The AI race between China and the U.S. pits China's state-driven, top-down approach against America's private sector-led innovation. While China invests heavily in subsidies and practical applications, the U.S. focuses on breakthrough technologies and funding advantages. The outcome will test which model—centralized control or decentralized markets—better fosters AI dominance.
Key Takeaways
•China employs massive state subsidies and a top-down strategy to boost AI, aiming for economic revitalization and global leadership.
•The U.S. relies on private sector innovation, capital markets, and technological expertise, with a focus on achieving artificial general intelligence.
•China faces challenges including chip shortages, financial system constraints, and adoption hurdles, while U.S. models currently outperform Chinese counterparts.
•Chinese AI firms excel in open-source models and practical applications, such as integrating AI into consumer products and manufacturing.
•The race highlights a broader competition between centralized autocracy and decentralized democracy, with high stakes for technological, economic, and geopolitical influence.
The competition between China and the United States is about more than technology. Illustration by The Atlantic. Sources: Javier Zayas Photography / Getty; piranka / Getty.Government officials in Hangzhou have grand ambitions to make their city in eastern China a global center for artificial intelligence—and the funds to try to make it happen. In June, they pledged $140 million to subsidize AI firms that operate in town. Not to be outdone, Shanghai promptly followed in July with its own $140 million subsidy program, and inaugurated an “AI innovation town” two months later with low-cost office space for start-ups in the sector. In the south, Shenzhen was already doling out $70 million a year to support local AI firms and research, while Chengdu, in the west, invested $42 million in a start-up called Zhipu AI to bring a new model-training center and research facility to the city.
This frenzied spending spree follows a playbook that has proved successful in many industries in China: The state acts as cheerleader, financier, and protector, uniting the country’s bureaucrats, executives, and entrepreneurs in a mission that Beijing believes is vital to China’s future.
China’s top-down approach has made the country the envy of much of the world when it comes to industries such as manufacturing and infrastructure construction. But AI—and the technological innovation it demands—is something different. “The Chinese government is struggling to figure out how to support” the AI sector, Paul Triolo, a partner at the advisory firm Albright Stonebridge Group who specializes in Chinese technology, told me. The U.S. government, in contrast, “is trying to get out of the way and create an environment in which capital markets and these very innovative companies can run with the ball.”
That makes the contest for AI not just between engineers and algorithms but between China’s centralized autocracy and America’s decentralized democracy. It pits American private enterprise and its immense financial resources, talent, and creative energies against a Chinese government determined to dominate this new technology at any cost. The future of AI is too uncertain to know just yet which approach will win out. But as things stand, even the most disciplined and dedicated policy makers in China may struggle to compete with the wealth, expertise, and experience of America’s private tech sector.
Given the ways in which AI promises to reshape nearly every aspect of our lives, the stakes for this race are high. Whichever country claims the lead could gain an edge not just technologically and economically, but also diplomatically, militarily, and in any area that relies on ingenuity.
In China, however, this race feels existential. After decades of robust growth, the country’s investment-led economic engine is running on empty. The Communist Party plainly hopes that AI technology will offer a solution to the country’s economic malaise—a way to restore growth without resorting to the kinds of reforms that might loosen the party’s grip on power. With its society aging and public discontent mounting, Chinese leaders are keen to put AI to work. If their strategy fails, the divisions between the party and public may deepen, leading to even greater repression and a more uncertain future.
All AI companies are racing to develop smarter large language models and the expensive infrastructure to support them. But in the U.S., the strategy is primarily to build bigger and more powerful models as rapidly as possible, to hasten both the adoption and the profitability of AI. The ultimate ambition is to achieve “artificial general intelligence,” whereby a machine can match the cognition and problem-solving of a human brain.
In China, the prevailing assumption is that AI models are already good enough to deliver benefits now. In August, the State Council, China’s top governing body, pledged to promote “the broad and deep integration of AI across all industries and areas of the economy and society” by 2035. China’s leadership wants every civil servant, lab scientist, factory manager, corporate executive, and army general to be harnessing AI asap.
In the U.S., the hope is that artificial general intelligence “will provide a cross-sectoral advantage across the economy, across the military,” Scott Singer, a fellow who focuses on technology at the Carnegie Endowment for International Peace, told me. “For China, it’s much more about how a conventional, meaningful boost to your economic, and perhaps military, productivity creates competitiveness.”
