AI triggered a software sell off
Quarterly reports continue to arrive, and at the time of writing, far from all companies have reported. No clear overarching trends are yet visible in the results published so far, as most companies largely tell their own individual story.
If one nevertheless attempts to identify certain patterns, a growing number of small and mid cap companies mention improving sales trends in Sweden. Another recurring theme is increased market scepticism towards companies making very large investments related to data centres. The major US technology companies, including Meta Platforms, Alphabet, Amazon and Microsoft, have together announced investments of approximately USD 650 billion, equivalent to the entire annual GDP of Sweden, to be deployed over the current year.
A year ago, the market praised large scale AI investments. Today, announcements of this magnitude are increasingly causing investors to hesitate, particularly in the software sector.
Software stocks have recently underperformed, as artificial intelligence is expected to challenge existing business models. It remains unclear how this will materialise in practice, but the argument is that AI tools, given relatively simple prompts, can generate fully functional applications and thereby lower barriers to entry and intensify competition.
Microsoft declined by 12 percent in connection with its earnings release and is down 17 percent year to date. One could argue that Microsoft was trading at an elevated valuation, meaning it did not require much negative news to trigger a correction. However, the sell off has been broad based. Even software companies that initially traded at comparatively low valuations have been affected. Sinch is one example of a company that has been drawn into this downward movement despite starting from a low valuation base.
There is considerable discussion regarding the magnitude of productivity gains that AI may ultimately deliver. US productivity has begun to trend upward over the past quarter, but we do not believe artificial intelligence is yet the primary driver of this development.
Historically, the most significant productivity gains from the IT revolution occurred between 1995 and 2005, despite the bursting of the technology bubble well before most of these gains were realised. The question is how far into the AI cycle we currently are. OpenAI launched ChatGPT in late 2022, and the technology has therefore been widely accessible for just over three years.
New technologies often lead to an initial decline in measured productivity. Investment costs arise first, followed by a learning phase, and subsequently a period during which legacy production processes are maintained in parallel with the development of new systems. Hours worked per unit of output may increase before efficiency improves over time.
Our assessment is that we remain relatively early in this productivity cycle, and that more substantial gains are likely several years ahead. So far, few companies have clearly highlighted measurable productivity improvements linked to AI in their quarterly reports.
We also observe that markets can rapidly shift between narratives. Stocks strongly favoured under one prevailing theme can quickly fall out of favour when the narrative changes.
Our conclusion is therefore that diversification is more important than it has been for some time. Chasing highly valued stocks carries elevated risk, as it is impossible to predict where the next bout of excess enthusiasm, or the next correction, will emerge.