Philippe Aghion: AI’s growth potential depends on competition

New technology can open the door to innovators or entrench the companies already in charge. The economist explains why his optimism depends on the institutions we build.

A blurred, glowing white silhouette of a galloping horse against a gray background.

Economist and 2025 Nobel laureate Philippe Aghion has spent much of his career studying creative destruction: how innovation drives growth by displacing legacy technologies and incumbent firms. With Peter Howitt, he turned the idea popularized by Joseph Schumpeter into a formal framework explaining how this process can sustain economic growth. In a conversation with the Review, Aghion argues that AI’s impact on the economy will depend on the policies that governments put in place.

This conversation has been edited and condensed from two interviews.

AI promises abundance. But what will determine which future we get?

I often compare a technological revolution to a horse: The horse can kill you, but it can also take you wherever you want to go.

Philippe Aghion: It all depends on the institutions we put in place. I cannot tell you a priori whether the outcome of AI will be good or bad. I often compare a technological revolution to a horse: The horse can kill you, but it can also take you wherever you want to go. It depends on how you manage and steer it. It is the same with AI. If we put the right institutions in place, we can get very good outcomes from the AI revolution. If we do not, then the worst is to be feared.

AI has enormous growth potential because it can automate tasks—not only in the production of goods and services, but also in the production of ideas. It makes ideas easier to find. But the danger is that we end up with only a few large firms monopolizing the AI revolution. They could become so dominant that they discourage new innovators from entering the market.

The Humanist Review of AI, Issues 01 and 02

Conversations at the frontier, in your inbox

Subscribe

Creative destruction depends on new innovators displacing incumbents. What happens when yesterday’s innovators become powerful enough to close the door behind them?

Aghion: Every company wants to do that. It is at the heart of the creative destruction model. Yesterday’s innovators always want to prevent new innovations because they do not want to become subject to creative destruction themselves. The question is what governments will do. Will they collude with incumbent firms, or will they impose genuine competition policy? Governments must do the right thing. But civil society, voters and the media also have an important role in exposing collusion and making sure that new entry remains possible.

AI development is increasingly concentrated in a handful of powerful companies. How can competition policy keep the field open to new innovators?

Aghion: There is competition between countries, but there is also competition between firms. Both matter. Without competition, one or two firms can monopolize the market and discourage everybody else. That is why we must insist on open source and data sharing. When mergers and acquisitions are proposed, competition authorities must make sure they will not stifle future entry and innovation.

Your research shows that policy can shape innovation, including by enabling new discoveries to build on earlier ones. Does AI change what governments need to do? Do we now need stronger policies to support research and innovation?

Aghion: We do, of course. Europe is falling behind the United States and China because we do not provide enough long-term funding for basic research. We need more of it because we need to encourage risk-taking and failure. Failure should not be seen as a drawback; it should be seen as a virtue. We need more venture capital to encourage risk-taking in startups. We need more institutional investors—pension funds and mutual funds—to encourage risk-taking by larger firms. And we need a DARPA, the Defense Advanced Research Projects Agency, as a way to reconcile industrial policy with competition policy, and public procurement with competition policy. I think we need all those things more than ever.

Has your thinking about growth and innovation changed over time?

Aghion: I would never have thought that mathematical problems would be solved so quickly [by AI]. Claude is checking my proofs and saving me so much time proving theorems and results. He’s just amazing. It’s going much faster than I ever thought. So that reinforces my conviction that we need a good education system, good competition policy to make sure that we don’t have one or two firms monopolizing the system and good flexicurity systems.

Every technological revolution destroys jobs and creates new ones. But what makes AI’s impact on work different? And are we condemned to repeat the same debates?

That may be the specificity of AI: It threatens some jobs that were not threatened before.

Aghion: The basic debates are the same. AI destroys some jobs because it replaces tasks, and a job is a collection of tasks. In some occupations, AI may replace most of those tasks and put the job in jeopardy. But AI can also create new jobs. The difference with AI is more about the kinds of jobs that may be threatened. AI can replace tasks in highly skilled occupations. Paradoxically, many manual and interpersonal occupations are less threatened. Nurses, plumbers and cooks may be more resilient. That may be the specificity of AI: It threatens some jobs that were not threatened before. But there are labor market policies that can help us adapt. The question is, what do we do about it? Education is key. You need an education system where you learn to learn, even if the knowledge you acquire might become obsolete very fast. You need a whole part of schooling without AI, where you learn to read, write, calculate and demonstrate theorems on your own, because that makes you [able to] adapt. A big lesson is to reform the education system.

Apart from changes to the education system, what are examples of adequate labor-market policies? How should society respond to AI’s effects on the labor market?

Aghion: We need a flexicurity system. We need to emulate what they do in Denmark. When you lose your job in Denmark, for two years you receive [up to 90%] of your salary. The state retrains you and helps you find a new job. More than ever, we will need this kind of system with AI, because there will be job destruction and job creation. We need to smooth people’s transition from old jobs to new jobs.

Some people advocate for universal income. What I don’t like about universal income is the idea of receiving an income to do nothing and stay at home watching television. I think it is very important to have a social activity. So I would prefer income to be conditional on either training or working. That brings you closer to the flexicurity system in Denmark. Another point is that jobs such as plumber, cook and other manual occupations will survive AI very well, so we should encourage them more. In France, over the past decades, we have tended to look down on those jobs. I think we should rehabilitate them, as Switzerland has done. They have a very good system: You can become a fantastic plumber and still have a bridge back into general higher education. Soft skills are also very important: People need to know not only how to perform their own tasks, but also how to interact with others in the firm. We should encourage firms to strengthen those skills. All these policies would help us to make the best of the AI revolution.

If AI creates extraordinary wealth for a very small number of people, how can that abundance serve the wider public without undermining innovation?

Aghion: Some people may become extremely rich very quickly. I think they should have strong incentives to put part of their wealth into foundations supporting areas such as education, research and defense. I see this as an alternative to the Zucman tax [a wealth tax proposed by UC Berkeley and Paris School of Economics professor Gabriel Zucman]. A young company may have a very high valuation, but that valuation can shrink if a bubble bursts. This is unrealized wealth. If you tax the entrepreneur on that valuation, they may have to borrow to pay the tax rather than borrow to invest. They will lose out to competitors who do not face the same tax. I favor reforming inheritance tax. People could be encouraged to put part of their wealth into foundations before they die, with the rest taxed at a lower rate.

Would you describe yourself as an optimist when it comes to AI?

I am not a kind of optimist you find in Voltaire. I am more of a fighting optimist.

Aghion: I am a guarded optimist. I am not a kind of optimist you find in Voltaire. I am more of a fighting optimist. I think AI can be put to good use with sound institutions and sound government policies. Then you can harness the growth potential. Of course, if you don’t, you get the worst from AI: mass unemployment, skyrocketing inequality and environmental disaster. It all depends on how we use AI.

Philippe Aghion

Philippe Aghion is a French economist and professor at the Collège de France and INSEAD. With Peter Howitt, he developed a theory of economic growth through creative destruction, for which they shared the 2025 Nobel Prize in Economic Sciences with Joel Mokyr. His books include “The Power of Creative Destruction,” which he co-authored with Céline Antonin and Simon Bunel.

English (United States)
Your Privacy Choices Opt-Out Icon Your Privacy Choices
Consumer Health Privacy Sitemap Contact Microsoft Privacy Manage cookies Terms of use Trademarks Safety & eco Recycling About our ads