gamble that has not yet been won. That the growth-employment decoupling is real and sustainable. That new growth will compensate for the old world waiting, and that social order will withstand the shock of the transition. Let me tell you, as you can see, there are still two or three steps before we get there. A whole program with a very large number of unknowns, but which remains the path the world is currently embarking on, whether we like it or not. I don't know how all this will evolve, but what is certain is that we are living in a particularly interesting time.