There is an elephant in the room that no one is talking about. But the AI is going to wake it up and it will become an elephant in a china shop. Technology deflation is coming, but what is it really?
Artificial Intelligence is going to boost productivity and, with it, overall prosperity. The phrase may have its nuances, but it is commonly accepted. If machines do more work in less time, the economy should grow faster and we should all live better. However, when one looks at historical productivity data in advanced economies, the evolution is puzzling: they grow little, sometimes not at all, just in the period of greatest digitization in history. Is this going to continue to be the case?
Before answering the question of what AI will do to the economy, it is worth clarifying a basic assumption. Productivity is not the same as welfare, nor is it even automatically synonymous with GDP growth. Productivity measures how many units of value are produced per hour worked, but “value” is measured in money. And that is where the problem begins.
Zero marginal cost and price erosion
AI takes to the extreme a dynamic we already knew with software and the Internet: zero marginal cost. Training a model is expensive, but once trained, generating a report, a design, a text or an additional line of code costs practically nothing. From an economic point of view, this has considerable consequences.
In open and highly competitive digital markets, price tends to approach marginal cost. If it costs almost zero to produce an additional unit, the price also tends to zero. This is not theory: we have been seeing this for years in digital services, online content and e-commerce, where players like Amazon have pushed competitive pressure to the limit.
In non-digital markets, the effect is smaller (you still have to grow wheat to make bread) but equally appreciable.
The result is paradoxical. There are more services, more functionalities and more “useful things” than ever, but each of them generates less revenue than its analog equivalent. One software replaces one machine, one subscription replaces dozens of physical services, one algorithm does the work of several qualified profiles. From the user’s point of view, it’s fantastic. From an aggregate billing point of view, not so much.
This is where what I have once called the Matrix effect comes in: we replace expensive objects, processes and services with cheap digital solutions that bring a lot of use, but little money into economic circulation. Perceived well-being can rise while GDP stagnates.
Productivity without wages: an unstable equation
The discussion often remains on the surface: “if AI increases productivity, wages will eventually go up”. That phrase encapsulates a historical assumption that is no longer valid. For decades, productivity and wages advanced more or less in parallel. Since the 1980s, that relationship has been breaking down.
AI accelerates this rupture for a simple reason: it replaces wage labor with technological capital. A company can produce the same, or more, with fewer people on the payroll. Productivity per worker rises, but the wage bill does not necessarily rise. And here we enter a field that is rarely addressed in technological debates: the actual functioning of the monetary system.
Money, debt and the role of credit
In modern economies, money does not appear by magic, nor is it mostly printed by the central bank. In practice, most money is created when banks grant credit (more than 90% of the money supply or M3). A mortgage, a loan to a company or a consumer loan generates new money in the system. When that credit is repaid, the money disappears.
This mechanism works reasonably well as long as there are expectations of stable future income. A worker with predictable employment goes into debt to buy a house. A company with growing sales takes out credit to invest. But what happens when AI starts sending people out of work or, at the very least, introduces a structural sense of job insecurity?
If wages stagnate or if a growing part of the population fears being replaced by algorithms, the incentive to take on debt is reduced. Less new credit means less new money. At the same time, old credit keeps being paid back, destroying money. The system starts to run out of fuel.
It is not a classic financial crisis, but something more silent: a monetary contraction derived from extreme efficiency. A system designed to grow through debt encounters a technology that reduces the need for human labor… and thus the basis on which that debt was built.
Technological deflation: abundance without prices
Thus, AI exerts a very powerful deflationary pressure. If production is cheaper and faster, prices tend to fall. If wages do not rise at the same rate, or even fall in relative terms, consumption suffers.
This type of deflation does not resemble that of a traditional crisis. There is no shortage, but abundance. The problem is not that there is a lack of goods or services, but that there is a lack of money circulating to pay for them. It is a technological deflation, difficult to combat with the classic tools of monetary policy.
When interest rates stop working
For years, the standard response to any economic slowdown has been to lower interest rates. But when rates are already close to zero and companies and households still do not want to take on debt, this lever loses its effectiveness. It is at that point that measures that once seemed exceptional come into play.
Quantitative easing (QE) consists of the central bank creating money to buy financial assets. The problem is well known: much of this money is trapped in financial and real estate markets, inflating asset prices, but without reaching the daily economy. We saw this with the large quantitative easing in Europe, the US and Japan since 2014.
That’s why ideas like helicopter money are starting to sound not so bad. helicopter moneythat is, direct transfers to citizens without going through the banking channel. If money is not created via credit because no one wants to get into debt, it can only be created and distributed directly.
Central bank digital currencies (CBDCs), such as the European Union’s Digital Euro, are a very useful tool for such policies. They allow for direct, programmable transfers, potentially linked to basic income schemes. However, they could make commercial banks dispensable and would give the state inordinate power over citizens and companies, which is why they generate no small amount of controversy.
AI, delivery and the real bottleneck
At this point, a key distinction should be made explicit. The problem is not that AI does not generate value. It does, and generates a lot of it. The problem is how that value is distributed in a system designed around wages and credit.
If machines produce almost for free, someone has to be given purchasing power to keep the economy going. Historically, that someone has been the wage worker. If that mechanism weakens, another one has to be invented. It is not an ideological question, but an accounting one.
This is where many technological narratives fall short. They talk about efficiency, innovation and disruption, but avoid the hard core: the monetary and social architecture on which modern capitalism rests. Without an adaptation of that architecture, technological deflation is not a temporary anomaly, but a structural trend.
Conclusion: abundance without story
AI confronts us with an unprecedented situation. For the first time, we have technologies capable of producing enormous amounts of value with very little human labor. From a technical point of view, this is a success. From an economic and social point of view, it poses tremendous tensions.
There are very few of us who have ever talked about technological deflation and its effects on the economy, and it certainly cannot be said that we have changed the course of history. It is a term that is hardly mentioned, nor is there any other similar term that expresses the problem. However, we are heading headlong into brutal systemic change thanks to AI and it will no longer be possible to ignore this huge elephant in the room.
Technological deflation is not an inevitable catastrophe, but it is a sign that the traditional link between work, wages, credit and money is weakening. The challenge is not to slow down AI, but to rethink the mechanisms of sharing in a world where scarcity is no longer in production, but in the distribution of purchasing power.
Perhaps, as a final provocation, it is necessary to accept that in the future a relevant part of the money will have to “fall from the sky”. Not out of generosity, but because the economic physics of an ultra-efficient system no longer rests solely on the soil of traditional employment. The serious debate has barely begun.
