- Статьи
- Economy
- Cognitive Monopoly: how to live in a world dominated by the United States and China in AI
Cognitive Monopoly: how to live in a world dominated by the United States and China in AI
The competition between China and the United States in the field of artificial intelligence (AI) development continues to escalate. In June, the Chinese company Zhipu released the GLM 5.2 model, which was second only to the leaders from ChatGPT and Anthropic, surpassing the Google models. The second sensation, as in the case of DeepSeek, did not happen — investors reacted calmly to the news. Nevertheless, it is becoming obvious that only two truly powerful players remain in this market - the United States and China. Izvestia investigated whether the rest had fallen behind forever and what niches in the global economy of the future other countries might get.
Strength and weakness
At the moment, we can say that the leaders of the AI race have already turned into hegemons, and each of them has its own strengths. The American approach relies on enormous computing power and unlimited access to cutting-edge equipment. Corporations from the United States are solving scaling problems with tens of thousands of the latest chips. The Chinese industry operates under severe export restrictions. The lack of advanced hardware has forced developers from China to focus on mathematical and architectural optimization.
When DeepSeek's R1 model collapsed the capitalization of the NASDAQ index by $1 trillion in early 2025, the main argument was the stated cheapness of her education and work. However, the audit revealed the specifics of this cost-effectiveness. The new Chinese GLM 5.2 model offers a token price 10 times lower than American competitors (1 million DeepSeek V4 tokens cost $0.87, while the American Fable 5 has a value of $50). But there is a difference in the algorithms: the Chinese model spends 23 times as many tokens to complete one complex task. Given the lower quality of the final result, the actual cost of solving corporate problems using Fable 5 may even be more profitable. In addition, there is a difference between open and closed tests: in the first of them, the Chinese show almost half the gap.
At the same time, the American industry is now facing serious infrastructural barriers. AI training requires volumes of electricity comparable to the consumption of average states. In the United States, the development of data centers is being blocked by social opposition: polls show that 70% of Americans are categorically opposed to building data centers near their homes. The Chinese government is able to allocate land and build a power plant for the needs of a data center in a few months. And in general, China's energy capacity is now higher.
Monopoly on the march
The world economy knows examples of monopolies and duopolies. The mainline passenger aircraft market is divided between Boeing and Airbus. The field of advanced microchips is controlled by Taiwan's TSMC and the Netherlands' ASML under the patronage of the United States. The difference is that aviation or navigation systems are industry-specific technologies. Artificial intelligence is a general-purpose technology that automates cognitive labor.
The current volume of the direct AI market is estimated at about $500-550 billion, with total ecosystem and infrastructure costs exceeding $2.6 trillion. By 2036, the baseline scenario assumes the growth of the direct market to $7-9 trillion. In this configuration, the United States and China are forming a mechanism for collecting global digital rents. Control over the underlying models and computing infrastructure allows them to set standards for the entire planet, which is much more serious than dominating the oil market.
This process poses serious challenges for both developed and developing countries. For a long time, emerging markets have attracted investments due to cheap labor. The introduction of AI leads to large-scale automation of routine intellectual and service work. For advanced economies, the situation is even more dangerous, as they do not even have the advantage of cheapness.
"One of the effects of using AI is labor automation, in this regard, other countries lose the benefits of cheaper labor and productivity growth," says Vladislav Bukharsky, Associate Director for Sovereign and Regional Ratings at Expert RA agency. — The only limitation here is the unique human abilities: empathy, building relationships and ethical decisions are not yet amenable to algorithmization.
Adaptation strategy
For the remaining two hundred countries, the formation of an AI duopoly means a change of strategy. Trying to create your own frontier-level basic models is pointless because of the prohibitive cost of such projects.
— For medium-developed countries, the task of catching up with the United States and China is not worth it, it is expensive. Rather, their task is to create opportunities to adapt existing global solutions and industry—specific models to the needs of key sectors of the economy," notes Vladislav Bukharsky.
Local IT companies in Europe, India, or Latin America will rent access to American APIs or use open source Chinese models, training them to meet the needs of local banks, clinics, and government services. The success of this strategy directly depends on the availability of its own computing base. Every developed economy requires cloud clusters of data centers equipped with AI accelerators located on its own territory.
"Any AI developer should be able to connect to the facilities for a fee and train the systems in relative proximity, which is important for speed of work," explains Kirill Chernovol, a researcher at the Gaidar Institute's Laboratory for the Analysis of Best International Practices.
According to an OECD pilot study conducted in 39 countries, China (27) and the United States (22) are the absolute leaders in the number of such cloud clusters from international providers.
A strong middle peasant
Russia occupies a specific niche in this "bipolar world". The country has the necessary base to maintain the status of a technologically sovereign state of the middle level.
The main asset is an independent digital ecosystem built over the past twenty years and huge amounts of data in Russian. This allows us to train models with a deep understanding of the local cultural code and mentality.
The infrastructure base also demonstrates resilience. According to Kirill Chernovol, international providers are not represented in Russia, but the situation looks competitive due to national companies. The five largest Russian providers have about seven clusters with AI accelerators, which puts the country in 3-4 place in the world, comparable to Singapore and Japan.
An additional advantage of Russia is an excess of energy and a cold climate. The regions of Siberia and the Far East offer low temperatures and affordable generation, which drastically reduces the cost of cooling server racks. Given the global shortage of electricity, these territories have every chance of becoming sites for hosting data centers serving the needs of foreign partners.
At the same time, the domestic industry is facing serious constraints. The hardware shortage caused by the sanctions prevents the purchase of industrial batches of the latest Nvidia B200 graphics cards to create frontier-level supercomputers. There is a lag in the regulatory environment. The draft law on the development of AI, which regulates critical issues of access of algorithms to personal data and copyright objects, was submitted to the State Duma only on June 25. A broad look at the infrastructure gives a less optimistic picture: if we count all existing clusters of data centers with AI training support, then the OECD rating.AI Index of 21 countries, Russia is among the last.
In the coming years, Russia will find itself in the role of an "intermediate" digital market. Domestic developments will lag behind advanced American systems by one and a half to two years. This margin of safety is sufficient to solve the vast majority of corporate and government tasks within the country. If the industry is allowed to develop normally, the economy will avoid the need to pay "tribute" to Washington and Beijing, while maintaining control over its own infrastructure.
Переведено сервисом «Яндекс Переводчик»
