Several years ago, when chess computers emerged, no one expected that these computers would eventually beat grandmasters. But these machines kept improving, and today the computer defeats the human chess player.

Now, the tech company Ultramarin, in collaboration with Goldman Sachs and Baader, has launched the AI-Enhanced Eurozone Equities UCITS ETF, the first actively managed ETF entirely controlled by Artificial Intelligence (AI). Other firms also use AI not merely as support in portfolio management, but as the decision-maker instead of a human fund manager, both for mutual funds and actively managed ETF formats.

For example, at Acatis AI Global Equities, AI is responsible for selecting stocks in the fund. Similarly, Oddo BHF Artificial Intelligence uses AI to identify the most promising companies related to the AI theme.

Depending on the focus of a fund, the AI’s investment decisions might involve selecting individual stocks or determining the investment level in stocks or bonds. This naturally raises the question of whether AI is better than a human as a fund manager. For instance, is the AI better than Klaus Kaldemorgen and his team or better than Bert Flossbach and his team, just to name two prominent and successful fund manager.

The answer is complex and cannot be a simple yes or no. Firstly, there isn’t a single AI acting as a portfolio manager. Instead, there are AI developments for portfolio decisions created by various teams of rocket scientists. One AI algorithm might be convincing, another might not be, similar to human managers.

Secondly, capital markets are complex, dynamic systems. All AI developers conduct backtests, usually spanning several years. A clearly positive backtest result is certainly something. Ultramarin also points to a significantly positive backtest for the ETF mentioned above, covering several years. However, in investment funds, performance with real money is what counts. And not just over one or two years, but long-term over five or ten years. Whether an AI-driven strategy truly works will have to be proven in practice over the coming years.

An interesting aspect is also the costs. For example, passive ETFs on standard equity indices benefit from having low ongoing costs for both institutional and retail investors, usually just a few basis points per year. Because costs affect performance.

So, how expensive will AI-managed funds be? What does it cost to develop an AI that is successful in portfolio management? The aforementioned AI-Enhanced Eurozone Equities ETF has ongoing costs of 65 basis points. While not extremely expensive, institutional investors often pay less for other ETFs and funds.

On the other hand, asset managers can reduce their production costs through increased use of AI at various levels. In recent years, more and more quantitative approaches have improved asset management and attracted significant funds. Often, a human fund manager still reviews the results of quantitative models and analyses. But what happens in extreme market situations if AI alone makes the investment decision? Shouldn’t an experienced human in capital markets ultimately make the decision?

Or is it alike to the chess computer situation? It’s conceivable that in a few years, powerful AIs will mainly make the investment decisions. Because, like the chess computer, they have proven to be better.