Why the AI Boom Could Keep Interest Rates High

ChatGPT was launched in 2022. It could answer simple questions, an improvement over Google, which could only find information.

The first reasoning AI, o1-preview, followed in 2024. AIs could now answer not only everyday questions, but also more complicated technical questions requiring deep thought.

Coding agents were the great innovation of 2025. Instead of writing code yourself, you could tell an AI what you wanted and let it build the software.

The fourth stage of the AI revolution, computer-use agents, are the breakthrough of 2026. AIs can now use applications and websites on your behalf: shopping through a browser, answering emails, organising files, or creating and editing spreadsheets.

My own AI usage has exploded this year, and I expect AI to diffuse rapidly into office work over the next couple of years.

This means much more demand for AI compute, chips, data centres and electricity.

All of this needs to be paid for. And this comes at a time when governments are already borrowing heavily.

My view has consequently become more cautious on long-term interest rates. I now expect the German 10-year yield to remain above 3%, while the US 10-year yield could potentially reach 6% next year.

This has led to some changes in the portfolio.

I have reduced exposure to property and telecommunications. The valuations of these stocks are particularly sensitive to long-term interest rates.

At the same time, exposure to Energy has been increased. Oil prices have risen sharply this year and oil companies are posting record profits.

I see a meaningful risk, however, that increasing supply eventually pushes oil prices lower in 2027. The Energy view has therefore been expressed through oil services companies. They benefit from higher oil and gas investment, but have less direct exposure to the oil price than producers.

Although my view is that interest rates will increase next year, high government deficits and rising bond yields are now discussed almost everywhere. That makes me cautious. What you read in the media is often a contrarian signal: by the time a theme is widely covered by mainstream journalists, professional investors have typically already positioned for it and may be looking to exit.

I therefore haven't sold all of the portfolio's property exposure. Interest rates will probably remain structurally higher, but I would not be surprised to see bond yields fall temporarily during the final quarter of this year.

One of the lessons from this year is that you don't need to own AI companies to be exposed to AI. Many companies — software and professional services firms among them — will be affected by AI directly. Even companies owning physical assets can be substantially exposed through AI's effect on interest rates.

The portfolio has to be resilient to any AI scenario, and the recent changes are intended to reflect that.