CFA Society Netherlands: On the winners and losers of AI

February 20, 2026

_{This column was originally written in Dutch. This is an English translation.}

Column by CFA Society Netherlands

AI is a wonderful technology that is likely to lead to higher economic growth. But it also has downsides, such as high energy consumption, increased inequality and greater concentration in equities.

By Raphie Hayat, Investment Strategist at a.s.r. asset management

I recently spoke at the CFA Society Netherlands ALM conference about the macroeconomic effects of artificial intelligence. On the way, I stopped to pick something up at the supermarket and realised that supermarket checkouts are a good example of automation. They used to be manned by secondary school students, but now they are operated by a combination of software and hardware. Supermarkets have been winners (albeit in a specific sense) of the previous wave of automation.

Will AI herald the next wave of automation? That is certainly the promise. Most economists agree that the use of AI will be good for economic growth. After all, the automation of tasks leads to productivity gains, although opinions differ on exactly how much. Daren Acemoglu, winner of the Nobel Prize in Economics in 2024, estimates the additional annual growth in productivity due to AI at only 0.07%. Philippe Aghion, winner of the same Nobel Prize in 2025, estimates a much higher figure in an article with fellow economist Simon Bunel, namely 1.1%.

Economic growth is therefore likely to be a “winner” of AI. But there are also losers. For example, AI is a huge energy guzzler: both the use of AI models and their training require a lot of power. An example: according to a recent article in The Economist, training Meta's latest AI model (Llama) cost 27.5 gigawatt hours. That is equivalent to the annual electricity consumption of 7,500 European households. Added to this is the energy consumption during the use of the model. Previous estimates indicated that asking ChatGPT a question costs ten times as much energy as googling something (3 Wh versus 0.3 Wh). More recent estimates (from EpochAI) question this, partly because computer chips have become much more efficient over the past three years.

Ironically, the latter could indicate that AI's energy consumption is increasing at the macro level, due to “Jevon's Paradox”. This states that efficiency improvements in the use of energy or raw materials do not always lead to lower consumption, but sometimes to higher consumption. For example, the more efficient use of coal in steam engines led to steam technology being used more widely, which increased total coal consumption.

AI is also likely to increase rather than reduce existing inequalities. Developed countries are better positioned to benefit from AI than emerging markets. Developed countries score higher on the IMF's “AI Preparedness Index”. Furthermore, recent economic studies indicate that AI helps people who are already very good at a particular field more than those who are not so good at it. AI therefore helps the existing winners more than the “non-winners”.

This has also been the case on the stock markets so far. Expectations of AI have led to the ten largest companies in the S&P 500 (mainly technology companies) now accounting for 42% of the index. In 2010, this was only 19%. The “winners” of the stock market have therefore become even more prominent. And that is a risk for investors, because the globally diversified index is now in fact a “bet” on the American technology sector.

I thought about this as I cycled back from the conference. I had thoroughly enjoyed the many insights and wonderful encounters. But with the stock market in the back of my mind, there was one thought that made me feel a little uneasy: perhaps today's winners will be tomorrow's losers.