The release of the DeepSeek chatbot in early 2025 sent shockwaves through the global tech community, drawing immediate comparisons to the Soviet Union's Sputnik launch in 1957. This "DeepSeek moment" sparked debate about whether China was rapidly closing the artificial intelligence gap with the West. However, a closer examination of the foundational pillars of AI progress—compute, algorithms, data, and institutional environment—reveals significant structural challenges that may prevent China from leading the AI-driven industrial revolution.
The AI Stack: A Staggering Compute Gap
At the heart of AI advancement lies computing power, or "compute," which is critical for developing algorithms and processing data. Here, the disparity between the United States and China is profound. The US controls approximately 75% of global AI computing power, dwarfing China's share of about 15%. This gap is widening due to exponential scaling in the US and severe constraints on China, primarily from export embargoes on advanced chips and funding limitations.
This compute advantage extends to cloud infrastructure, the utility layer for AI services. In Q2 2025, the trio of Amazon Web Services, Microsoft Azure, and Google Cloud commanded around 63% of the global cloud market. Chinese providers, including Alibaba Cloud (4%), Tencent Cloud (2%), and Huawei Cloud (2%), collectively held just 8%.
Beyond the Spectacle: Algorithms, Talent, and Data Realities
While DeepSeek's open-source model created a perception of a breakthrough that defied the "scaling law," evidence shows US frontier models continued substantial improvement by following it. The next frontier is developing AI with a "world model"—understanding physical reality—which is key for advancements like AI-controlled robotics. Success here depends on compute and top-tier talent.
The talent landscape is revealing. Although many elite AI researchers are Chinese and work on leading US teams, major international AI awards are rarely, if ever, granted for breakthroughs achieved at institutions within China. In a geopolitical climate resembling a Cold War, the sharing of disruptive breakthroughs is unlikely, potentially leaving China permanently behind due to compute restrictions and limited researcher mobility.
Regarding data, the US benefits from a vast corpus of English-language scientific publications and internet text. While China excels in surveillance and video collection, its reliance on censorship and bureaucratic silos undermines data quality and shareability. For world-model research, high-quality multimodal data from healthcare, industry, and the physical world is essential—another area where the US holds an advantage.
The Demand Deficit: Why Industrial Revolutions Need More Than Supply
Critically, industrial revolutions are fueled by both supply-side innovation and robust, sustained demand. Historically, they have emerged only within advanced democratic capitalist systems. The US, with the world's highest per-capita GDP and high labor costs, generates powerful, economy-wide demand for automation and innovation, underpinning the stellar performance of its AI stocks.
China faces a contrasting reality. Its economy is grappling with weak consumer demand, overcapacity, high unemployment, and persistent deflation—conditions fundamentally incompatible with the demand-driven cycle of an industrial revolution. AI-led automation cannot solve these institutional problems. Furthermore, massive government borrowing to finance AI and chip ambitions exacerbates concerns over severe debt and soft budget constraints, echoing Soviet struggles during the Cold War arms race.
The user and revenue metrics highlight another divide. OpenAI reports ChatGPT now has over 800 million weekly active users, with most growth occurring after the DeepSeek moment, and has reached $12 billion in revenue. In contrast, China's leading models (Ernie Bot, Qwen, DeepSeek) report between 10-150 million monthly active users, primarily domestically, often at low or no cost. DeepSeek's free, open-source strategy limits its ability to collect high-quality interactive user data, which is vital for algorithmic refinement. Scaling inference services globally remains extremely compute-intensive, a major headwind for Chinese AI amid trust, funding, and geopolitical issues.
Ultimately, sustained innovation requires free institutions, risk-taking investors, open debate, and market competition. In China, where the Communist Party "leads everything," demand is suppressed by state control of resources, limiting household income and entrepreneurial initiative. Capital is funneled into state-directed projects rather than open-ended discovery. While a "DeepSeek moment" can capture global attention, fostering a genuine, lasting industrial revolution is a different challenge altogether. AI is not a cure for deflation, and deflation is anathema to revolutionary economic transformation.
Di Guo is a visiting scholar and Chenggang Xu is a senior research scholar at the Stanford Center on China's Economy and Institutions at Stanford University.
