Echoes: Bubble or no bubble, AI investment is still driving growth

Natixis IM Market strategist Mabrouk Chetouane analyses the academic evidence to see whether we are indeed experiencing an AI-driven market bubble and, if so, whether it poses a real and present threat for investors.

It is essential to remember that the key characteristic of a financial bubble is that it is unobservable. We can only infer the existence of a financial bubble either through historical comparisons or through a set of information and variables whose selection is subject to heated debate.

That said, there are two indisputable facts about bubbles:

1. They are closely linked to investors' expectations that the price of assets held in their portfolios will continue to rise, enabling them to continue to make gains. Conversely, the bursting of a financial bubble is also based on expectations of a reversal in the price of these same assets.
2. They, especially a financial bubble, can only be identified once they have burst. In other words, we know only after the event that we were in a bubble. This gives the phenomenon a ‘quasi-quantum’ dimension, in other words, that of being in several states at the same time.

Bubble talk echoes many of the questions facing investors regarding the new technologies sector linked to artificial intelligence (AI). Does the tech sector represent a financial bubble? It is impossible for us to answer this question with a definitive yes. Yet this observation should not prevent us from conducting an analysis based on tangible evidence and reasoning to assist investment decisions.

Does the technology sector, and the associated risk of a bubble, therefore pose a threat to investors? At this stage, based on our knowledge and the information available to us, I don’t believe the market is in a bubble. Moreover, the opportunity cost of under-exposure to technology is a penalising factor in asset allocation.
 

Creative destruction and productivity

The 2025 Nobel Prize in Economics was awarded to Philippe Aghion and Peter Howitt for their work on the knowledge economy and their contributions to the mechanisms of creative destruction in economic systems¹. More generally, this work falls within the framework of endogenous growth models, which study the role of research and development and innovation on productivity and growth paths².

These theoretical frameworks establish two fundamental propositions:

1. Romer (1986 and 1990) develops a framework in which economic growth becomes a self-sustaining phenomenon, supported by non-decreasing returns to scale². In other words, there are positive externalities generated by research and development expenditure that provide lasting support for factor productivity and therefore economic growth.
2. Aghion and Howitt (1992) show how innovation creates a sustainable growth cycle in which new products and new production methods replace old ones, thereby stimulating economic growth. Growth can continue despite the negative effect of creative destruction (to use Schumpeterian terminology) that occurs during the economic cycle.

Romer, Aghion and Howitt therefore place the concepts of investment, R&D and productivity at the centre of the growth process. Over the last three years, we have seen a surge in investment in the new technology sector. More specifically, in 2024, nearly $230 billion was deployed by the leading companies in the technology sector (Amazon, Alphabet, Microsoft, Meta), representing 7.8% of US corporate investment. This figure is expected to reach $400 billion in 2025 for these same companies, an increase of more than 70%³.

Even if these expenditures do not exclusively concern research and development (investment in capital goods, intangible capital, etc), it is reasonable to believe that they will stimulate the economic fabric, generate positive externalities and ultimately support the productivity of production factors. As a result, they are likely to increase the income of the players and economies that have undertaken this effort.

In the G7, for example, AI could have a significant impact on labour productivity and, consequently, on the potential growth of economies over the next decade – as highlighted in the OECD research in Table 1⁴. These productivity gains are dependent on the speed of deployment and, above all, adoption of these new technologies within the economic fabric of each country.

Unsurprisingly, the US dominates this sector and is likely to maintain its lead regardless of the scenario considered. Assuming an additional productivity gain of 0.41 pp for the US, the additional income generated over the next 10 years would be around £1.4 trillion.

Furthermore, Fiori et al (2025) emphasise that the time lag between substantial investments and the introduction of new technologies into production systems tends to limit productivity gains⁵. Put simply, procrastination does not go hand in hand with higher productivity gains. In their study, the US, followed by the UK, still largely dominate European countries.

Without spending on R&D, which is the cradle of technical progress, the probability of maintaining a constant level of potential growth is low. This result applies regardless of the scale considered, whether it be a company or a country.

