Signal & Brus

🇸🇪 Svenska · 🇬🇧 English · 🇪🇸 Español

I believe we will see a substantial correction in the AI stocks that have surged on lofty expectations. At the same time, I think the most interesting winners may be found among companies that use the technology—and that the stock market does not yet associate with AI.

A technology can change the world even if the shares of the companies behind it have become too expensive. I can therefore be sceptical of AI valuations while remaining optimistic about what AI can do for corporate profitability.

Not everyone has to fall by the same amount

My view is that companies that have not been lifted by the AI hype should have less of that particular air to let out. When I look at Ericsson and Betsson, for example, I find it difficult to see why an AI correction in itself would justify much lower P/E ratios.

But that is an assessment, not a floor under their share prices. When markets are unsettled, investors may sell broadly. A share can also fall at an unchanged P/E ratio if earnings decline. Ericsson’s demand is affected by telecom operators’ investment; Betsson’s profitability by factors including taxes, regulation and competition. What matters is whether earnings hold up, not merely what the share costs relative to last year’s results. Companies that depend on growth expectations and high P/E ratios—or that even struggle to make a profit—also have further to fall when the market starts demanding real returns.

How much can AI change?

It is tempting to start with roughly two per cent annual growth and then add AI on top. But there is no universal two-per-cent rule for the world’s companies. Revenue, profit, GDP and productivity measure different things. According to the OECD’s 2026 productivity report, labour productivity in the OECD increased by 1.2 per cent in 2024, with large differences between countries.

In a study published in the Quarterly Journal of Economics in 2025, involving more than 5,000 customer-service employees, the number of issues resolved per hour increased by an average of 15 per cent after AI assistance was introduced. The gains were greatest for less experienced employees. This was an improvement in a particular operation—not 15 per cent annual growth for an entire company.

At the same time, a 2025 IMF modelling study estimates a cumulative improvement of about one per cent in total factor productivity in Europe over five years. It is a scenario based on uncertain assumptions, but a useful counterweight to the biggest promises.

The average effect may therefore be modest even as individual companies change substantially. Imagine a company with revenue of 100 and costs of 90: operating profit is 10. If AI reduces total costs to 87 after the cost of the technology, profit rises to 13—a full 30 per cent—without any revenue growth. It is only an illustration, but it shows why even small net savings can have a large effect.

Where might the winners be found?

I would look primarily for companies with large amounts of recurring information work, useful proprietary data and strong customer relationships. Here are a few areas worth investigating:

Insurance and banking. Claims handling, document review, customer service and fraud detection can become more efficient. Large case volumes make small improvements valuable. But incorrect decisions and inadequate data protection can quickly consume the savings.

Industry and logistics. Better production planning, quality control and preventive maintenance can reduce waste and downtime. Atlas Copco, Sandvik and Volvo are examples of companies I would examine more closely, not predetermined winners. The evidence must appear in costs, deliveries and cash flow.

Telecoms. Ericsson already describes AI applications for more efficient networks and lower energy use. Among other things, the company reports energy savings of 33 per cent in radio units with a particular suite of AI applications. That is the company’s figure for a specific application, not a saving across Ericsson as a whole. The opportunity lies in both better products and more efficient operations.

Digital services and gambling companies. Customer service, software development and regulatory compliance are possible areas of use. In its 2025 annual report, page 74, Betsson states that AI assistance is used for quality control in customer service, including communications about responsible gambling. This demonstrates actual use, but does not say how much profit might increase.

These are candidates for further analysis. The fact that an industry can become more efficient does not mean every company will become more profitable. Consulting firms that sell hours, for example, may produce work faster but also have fewer hours to invoice. And if all competitors become more efficient, customers may demand lower prices.

Who gets to keep the gain?

This is where my biggest objection lies: what happens if companies become dependent on a small number of AI and cloud providers? Once data, workflows and systems have been tied together, switching providers can become costly and difficult.

This is a concrete competition issue. In March 2026, the UK’s Competition and Markets Authority reported that its cloud investigation had found significant market power held by Amazon and Microsoft, as well as barriers to switching providers. The authority also described steps taken by the companies to reduce certain barriers. Cloud services are not the same as the entire AI market, but the dependency is relevant.

Falling share prices do not automatically raise service prices. My concern is that weaker funding and demands for profitability may increase the pressure to charge more—and that poor competition makes this possible. Providers could then capture a large share of their customers’ efficiency gains. Open solutions and easier switching could push in the opposite direction.

I therefore want to see more than an AI promise in the next company presentation: is revenue per employee rising? Is the margin improving after all AI costs? Is cash flow keeping pace? And can the company retain the benefit without becoming dependent on a single supplier?

I believe the stock market may have run ahead of the AI companies’ earnings. At the same time, it may be underestimating what the technology can do for other businesses. That is where I intend to look: among companies that can show that AI makes their business better.

This English translation was prepared with AI from the author’s Swedish article. The featured illustration in the original is AI-generated.