Anthropic, the American artificial intelligence (AI) company that produces the Claude model, is preparing to go public. The market points to a market capitalisation between 1.5 and 2 trillion dollars, according to sources, although the final range has not yet been set.
Moonshot AI, the Chinese AI company that produces the Kimi model, one of the best in the country, is also preparing for a possible IPO, according to The Wall Street Journal. The last known private valuation of this company is around 50 billion dollars, according to Bloomberg, which is to say, between 30 and 40 times less.
The valuation gap is dizzying and even more surprising when considering that in January 2025, the performance of the Chinese model DeepSeek had shaken investors, who feared that the United States would lose its dominance in this field. Just two years later, the markets no longer seem to fear an imminent reversal in American leadership. So, has China lost the AI race?
In some respects, the American advantage is certainly very clear. Currently, Chinese models can only compete with American ones in certain tasks. In computer programming, for example, the gap between the two sides of the Pacific is practically non-existent: according to the LiveBench 2.0 test from 25 October 2026, DeepSeek V4.1 Flash scores 96 points, nearly tied with the 97.7 points of Claude Fable 5.1 from Anthropic. However, in long agency tasks, the best American systems approach 78 points, compared to 41 for the best Chinese model, in this case DeepSeek. A considerable difference.
American players have a second major advantage: their models are already integrated into a wide ecosystem of professional computing applications. If authorised, an agent connected to Microsoft 365, Salesforce, or Amazon Web Services, for example, can access the company's internal data to act. Chinese providers do not enjoy the same level of integration into Western companies' systems, which are often reluctant to trust them with their sensitive data.
Finally, the United States maintains a significant advantage in computational power. According to Epoch AI, the United States accounted for around 75% of the global potential of publicly identified large AI centres in May 2025, compared to 15% for China. It also benefits from preferential access to cutting-edge technologies: Nvidia accelerators, TSMC chips, state-of-the-art machines from ASML, and advanced memory manufactured by SK Hynix, Samsung, and Micron. China, on the other hand, faces restrictions that limit its ability to acquire this high-end equipment.
However, China is not doomed to play a secondary role in AI. Firstly, its models are much cheaper. According to LiveBench 2.0, a task performed with DeepSeek costs 0.18 dollars, compared to several dollars for the most advanced models from Anthropic.
On the other hand, China has a considerable advantage: the scale and speed of development of its electrical system. For years, the country has been massively investing in renewable energy, coal, and nuclear power plants, while the United States suffers a considerable lag in this type of infrastructure. This dynamism can sustain the expansion of AI, provided that China manages to overcome the challenge of transporting electricity on a large scale between its main production and consumption centres. According to the International Energy Agency's reference scenario in its April 2025 Energy and AI Report, the electricity consumption of Chinese data centres could grow by around 170% by 2030, compared to approximately 130% in the United States.
In summary, the currently uncontested American dominance in AI faces vulnerabilities precisely where China concentrates its main strengths. In many respects, the two countries seem like inverted mirrors in their strengths and challenges.
Thus, the duel between the two powers has not yet been definitively resolved. Judging by what China has achieved in the industrial sector over the past twenty years, it is perfectly possible that a more austere and less flashy strategy, but backed by a planning state as a crucial priority, could end up being competitive in the long term.