AI over-investments

AI: a story of over-investments and long-term economic value.

There are large long-term economic benefits from using AI - uses cases are many. The value created by AI services will grow over time, as e-commerce has slowly but steadily taken its place in the retail industry. (see illustration)

Meanwhile we are seeing a very large investment in infrastructure (data centres, GPUs) to power AI models and services. Some of this is over-investment, driven by the fear to miss out, as acknowledged by several industry leaders. Without a crystal ball, it is hard to know where to stop and what will trigger the start of a rationalisation period. But...

There is a physical limit to everything: eventually, companies will slow down their investments, digest what they have and focus on sweating these assets. Hype will go away, we will get used to having these new tools; we will expect them to work like we expect light to come on when we switch a button -- being invisibly useful is the best outcome for a technology.

Investors will ask -- when will this peak happen. I think we can triangulate the timeline. How many large companies in the US/Europe have already announced large investments in AI, as a total? Assess whether they are likely to repeat this level of investment next year. Compare with consensus expectation of revenue growth for the leading infrastructure suppliers. Wait a quarter (or two?) until they announce their results.

---