Skip to main content
Advertisement
Live broadcast
Main slide
Beginning of the article
Озвучить текст
Select important
On
Off

The capitalization of the largest US technology corporations continues to break records, supporting the entire American stock market. At the same time, the only driver of this growth is investors' belief that the multibillion-dollar costs of artificial intelligence will bring huge profits in the near future. At the same time, Chinese developers are releasing one neural network model after another, and they are much cheaper than American ones, and they are also open-source. This highlights the key problem of the US AI sector - its huge cost in terms of both costs and stock price. Izvestia investigated whether Chinese breakthroughs could lead to a new crisis in the stock markets.

How much is a token for the people?

In the AI economy, the basic unit of calculation is a token, a piece of text or code processed by the model. Nvidia CEO Jensen Huang called the token a new commodity, since it is through it that all AI services for developers and corporate clients are charged, while ordinary users usually pay a fixed price for a subscription.

Until recently, the main part of the market was represented by chatbots that worked on the "question–answer" principle. The consumption of tokens was relatively low. But now the industry is switching to agency-based artificial intelligence. AI agents don't just generate texts, they independently perform complex multi-stage tasks: they write program code, conduct market research, manage databases, and interact with external applications without human intervention.

Сотрудник
Photo: IZVESTIA/Polina Violet

This transition is accompanied by a rapid increase in the consumption of computing resources. According to joint calculations by researchers from MIT (Massachusetts Institute of Technology), Stanford and Google DeepMind, it takes up to 3.5 thousand people to complete a single complex agent task. There are several times more tokens than for a simple logical response in a chatbot. A normal dialog can cost a couple thousand tokens, whereas the work of an autonomous agent writing software or analyzing the market instantly spends millions. AI agent developers are already spending hundreds of millions of tokens daily.

The cost of 1 million tokens is becoming a determining factor for any technology business. And it is in this field that Chinese developers offer conditions that make American models less competitive. AI labs from China are trying to keep up with American ones in productivity, while noticeably surpassing them in cost and openness. At one time, DeepSeek blew up the markets, but now the successes of the Chinese have become routine.

A parade of models

In June 2026, the Beijing Zhipu Laboratory (also known as Z.ai ) introduced its latest GLM-5.2 system, stating its intention to make advanced intelligence publicly available. Almost simultaneously, Moonshot AI released the Kimi-K3 model, and technology giant Alibaba announced a preview of Qwen 3.8.

All of these releases are very impressive. In terms of the number of parameters — the degree of complexity of the model obtained in the learning process — Chinese novelties surpass most well-known analogues. Kimi-K3 has 2.8 trillion parameters, making it the world's largest open source AI system. Alibaba's Qwen 3.8 operates 2.4 trillion parameters. The American industry leaders, OpenAI and Anthropic, prefer to keep this data of their systems secret.

But the main difference is in pricing. 1 million output tokens for the GLM-5 model.2 from Z.ai It costs $1.92, while the Anthropic requires $50 for a similar volume for its flagship Claude Fable 5 model. The difference exceeds 25 times. An even more aggressive policy is being pursued by the DeepSeek laboratory, which has announced a 75% reduction in token prices for its API. Its DeepSeek-V4 Flash model offers prices at which the use of the Chinese system costs 1/34 of the cost of American competitors.

Доллары купюры
Photo: IZVESTIA/Anna Selina

The research company Artificial Analysis conducted stress testing, in which the leading models performed 657 complex office and administrative tasks requiring a huge expenditure of tokens. The results were remarkable. Completing these tasks on the Opus 4.8 model from Anthropic cost $1 thousand. Using GLM-5.2 from Z.ai it cost $270. The DeepSeek-V4 Flash model coped with the entire amount of work for just $14.

At the same time, the latest Chinese model Kimi-K3 not only showed outstanding cheapness, but also surpassed Fable 5 from Anthropic in terms of the number of successfully completed tasks without errors, costing the customer three times cheaper.

