The market has experienced a wave of selling due to AI. In which sectors are investors looking for a safe haven?

Last week was the worst week for the S&P 500 since November 2025, losing 1.39% in five days, the Nasdaq Composite was down 2.1% and the Dow Jones was down 1.23%. In February, one of the reasons for the storm in the markets was a reassessment of the prospects of entire industries because of the development of AI. For investors, it's now a technology that must actually prove its ability to be profitable. What should they bet on now?

Investors have moved from the idea that they are working with AI as the future to demanding that this future be tested by accounting. This is most noticeable in the tone of analysts at major investment companies and banks. Next to technology terms, the words selective, monetization, payback, margin pressure are increasingly heard. This change indicates growing demands on the quality of development of AI instruments and the extent to which this growth is converted into fixed or expected profits.

When it comes to company valuations, there is less room for a "just good" report. Even strong numbers may not save the day if management is cautious in its forecasts or if the market believes that the next phase of growth will require disproportionate spending.

An example is software companies, which themselves have been pushing the artificial intelligence agenda. Not so long ago, they pitched it as a way to save money: fewer people to support, faster to write code, cheaper to release new versions. Now the main question is different: whether AI will absorb part of the products themselves, especially those that are not the "heart" of the business (where data is stored and operations take place), but simply make work more convenient.

In the same way, the earning model can change. Even if a product is needed by the customer, AI changes how it is paid for. For example, for information and service providers, the subscription model in the market may be replaced by a "pay for results" logic, i.e., for contribution to the result of customers' work (e.g., customer support, speed of processing requests, etc.) For companies that grew by selling licenses, this is a painful change in the rules of the game.

Finally, before, defending market position was about integrations, user habit and interface. Now, businesses will compete not only with similar programs, but also with AI platforms that can take away features from one product after another.

As a result, the IT business found itself in the center of the "soft-apocalypse" on the market and a victim of its own interest in AI. If customers are more willing to pay for results rather than for the technology itself, the concept of product value will change. As long as customer companies are actively implementing digital assistants and optimizing internal processes, they are operating in the old model, buying solutions on the side. But very quickly, they may start to refine the product by the vendor or themselves, which no one in the industry has done directly before.

In parallel with the reassessment of the impact of AI, some investors began to look towards companies in the real sector: transportation, industry and infrastructure. JPMorgan Chase described this dynamic as a "healthy" rotation: Private Bank strategist Stephen Parker described the sell-off in software companies and part of the technology sector as a sign of market expansion rather than a sign of "the end of the subject". In his estimation, money is moving away from too narrow a bet on AI and looking for opportunities in more "earthly" segments, such as industry and infrastructure.

UBS looks at the situation the same way. Its analysts point out that investors may be scared off by the scale of AI implementation costs and unobvious payback periods, as well as the risk that new AI tools may take away part of the revenue from software developers.

"It's quite possible that we are only seeing the beginning of a trend toward equalization of valuations (e.g., by P/E ratio) of premium sectors like IT, consulting, etc. and the 'old' economy - raw materials extraction, consumer goods production, pharmaceuticals," manager Alexei Tretyakov wrote in his Telegram channel War, Wealth & Wisdom on Feb. 9.

Last week, for example, Union Pacific (up 3.2% for the week in a falling market and 12.7% YTD) and FedEx (up 1.5% and 29.7%, respectively) acted as a clear "bet on the physical economy," while Caterpillar (up 6.6% and 35.14%) and Vulcan Materials (up 1.2% and 14.9%) were the standouts on the transportation and logistics side.

But it is important that transportation or logistics is not always automatically a safe haven. Reuters wrote that the wave of fears around AI last week affected, among other things, the shares of transportation companies. The market believes that AI can change the economics of individual representatives of this business, up to the point of changing the model of their work and leaders. So the rotation is not based on the principle of "everyone from the tech sector to the real economy", but more selectively - to the most understandable business models.

There have already been moments in the history of technology cycles when the market stopped "buying the dream" in its entirety and switched to buying piecemeal. For AI, this moment is increasingly reminiscent of the final phase of the dotcom revolution of the early 2000s: the technology of the Internet was already pointless to challenge, but that didn't mean that every company with the right presentation deserved a premium to the traditional grading scale.

Investors are now asking important questions: How much capital do we need to sustain growth (and how much more will we need to invest)? Where exactly is the future revenue coming from: infrastructure, applications, data, services? How much of the effect of implementation is cost savings and how much is potential revenue growth. And most importantly, when will the forecasts of these effects become visible in company reports?

If we look at what's happening not through indices, but through investor reaction to reports and forecasts, we can form a rather applied list of criteria for winners in AI sectors.

The first group is "shovels and picks", i.e. infrastructure for AI. We are talking about companies that sell capacity and "hardware" without which the AI economy simply does not scale: chips, servers, equipment for data centers. Demand for them is easier to measure and verify: there are deliveries, there are contracts, there is capacity utilization, and this is usually visible before application services have mass "success cases". Typical examples are Nvidia, which sells GPUs for training and running models, Arista Networks, which supplies networking equipment for data centers, Equinix, a data center operator.

The second group is companies where AI is already affecting financial results: revenue, expenses and profits. The market is more willing to pay when it sees the effect of AI implementation in numbers: conversion increases, customer churn decreases, operations speed up, efficiency and profitability increase. Meta Platforms is often cited as an example (although analysts warn that the company is close to the "bear market" phase), where algorithms and AI in recommendations and advertising are directly reflected in monetization. Among them is Amazon, which lost almost 5.5% of its capitalization over the last week - in this company, AI affects both the efficiency of logistics and the revenue from cloud services.

The third group are companies that will be able to withstand price pressure and maintain their margins. AI dramatically accelerates competition: something becomes easier to repeat, and customers start demanding more results for the same money. Therefore, the winners are those who have the margin margin margin, control over the cost of computing, and the ability to build AI in a way that adds value to the product.

Examples of such "enduring" players are Alphabet, Google's parent company, which has strong infrastructure, data and distribution. Another is Adobe, which can embed AI into Creative Cloud and hold the price thanks to its ecosystem and user habit. Shares of cybersecurity company CrowdStrike added 8.6% in value last week during the tech sector's market crash. In cybersecurity, AI is improving the quality and speed of threat detection, and it is difficult and expensive to change vendors for large customers.

Regardless of which group a company of interest to investors is located in, the major change in 2026 for it is simple: the market is no longer debating whether AI will be everywhere, it's evaluating exactly who will make money from it and at what cost.

This article was AI-translated and verified by a human editor