Why China can’t win the AI-led industrial revolution
The People’s Republic was founded on the principle that the Communist Party of China “leads everything.” That remains true today.
AI is widely recognised as the core technology in an emerging industrial revolution that will probably transform every facet of the global economy. UN Trade and Development (UNCTAD) estimates – conservatively – that the global AI market will reach $5tn (€4.3tn) by 2033, thanks to average annual growth of about 31%. The International Monetary Fund predicts that the technology could boost global GDP by 4% over the next decade, with the United States gaining as much as 5.4%. AI’s impact on science, innovation, the military, and geopolitics is already significant, reinforcing the sense that the race for AI dominance is also a race for global dominance.
In this context, the Chinese startup DeepSeek’s release of a highly competitive chatbot caused a sensation in early 2025. Dubbed the “DeepSeek moment,” it immediately prompted analogies to the Soviet Union’s launch of Sputnik in 1957. But do such spectacles really mean that China is closing the gap with the West?
In considering that question, it is important to bear in mind that no industrial revolution has ever emerged outside advanced democratic capitalism. This is no accident. Like its predecessors, the AI-driven industrial revolution requires robust institutions to ensure secure property rights, enforceable contracts, the ability to attract and empower talent, efficient allocation of resources, and – crucially – sustained demand. The last element is often overlooked in analyses of China’s progress in AI.
The People’s Republic was founded on the principle that the Communist Party of China “leads everything.” That remains true today: The CPC controls courts, markets, banks, universities, and the media, and even commands private firms. Under such powerful party-state rule, the regime can mobilise massive resources and produce shining stars like DeepSeek (or Sputnik, in the Soviet case). An industrial revolution, however, depends on more than isolated breakthroughs; there must be a series of disruptive innovations in technology, business models, and institutions that build on one another. The Soviet experience makes this clear. The USSR and its satellites in Eastern Europe could not keep up with the West during the third industrial revolution, and this failure eventually contributed to the collapse of their communist regimes.
To understand where China stands in the AI race, we must examine some basic facts. Technically, advances in AI rest on three foundational elements: compute, algorithms, and data. Among these, compute (raw data-processing power with advanced chips) is arguably the most fundamental, since it determines the capacity to invent and develop algorithms and generate or process data.
The compute gap between the leading US and Chinese large language models is staggering. At the national level, the US controls about 75% of global AI computing power, compared to China’s 15%. And the gap continues to widen as AI compute scales up exponentially in the US, while China’s compute remains severely constrained by export embargoes and a lack of funding.
Compute also underpins AI customer services through cloud platforms, which are considered infrastructural utilities for the sector. Here, too, the US dominates. In the second quarter of 2025, Amazon Web Services, Microsoft Azure, and Google Cloud together held about 63% of the global market, while Chinese providers held only 8%, with Alibaba Cloud at 4%, Tencent Cloud at 2%, and Huawei Cloud at 2%.
With respect to algorithms, open-source AI models help narrow the information gap by making technical resources widely accessible. DeepSeek capitalised on this approach, and the initial shock it caused stemmed from a perception that it was a genuine technological breakthrough: an algorithm that can perform well without following the “scaling law,” according to which model performance follows compute and data resources. But reality proved otherwise. The damage that DeepSeek’s debut did to Nvidia’s valuation lasted only a few days. Hard evidence shows that US frontier models continued to improve substantially by following the scaling law.
Looking ahead, a major goal in the pursuit of artificial general intelligence algorithms is the development of AIs with a “world model,” meaning the capacity to understand and learn from physical reality, rather than solely from statistical patterns embedded in language. Among other things, major breakthroughs in world-model research will lay the foundation for AI-controlled robotics. Since this leg of the race remains in its early stages, no one can predict the eventual winner. But we can examine the two prerequisites for achieving successive breakthroughs. We have already discussed compute; the other is talent.
