After money

In his interview with The Economist, Elon Musk is not merely talking about artificial intelligence. He offers a complete economic theory: machines will do every job, supply will become nearly unlimited, work will become optional and money will eventually lose its importance.

The idea deserves better than ridicule. History even partly supports it: whenever we learned to reproduce an object at scale, the portion of a life required to buy it fell. But Musk moves too quickly from “a machine can produce it” to “everyone can obtain what they want”. Between those two claims sit ownership, energy, collective rules and human beings themselves.

Before imagining the disappearance of money, let us ask a more concrete question: how long do we still have to work to obtain what we consume?

01 · The price of our time
What can be copied gets cheaper. What remains scarce resists.

This article uses a deliberately French benchmark: the SMIC, France’s statutory minimum wage. It does not represent every household or the average standard of living. It tracks a narrower question: how long must someone at the bottom of the wage scale work to access the same goods? Directly comparing francs in 1950 with euros in 2025 would tell us far less.

The measure is imperfect. It uses the gross minimum wage, not the income actually available after contributions and tax. France’s 1950 SMIG and today’s SMIC are not a politically neutral series either: their increases reflect minimum wage policy as much as productivity. With those limits kept visible, the orders of magnitude are still revealing.

The progress is striking. A baguette represented about fifty minutes of the Paris minimum wage in 1950, compared with just over five today. One hundred litres of petrol required about sixty minimum-wage hours in 1960, compared with fourteen to fifteen in 2025. Even electricity, despite becoming more expensive since 2010, remains far more accessible than at the start of the series.

Housing tells a different story. Its price relative to income remained close to its historical trend for several decades before breaking away after 2000. This helps explain why so many people can feel poorer amid undeniable progress: we occasionally replace a screen that has become cheap, but we pay for housing, energy and services every month.

Progress has mostly reduced the price of what factories can copy. It has done far less to reduce the price of what depends on a place, a finite resource or an hour of human time.

02 · The white-collar revolution
AI is not yet replacing the worker. It is changing what one hour can produce.

Automation first hit manual workers in controlled environments: factories, warehouses and logistics chains. Office workers had already lived through spreadsheets, word processors and the internet. But those tools mostly helped them produce the deliverable. Generative AI can now produce a first version of the deliverable itself.

This is why the wave is reaching lawyers, consultants, translators, developers, designers and support agents so quickly. Their raw material is already digital: text, images, documents, data and software. Part of an hour spent on research, writing or coding can sometimes be compressed into a few minutes.

This is not yet the disappearance of the professional. It is the gradual disappearance of part of the time they used to sell. Judgement, verification, responsibility and relationships remain work. But the content of an hour is beginning to change.

In plain terms, the time that previously allowed one hundred requests to be handled now allows about one hundred and fifteen. This is an observed productivity increase, not a forecast. Yet on its own, it tells us nothing about how the gain will be distributed.

The worker might finish the same amount of work sooner, reduce their effort, focus on more interesting tasks or negotiate better pay. The customer might pay less. The company might produce more, increase its margin or hire fewer people. Shareholders might capture most of it. The real political question is therefore not only what AI can produce, but who receives the time it frees.

And fifteen productivity points do not mean a fifteen-point fall in price. A company that controls the models, the data or market access can turn technical abundance into rent. The cost of production can collapse without the price paid by the customer falling at the same pace.

I think robots will eventually take on cleaning, cooking and monitoring just as AI is taking on office work today. But the order matters. A robot can learn a repetitive movement in a factory more easily than it can navigate a cluttered kitchen with a child and a pet nearby. Screens will come before homes, controlled environments before everyday life.

Craftsmanship will not disappear as a result. IKEA reduced the functional price of furniture without eliminating the cabinetmaker. When function becomes abundant, provenance, customisation and the fact that a particular human made the object become more valuable. Automation destroys some forms of scarcity and creates others.

03 · The political choice
A machine can be better, without being considered legitimate.

Musk understands better than almost anyone how quickly a technology can progress. But he sometimes treats society as a second machine, one that will inevitably adopt the most effective solution.

Yet we do not only expect the best statistical diagnosis from a doctor. We want an explanation, an accountable person and someone to speak to. In justice, a decision can be consistent and still feel illegitimate if nobody can challenge it. In war, an error is not an imperfect output. It is a death. In these domains, we may choose to keep the final word even if the machine becomes better.

These figures do not prove that the public will always prefer a human to a more effective AI. They show something more precise. A majority of Europeans can expect AI to have a positive effect, while part of the US public can recognise a procedural advantage without accepting that it decide alone. We can recognise a quality in the machine and still refuse to surrender the decision.

This resistance will not only be legal. It will be intimate. A technology that frees us from a chore is desirable. An intelligence that writes, decides or creates better than we do raises a more brutal question: does it make us freer, or merely less necessary?

I still believe automation will ultimately benefit humanity as a whole. But I believe neither in the 2036 timetable nor in a transition without conflict. Before work becomes optional, millions of people could lose some of their bargaining power while a handful of companies own the machines. Abundance in production is not yet abundance in access.

A universal basic income can serve as a bridge, but it does not build an apartment, a power station or an additional doctor. Our world remains finite. Producing more is not enough when the problem is the location of a home, the capacity of a grid, a specialist’s time or ownership of the machines.

Even if we overcame all those obstacles, abundance would not end competition. Not everyone can own the best-located home, receive all the attention, command the most trust or wield the most power. Our needs can be met, but our desire for distinction continually recreates scarcity.

Then comes meaning. I do not think we will stop acting because we no longer need a salary. We will continue in order to learn, care for others, be recognised and move society forward. We will simply have to stop confusing employment with usefulness. The real danger of an abundant world would not be that humans have nothing left to do. It would be that, after handing all their functions to machines, they no longer know why they matter.