China’s leaders have had their eyes fixed on AI for nearly a decade. In 2017, the State Council introduced a national plan for AI development with the intention of “making China the world’s primary AI innovation center.” The Chinese government has spent at least $200 billion over the past decade on supporting the AI sector. Adding related state funding for chips and other industries probably puts this sum at more than $300 billion, according to estimates from the Council on Foreign Relations using data from a study published by the National Bureau of Economic Research. China now presents the only real threat to American leadership in AI.
This kind of state muscle is great for making steel and cars, but it doesn’t necessarily work when it comes to conceiving and adopting something as complicated as AI. “It’s much harder for the government to almost mandate or just push through adoption at a large scale” in AI versus other industries, Jeffrey Ding, a specialist on China’s AI sector at George Washington University, told me. When it comes to infrastructure, “you can build the bridges; you can invest in high-speed rail.” But to get people to use AI systems, “it has to make sense for them from a profit perspective, from a market perspective.”
More perniciously, China’s leaders have also refused to liberalize the country’s antiquated financial system. Bureaucrats have stunted the development of stock markets, and their distrust of private enterprise has scared off investors and venture capitalists. This means that Chinese AI companies cannot raise the kinds of funds that their American competitors do with ease. According to an analysis by Ding based on data from a Chinese research institute, four top U.S. tech giants—Google, Microsoft, Meta, and Amazon—have together invested over eight times more in data centers and other infrastructure necessary to support AI than China’s seven leading internet companies combined. And state funding doesn’t appear to be making up the difference.
China’s AI sector is further hobbled by Washington’s curbs on selling advanced American AI chips to Chinese companies, because China’s semiconductor industry is not yet capable of producing equivalent alternatives. Over the objections of many national-security experts, President Donald Trump has allowed the U.S. semiconductor giant Nvidia to sell more AI chips to China, but he has maintained the ban on the most powerful American chips, at least for now. Chinese efforts to boost its own chip industry have fallen short.
These constraints place China’s AI firms at a disadvantage. Experts generally agree that American AI models perform better than their Chinese competitors. Without sufficient investment or chips, the Chinese industry may never catch up. If an American firm does create an AI model that rivals human intelligence “using breakthrough technologies that others don’t know, then they may really squash the rest of the world,” Kai-Fu Lee, the chairman of the venture-capital firm Sinovation Ventures and an AI expert in China, told me.
But the race is far from over. Despite these considerable obstacles, Chinese AI firms have proved adept at keeping pace with their richer U.S. peers. DeepSeek startled Silicon Valley last year with a model that rivals ChatGPT but built with far less money and computing muscle. The Chinese AI industry has also chosen to be “open source,” which means that the source code of models from DeepSeek and Alibaba is freely available to study, improve, and share, and is cheaper to use than U.S. alternatives, which are largely kept private.
“The American approach is like, I’m a genius; I’m going to win the Nobel prize; I’ll invent something and beat everyone to it,” Lee said. “The Chinese approach is We’re all good students; we’re not geniuses; we’re going to do our homework together.”
Chinese entrepreneurs and engineers are channeling their efforts into more practical applications of AI that promise a wider reach. About 70 percent of users of a Chinese AI-video generator called Kling, for example, are outside China. Chinese companies are also moving quickly to capitalize on the country’s vast manufacturing capacity by deploying AI in everyday objects, such as cars, eyeglasses, toys, and transcription devices. Tom van Dillen, the managing partner of the Beijing-based technology consulting firm Greenkern, told me that 17 of the top 20 car brands in China have integrated DeepSeek’s AI into their models. “The ability to take a lot of these new technologies and put them into use in creative ways is superior” in China than in the United States, he said.
Given the questions that plague the future of AI, what may seem like clear-cut comparative advantages may prove not to be. Perhaps China’s aggressive state spending, open-source AI coding, and interest in ordinary use cases for the technology will ultimately give the country an edge over American firms that are expensively chasing revolutionary applications for a more theoretical product. Perhaps the sheer diversity of AI and its potential applications makes it wrong to conceive of this competition between the U.S. and China as a winner-takes-all binary.
Or perhaps this race will indeed reinforce which political and economic model best enables technological innovation: China’s aggressive meddling or America’s intensive enabling. The U.S. would seem to have the upper hand here, but it would be unwise to count China—and the Communist Party—out.