Here, the data are revealing. While many of the most talented AI researchers are Chinese – constituting a significant share of leading AI teams in the US – it is exceedingly rare, if it has happened at all, for major international AI awards to be granted to researchers whose breakthroughs were achieved at Chinese institutions. This might not matter too much if essential breakthroughs remain open-source, and cross-border mobility continues. But with Sino-American relations assuming more of a Cold War-like character, disruptive breakthroughs are unlikely to be shared, and China – hampered by compute restrictions and limited mobility – could fall permanently behind.
With respect to data, the US benefits from a substantially larger collection of publications – especially in science and technology – and internet text, not least because these sources are overwhelmingly in English. Though China does excel in surveillance and mass video collection, its reliance on censorship, political control, and bureaucratic silos tends to undercut data quality and shareability. In terms of pursuing world-model breakthroughs, high-quality multimodal datasets – health-care, industrial, and physical-world data – are essential, and the US has the advantage. Moreover, additional data generation and processing depend heavily on compute, thus reinforcing the US advantage.
These comparisons show that the tech engines of AI progress – compute, algorithms, and data – are deeply interdependent. While China has made rapid advances, its structural constraints remain significant. To assess its place in the next industrial revolution, one must consider two critical issues. First, industrial revolutions are always driven by both supply and demand. Wealth and strong demand for successive innovations are indispensable. That is why industrial revolutions have occurred only in advanced economies operating under democratic capitalist systems.
With the world’s highest per capita GDP and correspondingly high labour costs, the US generates strong demand for automation across all sectors, including in research and development itself. It is these macro fundamentals that have underpinned such strong performance in US AI stocks.
By contrast, China’s economy has been trapped in a vicious cycle of weak demand, overcapacity, high unemployment, and persistent deflation, which is fundamentally incompatible with any industrial revolution. AI-led automation offers no remedy for such problems, which are rooted in the country’s institutional foundations. The massive government borrowing used to finance China’s bid for AI and chip dominance has only deepened concerns about its already severe debt burden and chronic soft budget constraints – problems reminiscent of what the Soviet Union faced during the Cold War arms race.
Moreover, AI services are essential for creating a virtuous cycle between revenue and development of the AI ecosystem, as well as for generating the interactive data that are so critical for continuous algorithmic improvement. Here, we find another stark gap: OpenAI reports it has surpassed 800 million weekly active users worldwide for ChatGPT, and most of these users showed up after the “DeepSeek moment.” In terms of revenue, OpenAI has reached $12bn (€10.2bn), and all leading US AI companies have been growing at 300% annually.
Meanwhile, China’s leading models – Baidu’s Ernie Bot, Alibaba’s Qwen, and DeepSeek – have so far attracted 10-150 million monthly active users each, mostly within the domestic market and often at very low prices or for free. DeepSeek’s open-source, free-download strategy may help expand its ecosystem, but it reduces opportunities to collect high-quality interactive user data. Ultimately, increasing one’s market share requires scaling inference services (processing a query) to users, which is extremely compute-intensive. Compute restrictions will remain a major headwind for Chinese AI models seeking a global user base, especially when one also considers issues such as a lack of trust, insufficient funding, and geopolitical constraints.
Sustained innovation requires free institutions and robust demand. Breakthroughs come when entrepreneurs and scientists are empowered by independent courts, supported by risk-taking private investors, and tested through open debate and market competition. In CPC-controlled China, demand is suppressed because the state controls key resources that limit household income and entrepreneurial initiative, and capital is funnelled into state-directed projects rather than open-ended discovery and innovation. While a “DeepSeek moment” may capture our attention, achieving long-term competitiveness and fostering a genuine industrial revolution is another matter entirely. After all, AI is not a remedy for deflation – and deflation itself is fundamentally incompatible with any industrial revolution.
Di Guo is a visiting scholar at the Stanford Center on China’s Economy and Institutions at Stanford University. Chenggang Xu is Senior Research Scholar at the Stanford Center on China’s Economy and Institutions at Stanford University.
CONNECT WITH US TODAY
Be the first to know the latest news and updates