Shot in the leg

The spread of Chinese software is supported by the actions of the American administration itself. In an effort to protect its technological superiority, the White House banned non-US residents from using the advanced Claude Fable 5 model from Anthropic on June 12, 2026. For the first time, access to commercial AI was blocked by a government decision. The reaction was the opposite. Companies around the world have realized the danger of being completely dependent on AMERICAN proprietary systems, access to which can be revoked at any time. Against this background, open Chinese models, whose parameters are publicly available for download and local launch, look like a safe harbor.

Even American corporations are starting to rethink their strategies. According to the Ramp payment platform, there has been a sharp increase in the number of companies paying for DeepSeek services in the United States. Microsoft is already exploring the possibility of integrating Chinese models into its flagship Copilot office assistant due to the exorbitant costs of using American systems.

ИИ
Photo: IZVESTIA/Eduard Kornienko

Beijing skillfully uses the price advantage of its models for geopolitical influence. At the World Conference on Artificial Intelligence in Shanghai, Chinese President Xi Jinping announced the creation of AI cooperation centers within six regional organizations covering almost the entire Global South. For developing countries seeking modernization but without billions of dollars in budgets, cheap Chinese AI with open source becomes the only solution.

This trend is also confirmed by the scientific community. The publication in Nature indicates that researchers around the world are massively switching to Chinese open models (Qwen, GLM, DeepSeek). They are powerful, free, or very cheap, and can be run on local hardware, which completely removes research teams from U.S. export controls.

For Russia, which is under unprecedented sanctions pressure and cut off from American cloud services, Chinese open sourcing has become a key factor in the survival of the domestic AI industry. Unable to legally use closed models from OpenAI or Google, Russian developers and large corporations are actively downloading open weights of Chinese systems, training them on their own data to create sovereign corporate platforms. This allows us to maintain the pace of digitalization of the economy in conditions of severe hardware limitations.

The trillion dollar collapse

However, Chinese models still have vulnerabilities. According to research, they are less effective in processing complex logical tasks that require days of offline analysis, and sometimes they spend more tokens to "think about" the answer than their American counterparts. Nevertheless, for 90% of routine office tasks, their capabilities are already more than enough.

Сервер
Photo: IZVESTIA/Sergey Lantyukhov

For Wall Street, this situation is a serious thunderstorm. The market capitalization of American tech giants is inflated by expectations that they will be able to sell AI at high prices in an attempt to recoup their trillion-dollar investments in Nvidia chips and the construction of data centers. If China makes basic intelligence a publicly available and very cheap commodity, the profitability of the American AI business will go into negative territory, and the dominance of Silicon Valley may be undermined. Considering that the entire US stock market is now supported by technology corporations, the scale of the collapse may be prohibitive. In January 2025, when the release of the DeepSeek R1 model wiped $1 trillion from American stock markets in one day, it could have been just a rehearsal for this scenario.

Despite the seriousness of the threat, the likelihood of a total disaster for the American AI industry, and therefore (nowadays) the stock market, is somewhat exaggerated. There are a number of barriers that will be almost impossible to overcome in many cases.

The first is corporate security and regulation. Large Western businesses (banks, medical corporations, and the military-industrial complex) never integrate the models created in China into their closed circuits. The risks of hidden bookmarks, data leaks, and outright bans from U.S. and EU regulators completely cut off Chinese open source from the premium corporate segment of developed countries. If someone wants to do this, they can get a shout-out from Washington. Even the most powerful corporations are by no means omnipotent and are afraid of political leaders, especially when it comes to national security issues.

Ноутбук
Photo: IZVESTIA/Eduard Kornienko

The second limitation is the ecosystem. American giants offer customers not just a neural network, but a deeply integrated environment: cloud storage, cybersecurity systems, office suites (such as Copilot integration into Microsoft products). Customers pay for a seamless experience and technical support, not just for the raw generation of tokens.

At the same time, interesting conflicts may arise outside the United States. Given that relations between Americans and a number of European countries are far from ideal, many of them will at least try to ignore the "urgent recommendations" of the United States. Who would give up an obvious benefit? And this will only increase transatlantic economic tensions.

Переведено сервисом «Яндекс Переводчик»

Live broadcast