The difference in per capita income between the US and the eurozone has increased by nearly 40% over the past 20 years – Fiori et al (2025) attribute part of this gap to significantly higher productivity gains in the US⁵. However, improved productivity gains are dependent on investment dynamics, all other things being equal.

In this area, companies in the eurozone lag significantly behind their American counterparts. In the third quarter of 2025, the gap in private investment between the two zones reached 35 points in favour of the US⁶. What, then, is the link with the risk of bubbles?

Excessive expectations and the energy sector

Excessive capital accumulation, in the Marxist sense, is the result of overinvestment. Initially, this leads to sub-optimal allocation of capital and can then lead to various forms of excess, such as overproduction or a collapse in capital returns, which inevitably results in economic and/or financial crises.

Similarly, this overinvestment can raise expectations of future income growth and result in exponential price growth – a characteristic sign of a bubble. These expectations could be deemed excessive and lead to a sharp fall in prices.

Nevertheless, there is no objection to comparing the profits made by companies in the technology sector with those of the rest of the US stock market. These companies, which are in a phase of expansion, have largely outperformed companies outside the technology sector and exceeded analysts' expectations.

Despite spectacular earnings growth, the valuation levels of technology companies in the US are actually lower than at the end of the 2000s. As shown in Figure 1, another fact is that the rise in the valuation of this sector has been more gradual, even if a few hiccups have disrupted this upward trend.

Although significant investment efforts have been made, maintaining a dominant position in the areas of computing capacity and information and data storage will require further investment. Yet there is one factor that could potentially hinder the sector's long-term viability: energy.

As Figure 2 shows, installed electrical capacity in the US stood at 1,294 gigawatts in 2024 – an increase of only 3% compared to 2023⁷. In China, installed electrical capacity reached 3,217 gigawatts over the same period, 2.5 times that of the US.

The energy issue is strategic, even vital. Logically, we can expect significant investment in this sector in the US to ensure that it does not become an obstacle, which would ultimately be a positive externality for the entire economy.

Figure 1:  US Stock market valuation - Price Earning ratios 

This new ‘industrial revolution’ naturally raises questions about the relevance of investment decisions. The time frame required to see these advances materialise in terms of productivity and, more visibly, in corporate earnings, is still long.

I believe that the lead enjoyed by the US and the potential for the spread of these new technologies will enable the emergence of new players, possibly in the form of a duopoly or oligopoly, which will lead to greater adoption of these new technologies – a prerequisite for maintaining productive economies.

So, what’s the verdict? In summary, at the time of writing, all the analysis suggests the current situation does not constitute a bubble. But it’s worth acknowledging that not everyone agrees with this sentiment

Interviewed in December 2025

¹ Aghion, P and P Howitt (1992), “A model of growth through creative destruction”, Econometrica 60(2): 323-51.

² Romer, P M (1990), “Endogenous technological change”, Journal of Political Economy 98(5): 71-102.

³ Wall Street Journal, ‘Big Tech Is Spending More Than Ever on AI and It’s Still Not Enough’, October 2025,  https://www.wsj.com/tech/ai/big-tech-is-spending-more-than-ever-on-ai-and-its-still-not-enough-f2398cfe

⁴ Filippucci, F, Gal, P, Laengle, K, and Schief, M, 2025, ‘Macroeconomic productivity gains from Artificial Intelligence in G7 economies’, OECD Artificial Intelligence Papers, No. 41, OECD Publishing, Paris, https://doi.org/10.1787/a5319ab5-en

⁵ Giuseppe Fiori, Colleen Lipa, William Wu, 2025, ‘Investment as a Source of Productivity Growth," FEDS Notes. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/2380-7172.3920

⁶ Mabrouk Chetouane, Bloomberg, Natixis Investment Managers, 2025

⁷ Haag, Alex, 2025, ‘The State of AI Competition in Advanced Economies’, FEDS Notes. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/2380-7172.3930