Artificial intelligence (AI) has undergone a major transformation since large language models (LLMs) arrived on the scene a few years ago. In all countries, this rapid growth is likely to affect economic activity through several channels in the short term.

In the United States, since late 2022 and the launch of ChatGPT to the general public, the surge in demand linked to AI has primarily been reflected in accelerated business investment in digital assets and related infrastructure (IT hardware, software and data centers). While investment in data centers – which has a small weight in GDP – has surged since late 2022, corporate investment in digital assets only began to ramp up in 2025. All in all, digital investment accounts for a third of US growth (+0.7 percentage points in 2025). In France, the increase is expected to be smaller, at around +0.1 percentage points in 2025. In both countries, the statistical picture remains incomplete, as AI-related expenditure is split between hardware, software and infrastructure.

Across the Atlantic, the rise of AI may bolster demand in the short term through a second channel: rising financial asset prices and “wealth effects” on household consumption. These wealth effects are estimated to have boosted growth by 0.1 percentage points in 2025. In France, however, the impact of this channel appears negligible, as wealth effects are traditionally weak and the savings ratio is historically high.

High exposure among US households and signs of accelerating growth in certain valuations within the technology sector are increasing the risk of a downturn similar to that seen during the dot-com bubble of the early 2000s. If a sharp correction were to occur, it would have a major impact on US economic activity. However, the IMF believes, based on price-to-earnings ratios, that the risk of overinvestment and a financial bubble linked to AI is lower than it was during the dot-com boom.

In addition to the main demand channels, the widespread adoption of AI could also affect economic activity on the supply side by boosting productivity gains. In particular, sector-specific indicators suggest a downturn in employment over the past two years in the most exposed sectors, most notably in computer programming, where value added is on an upward trend in both the United States and France. In addition, in both countries, the adjustment in employment in this sector is likely to affect young people entering the labour market...