Dear Tánaiste Harris—dear Simon—thank you for inviting me to the land of Guinness. As we heard yesterday, Guinness has a special place in Ireland’s history. It also has a special place in my own life. When the wind of perestroika blew into Bulgaria, one of the outcomes was the opening of hard-currency shops. My very first purchase in hard currency was a can of Guinness. As I drank it, it was inconceivable that Bulgaria would be a member of the EU and the euro would be our currency. But it happened.
So my starting point today is that Europe is capable of unthinkable transformations.
This is, again, the task we face with AI. To maintain Europe’s cherished social model, it needs to grow faster—and take advantage of the leaps in technology that make it possible.
Love it, hate it, or fear it, the new frontier is defined by AI—and with AI, opportunity meets risk on an epic scale.
To extract the productivity benefits of AI, economic agents must rethink, re-engineer, and reinvent their product lines, business processes, and workflows. As with the railways, electricity, or the internet, it is these changes—not the technology alone—that are transformative.
And yet what makes AI different is its speed: it moves at a breathless pace, and so too must we—you snooze, you lose.
For Europe, keeping pace means becoming an integral part of the AI value chain. ASML in the Netherlands and Mistral in France are just two cases in point. Yet still, European countries trail the U.S. and China on AI, as well as AI suppliers in Asia like Taiwan Province of China, Korea, Malaysia, Thailand, Singapore, and Japan.
The U.S. and China, of course, are the world’s leading AI providers, creating the pay-by-token frontier models and the open-source solutions just one step behind. The others are AI builders, each with its own special role in the value chain.
Naturally, all these economies are also AI adopters—although that is a group any country can join by outsourcing.
Europe, with its size, has to aspire to be all three.
And it has to do so comprehensively. So let me describe the four main channels through which AI will affect economies:
- First, in the short run, the AI building boom is a positive demand shock—with strong private investment lifting both growth and inflation in economies in the value chain. The inflation effect is amplified by AI’s high energy intensity paired with the still-unfolding global energy supply shock.
- Second, in the longer run, the expected productivity boost from AI is a positive supply shock. We estimate AI could lift global annual potential growth by some 0.1–0.8 percentage points eventually, with forthcoming research confirming that Europe can and should aspire to be in the upper half of this range. Whether the long-run effects are inflationary or deflationary is a complex question, one that depends, first, on the leads and lags between the productivity gains and the increase in consumption from higher expected permanent income and, second, on whether AI through its labor-market effects collapses large segments of consumer demand.
- Third, AI could hit the labor market like a tsunami, affecting up to 60 percent of jobs in advanced economies. Recent IMF research finds that, already, about one in ten job vacancies in advanced economies asks for at least one new skill. These new skills boost average wages and employment but deepen polarization. Who benefits? High-skilled and low-skilled workers—the latter because higher incomes and buoyant stock markets drive higher consumption of services. But we get a hollowing out of the middle.
- And fourth, somewhere in the transition between the short run and long run, we will likely traverse the period of maximum risk. I am reminded of a dictum called Amara’s Law, which states that we tend to overestimate the effect of a new technology in the short run and underestimate its effect in the long run. It implies disappointment is more likely to be frontloaded—and therein lies financial stability risk. At the same time, as private AI providers run ahead of official rules and oversight, some are warning that the frontier models may soon slip beyond human control. It is not hard to imagine how this could cause havoc in the financial system and, at worst, grow into an extinction-level threat.
READ MORE: https://www.imf.org/en/news/articles/2026/09/20/sp091926-europe-and-global-ai-race
