The AI rally in the American market illustrates the story of hyperconcentration of capital, the distortion of the market structure and the growing dependence of indices on a narrow circle of the largest innovative companies. Shares of AI players have become the main driver of growth in the Nasdaq technology index, but at the same time the main source of vulnerability. When a handful of big names move the entire broad market, the index’s formal liquidity masks its real fragility and vulnerability.
Goldman Sachs estimates that Magnificent Seven provided more than 60% of the S&P 500’s total return in 2023, and their sales are expected to grow by an average of 12% per year until 2026, against 3% for the remaining 493 companies in the index. At the same time, Nvidia in 2025 became the largest weight in the S&P 500, about 7.5%; that remains about the same to this day, which in itself shows how the market is tied to several AI stories. This is an important detail. When several securities move the entire index, the market formally remains wide, but in fact it becomes much less stable and prone to sharp movements on the news or a sudden change in the mood of large investors.
This is where the main question arises. How sustainable is the current growth if the companies’ valuation already reflects not only future revenue, but also an almost ideal AI implementation scenario? This is especially noticeable in the example of market leaders, where investors pay not for the current monetization, but for the right to be inside the next technological order. The problem is that the technological structure and investment profitability do not always coincide in time.
Regarding how exactly to perceive the danger of the current situation in the context of hyperconcentration, I will say the following. First of all, it should be noted that the concentration has always been to one degree or another. And everyone always talked about its potential danger – but it was akin to, you know, shouting “wolves, wolves.” Another question is that if you seriously and honestly study this issue now, you will have to draw an unpopular conclusion that there are some alarm bells, and their volume is growing. That is, for example, there is OpenAI, or a ChatGPT product, which at first was generally some kind of magical revelation, but now it faces competition and is steadily losing the market.
According to the latest data, Anthropic, one of the leaders of the AI breakthrough, reported that during the new round of financing it managed to raise $65 billion, and its market value was $965 billion. With this assessment, Anthropic overtook OpenAI and became the most expensive startup in this area. It should be noted that this is not yet a public company. Hence the main question: what will happen to its market valuation next? At the same time, the company is burning money at a frantic pace for capex (capital investment). Accordingly, this is one of the main indicators of what exactly in the context of balance sheet stability will happen in the AI segment in the coming months, that is, this “bubble” will burst or succumb to control and avoid this fate.
There is a resemblance to the notorious 2000 dot-com bubble, but it is neither direct nor mechanical. Then the market overestimated the speed of commercialization of the Internet, today it can overestimate the speed and scale of the return on AI investments. That is, many companies, whose estimates after the dot-coms fell catastrophically, subsequently grew simply into giants – for example, Amazon.com. They still represent a meaningful share of the Nasdaq – both then and now. Another question is that their assessment has changed. At this stage, I wouldn’t expect some kind of dotcom level crash.
Nevertheless, we will most likely see a correction in short horizons.
Moreover, there are several reasons for her. First, these are the same fundamental assessments, the long-term distortion of which is always fraught with an orderly or chaotic adjustment of “common sense.” Returning to our AI segment as an example, on the one hand, it really marks a breakthrough in improving the efficiency of many activities. However, this raises the question of how this increase in efficiency correlates with the huge investments that have already been made there and which need to be invested even more, and in what time horizon they will pay off, and whether they will pay off at all. In this context, the OpenAI example looks relevant. That is, the company continues to increase the rate of internal investments, the team continues to work and receive a salary, and the product is already lagging behind several competitors.
Therefore, the first: investors need to be prepared for the fact that some individual product – not the entire segment as a whole – may itself be unsuccessful. Second: it can increase efficiency, but not enough to justify a gigantic investment in its development. That is, here are two main types of tests for the AI industry in the future that you need to constantly think about.
In addition, to declare that AI is some kind of golden key from all doors looks like an exaggeration at this stage. AI gives good aggregation and provides fast processing of some (not all) tasks on large amounts of data. On the other hand, if we are talking about algorithmic strategies and when we want to optimize parameters non-linearly, AI will cope with this best. However, if you set the task for AI as follows – “please find me an inefficiency that can be profitably traded on the basis of simply raw market data,” then it is extremely doubtful that AI is capable of such a practical analysis at this stage.
Another argument that the market is overheated is related to how algorithmic and institutional strategies work at the current stage. In the superheated phase, they do not so much extinguish the risk as they can intensify it. I myself launched an algorithmic trading strategy, which seemed to us very deep and stress-resistant, and it really was quite complicated. As a result, it worked for six months, and then we obviously did not have time to retrain the network, and it stopped providing the expected results. That is, stability and multimodality alone does not guarantee constant profit. In this sense, I remain a skeptic about the popular thesis that, they say, AI can demonstrate something so supernatural that it will directly turn the market and our entire way of life.
Speaking of signals that could indicate the inevitability of the start of the correction, the first signal worth watching is the possible loss of a stable market share by key AI platforms, primarily OpenAI. If the current leaders begin to concede noticeably to little-known competitors, the market will quickly reconsider the idea of the inevitability of super profits in the sector, based on their recognition alone.
The second is the general geopolitical uncertainty, which is getting closer to the recession in the largest economies in the world. For example, let us not forget that Iran still keeps the Strait of Hormuz, through which a fifth of the world’s oil supplies pass, closed. The question is that oil, if it is set at levels of about $150 per barrel, can be considered as a self-sufficient factor in the inevitability of an economic downturn.
Look at 2022, which, after the enchanting 2020 and 2021, and we had a lot of such years in the last decade, just turned into a big disappointment. These are risks that investors may now underestimate. In addition, in my opinion, important macro factors are also underestimated – in particular, that even regardless of geopolitical scenarios, a recession is possible in the European economy and in the economies of Asian countries, many of which are highly dependent on oil supplies, which suddenly became expensive. Offhand, these are the risks that do not look taken into account by the markets (underestimated, as we say).
AI already shows a typical retraining problem. On high-frequency and system strategies, it can work very well, but only as long as the data structure is stable. As soon as the market mode changes, the model begins to lose quality. This illustrates well the main risk of the current rally: the market may be right in the idea, but wrong in the scale and timing of monetization.
By Kirill Kuchinsky, financier, PhD (Econ.)