Expensive AI development drags companies into huge debts. They no longer have enough income. Of the 5.5 trillion dollars that are planned to be spent on AI, as much as 4.1 trillion will be borrowed. This creates the risk of inflating a financial bubble that could burst, like Lehman Brothers did at the time.
Technology companies have overtaken banks in terms of the share of credit risk in the US market and could plunge the world into a new financial crisis, Bloomberg reports.
The six largest technology companies now carry 8.4% of the total credit risk in the US corporate bond market, which is more than the six largest banks (7.3%), according to Barclays analysts. In other words, the influence in the economy is shifting from the financial sector to the technological one, which creates new risks for the entire system.
Previously, AI companies used to finance projects out of their profits, but now the scale of investments forces them to actively borrow. In total, JPMorgan estimates that of the $5.5 trillion planned to be spent on AI, about $4.1 trillion will be borrowed. Five technology giants – Amazon, Alphabet, Meta* (recognized as extremist and banned in Russia), Microsoft and Oracle – plan to spend more than $1 trillion on development next year. At the same time, their free money is shrinking: for Alphabet, it went into negative territory for the first time, and for Meta* it fell by 91%.
Company stocks are rising, but the debt market is often the first to notice problems.: The current tightening of loan conditions may be temporary, or it may be the beginning of a serious crisis.
Investors who lend money to AI companies have become more cautious and are demanding more payment for their risks, Bloomberg writes. Analysts recall the railroad boom in the United States in the 19th century, when the collapse of financing led to the financial panic of 1873.
"The comparison with the railway boom of the 19th century is not entirely relevant: this example is too archaic. A closer example is the dot-com bubble at the turn of the 2000s. Many companies disappeared then, but some of the players survived the collapse and continued to grow. Probably, even now many companies and business models will leave the market," says Alexander Firanchuk, senior researcher at the Gaidar Institute's International Trade Laboratory.
There is a risk of overestimating investments in AI, he agrees. "In the case of rapidly developing technologies, it is always difficult to assess how they will affect productivity and what the return on investment will be. At the same time, many tend to invest first, which creates the possibility of a bubble. Correction in the assessment of AI is possible, but when it will happen and whether it will be abrupt, it is almost impossible to predict," notes Firanchuk.
Why did IT companies start accumulating debts, meaning they invested more than they earned?
"Previously, IT giants really lived on their money, but now the scale is completely different - we are talking about trillions of dollars for the construction of data centers and the purchase of chips. There is no longer enough profit for such a scale, the alternative is to borrow while conditions are still favorable.",
– notes Yulia Katasonova, Senior Director of ratings of non-financial companies and ESG ratings of NRA.
"Previously, investments in AI were relatively small and they were easily repelled by subscriptions and licenses. But now, for the normal operation of large language models, an appropriate infrastructure is needed, for example, data centers, in order not only to store data in the cloud, but also to create the basis for the functioning of neural networks of future generations. And for data centers to function, you need the appropriate super-powerful computers, for computers - super-powerful chips, and for all this you need investments in electricity, since data centers consume a lot of it. Hence the growing debts of the AI sector, especially the American one," explains Natalia Milchakova, a leading analyst at Freedom Global.
According to her, AI really makes it easier for a person to solve many tasks, since it is already able to perform at least routine functions, helps enterprises save on costs, freeing up unskilled office staff.
"AI today is a loyal assistant to man, not his enemy. It is a tool for increasing business efficiency, sales efficiency and a possible increase in market share. Thanks to AI, enterprises and financial institutions are already reducing costs by 25-40%.
According to McKinsey estimates, from 2025 and in subsequent years, the contribution of AI to global GDP is about 1.2% annually," Milchakova notes.
Russia is also keeping pace with the trend in terms of implementing AI, the expert believes. A significant number of Russians use the neural networks "Alice" from Yandex, GigaChat from Sberbank and their analogues. In 2026, 55 subjects of the Russian Federation have already implemented AI under the smart cities development program. AI is actively integrated into the Gosuslugi service, and "unmanned" transport with AI is developing.
"Russia is also creating its own production facilities for chips and software for neural networks, and many large corporations are already building data centers or announcing plans to build them in the next three years. By the end of 2025, the volume of the Russian AI market reached 316.1 billion rubles (an increase of 29.7%), and the double–digit growth trend is likely to continue in 2026," Milchakova concludes.
Olga Samofalova
* The organization(s) have been liquidated or their activities are prohibited in the Russian Federation
