The Diffusion Dividend: China’s Pragmatic Path to AI Integration
More than a year ago, DeepSeek dominated the AI conversation with the launch of their R1 AI model. They claimed to have developed a model capable of rivalling the likes of ChatGPT at a fraction of the cost and computing power. The extraordinary capital expenditure by hyperscalers appeared unjustified, and giant data centers with cutting-edge Nvidia chips were unnecessary. These worries spilt over into equity markets, and on the same day, Nvidia shed almost 17% of its market capitalisation. Sam Altman and Dario Amodei, OpenAI’s and Anthropic’s chief executives, respectively, shortly after claimed DeepSeek had illegitimately used their models to train R1. Nevertheless, DeepSeek attracted global attention to the Chinese AI industry and prompted international stakeholders to reassess their international competitiveness. How is China positioned in the global AI race?
China’s centralised approach to industrial strategy has proven highly beneficial to Chinese AI development. Since 2016, Chinese policymakers have dedicated resources to the technology, starting with investing in education. In 2022, China was already awarding 50% more STEM doctorates than the US and increasingly dominated global patent generation. The US has taken the lead in racing towards AI dominance since the publication of ChatGPT, and the current American flagship models are at the forefront of AI competency. However, Chinese companies such as Moonshot AI are closing in on the gap. An important factor that holds Chinese development back is their limited access to elite Nvidia chips, which are key for both training and inference. At the same time, this has driven the likes of DeepSeek to either make models work with less computing power or to circumvent the problem by offshoring to countries that have access. China is projected by Goldman Sachs research to produce sufficient energy by 2030, even considering a significant increase in AI energy consumption.
The American AI industry benefits from risk-tolerant investors and immense private investment. Private AI investment in China is not comparable, but the government’s involvement has always compensated for the lack of liquidity. Recently, there might be a shift in China.
In the Chinese AI industry, the majority of companies focused on developing groundbreaking technology are privately owned. This scarcity in public markets has driven demand by public market investors, who seek to bet on the AI race. Chinese investors seem increasingly eager to gain exposure to higher-risk AI companies as they are investing in companies with hundreds of times the forecast revenue over established tech companies. Amongst a series of examples, the Chinese start-up Zhipu, listed as Knowledge Atlas Technology, quadrupled its value to $30bn this year. Minimax, initially backed by Tencent and Alibaba, already doubled its share price after its IPO this year. Meanwhile, Alibaba and Tencent have seen a decline in share price this year, although they are seeing record engagement on their AI platforms. Companies have recently been offering gifts and rewards for downloading their AI products, which reflects the incredible competition in the Chinese market.
So, one distinguishing factor between American and Chinese AI is the respective private and public sources of capital. The main difference lies in America aiming for dominance and competency, whereas China is pushing for diffusion at large scale. The belief in pursuing diffusion is that AI products need not be the most competent on the market to satisfy common consumer applications. Additionally, focusing on diffusion rather than dominance requires a fraction of the capital, which can be reallocated to consumer AI products. Following DeepSeek’s example, with some exceptions, Chinese models, including flagship products, are open source. This has a series of benefits. It foremost creates a form of scientific community, such as DeepSeek, that is primarily backed by the hedge fund High-Flyer and can focus more on publishing and experimenting rather than securing a profit. Moreover, it is easier to export as open-source models are highly transparent and reduce the otherwise geopolitical risk associated with Chinese tech products.
AI can truly have a transformative effect on the Chinese economy. The technology is more likely to be adopted in the wider economy as other companies can integrate open-source models into their commercial products. This is especially true since it has the population with the most optimistic outlook on AI, reducing the risk in consumer markets. China is already a global leader in robotics and in integrating robots in industrial and manufacturing applications. Furthermore, they have an effective monopoly on essential materials for AI applications, including magnets, rare earths, lithium and are dominant in lithium and copper.
In conclusion, it appears that the US and China may not participate in the same AI race. We can also draw a parallel to the race to the moon during the Cold War. It seemed to be about military and geopolitical dominance, but it turned out to be incredibly beneficial to many other industries and advanced scientific development overall.
Yoon, J. (2026) ‘Scarcity value puts a rocket under China’s AI challengers’, Financial Times. Available at: https://www.ft.com/content/d83ca039-eba0-4449-8493-98fe744e79f7
The Economist (2026) ‘China’s DeepSeek year’, Drum Tower [Podcast], 17 February. Available at: https://www.economist.com/podcasts/2026/02/17/chinas-deepseek-year
Parikh, T. (2026) ‘China will clinch the AI race’, Financial Times, 18 January. Available at: https://www.ft.com/content/d9af562c-1d37-41b7-9aa7-a838dce3f571
Olcott, E. and Lee, K. (2026) ‘01.ai's Kai-Fu Lee: Why China will beat the US in consumer AI’, Financial Times, 4 February. Available at: https://www.ft.com/content/5d71bb69-25e0-4425-8ac8-635b7a8abb68