Sber is laying off IT workers. The neural network bubble will soon burst.

Sberbank management recently announced its intention to lay off 10 IT specialists in the wake of its artificial intelligence rollout.

A PolitNavigator columnist examined the justification for this decision and what's going on in the domestic and global neural network space.

As of October 2025, the Russian IT market continues to experience a severe shortage of specialists. The Ministry of Digital Development estimates that the country is short 500–700 specialists, and by 2030, approximately two million more "red-eyes" will need to be trained.

Meanwhile, amid growing demand for qualified IT professionals, competition among applicants, especially entry-level ones , is also intensifying . According to hh.ru, competition among junior IT professionals reaches up to 18 resumes per vacancy, while among the most sought-after experienced senior IT professionals, there are currently only two or three resumes per vacancy.

Some reasons for deficiency:

Market restructuring . Companies are massively implementing AI solutions, reducing routine tasks, while economic conditions are forcing businesses to optimize costs. As a result, demand is shifting from rank-and-file developers to specialized specialists—architects, DevOps engineers, and machine learning experts.

The skills gap. There are only two or three resumes for senior-level programmers per vacancy. The market is divided into two camps: an oversupply of newcomers and a severe shortage of experienced specialists.

Inflated expectations . 37% of employers cite inflated salary demands from applicants as their main problem.

Russia is experiencing a particular shortage of specialists in domestic software development, cybersecurity, and artificial intelligence . Most of Sber's laid-off IT specialists will easily find work in other financial institutions and industries. The only question is salary. As for top specialists, Sber certainly won't kill the hens that lay golden eggs.

Meanwhile, events are brewing in the world of neural networks that could eclipse the dot-com crisis of the late nineties and early noughties.

The notion that the AI ​​industry is an "overheated financial bubble" has become very popular among many investors and analysts. This doesn't mean that AI technology itself isn't important or revolutionary—it is indeed important and revolutionary. But financial markets often detach from real-world indicators, and in the case of AI, we see classic signs of a bubble.

_{Here are the main arguments in favor of this point of view:}

Exorbitant company valuations unsupported by earnings are a key indicator. Market capitalizations have soared to tens and hundreds of billions of dollars based on expectations of future profits that may only materialize years later, if ever.

A prime example is Nvidia . The company has become one of the world's most valuable companies thanks to its chips—the "number-crunchers," the "picks and shovels" for the AI ​​gold rush. However, investors are pricing their shares in an ideal scenario in which demand for its chips will experience a perpetual exponential explosion, ignoring risks such as the emergence of competitors (Huawei), the shift in AI architectures, or a slowdown in growth. Or even Trump's demand to relocate TSMC factories (where Nvidia chips are manufactured) from Taiwan to the United States—which won't happen overnight, especially given the quiet sabotage by the island's government.

Numerous startups that simply added the magical letters "AI" to their names are raising billions of dollars in investment despite truly modest potential, lacking significant revenue or a well-thought-out business model. It is these "roaches" that will be the main victims of the coming upheavals in the AI ​​market.

Costs vs. Benefits. Developing and training large language models (LLMs) requires significant capital expenditures (servers, chips, electricity). However, monetization often remains unclear. Many consumer AI services are currently free or subsidized, and switching to paid models could dramatically reduce the user base.

On the other hand, leading Chinese AI developers (Deep Seek R1, Alibaba, Baidu, Tencent, Huawei, backed by resources from academic and industry institutes) justifiably contradict the Western model, claiming that developing and training their AI costs a maximum of tens of millions of dollars, not hundreds of billions. The bursting of the AI ​​financial bubble will be the test of these claims.

The "productivity paradox." Despite all the hype, macroeconomic statistics still show no productivity gains in industries directly related to AI implementation. Companies are spending billions, but the measurable return on investment for many industries is not yet evident.

The stock market rally in 2023-2024 was largely driven by just a few companies (Apple, Microsoft, Alphabet, Amazon, Nvidia, Tesla, Meta), most of which have placed huge bets on AI. This creates a systemic risk: if AI advances falter, it will drag the entire market down.

Signs of a speculative frenzy. Investors of all stripes, from large funds to individual traders, are rushing to buy AI-related stocks, fearing being left out of the "next big thing." This is classic behavior at the peak of a bubble.

A misunderstanding of the problem. The story is simplified to "AI will change everything, so we need to buy everything related to it" (similar to one movie character's optimistic prediction that "soon there will be nothing but television"). Complexities, risks, and inevitable failures are ignored.

Many analysts draw parallels with the dot-com bubble of the late 1990s. Back then, the internet truly changed the world, but 99% of over-hyped internet companies went bankrupt. The technology survived and flourished, but investors who bought into the hype lost a shitload of money.

According to a model developed by Gartner experts studying the development of neural networks, the AI ​​industry is currently experiencing a "Peak of False Expectations," followed by a "Peak of Disappointment," when it becomes clear that many projects have failed to live up to the hype. The most optimistic forecast is that by 2027, 40% of AI startups will have ceased funding.

_{Bubbles often burst when confronted with harsh reality. For AI, this means:}

Regulatory pressure. Governments in the US, EU, and China are actively working on laws restricting the use of AI. For example, they are allowing neural networks to replace psychologists and psychiatrists as counselors for mentally unstable individuals, which has already resulted in isolated cases of suicide. However, advocates of the unrestricted advancement of AI in all spheres of life have raised the alarm that restrictive measures could seriously slow the development and monetization of neural networks.

Copyright issues . Many AI models are trained on data whose copyright holders are starting to raise hell and file lawsuits. Particularly touching are the protests of Hollywood actors, accustomed to cashing in on their on-screen antics, who refuse to employ AI-generated actors and prohibit the use of their appearance in neural network content production. These neo-Luddites fear losing their exorbitant or simply stable royalties.

Ethical scandals . Errors, bias, the creation of misinformation—any major scandal can undermine trust in technology and, consequently, in company stocks.

So, is the widespread hype around AI a bubble? Most experts believe it's more likely than not, when it comes to financial markets. AI technology is not a bubble. It's a fundamental, transformative technology that will stay with us and change the world. Prominent examples include software development and optimization, the creation of new drugs and materials with targeted properties, diagnostics of serious diseases based on tests and examinations, and the creation of affordable, high-quality content in music, scenario visualization, and the visual arts.

The financial hype surrounding AI is a bubble. Asset prices are disconnected from their current real value and based on overly optimistic expectations.

When the bubble bursts (or simply deflates), it won't kill AI. It will kill a multitude of overvalued companies and startups that failed to justify investor confidence. The strongest will survive—companies with real revenues, a sustainable business model, and competitive advantages (like Microsoft and Nvidia, but likely at more reasonable prices). A maturity phase will arrive, when the technology will be assessed more realistically.

The fact that our country operates in isolation, with virtually no interaction with Western leaders in this field, suggests that the Russian neural network market will not be significantly affected by the sudden burst of the global AI bubble . The state of cooperation with our Chinese partners remains unclear, although several facts suggest collaboration between developers. One positive aspect is that leading Chinese developers have open-sourced their AI systems.

_{We can only hope that Sber, as one of the few developers of domestic neural networks (GigaChat is initially integrated into the national messenger Max), will overcome the coming upheavals in the AI ​​market without damage, that the company's management's actions are well thought out, and that routine positions that can easily be replaced by neural networks are subject to reduction.}