China's early success in global AI competition, bolstered by continued massive state investment and other advantages, could help it extend its dominance in international markets for manufactured goods to the software realm.
China is the world's manufacturing powerhouse. In 2024, it exported $3.6 trillion worth of manufactured goods—about as much as the combined exports of the United States ($2.1 trillion) and Germany ($1.6 trillion). Yet, in services, China's story is different. Its $384 billion in services exports in 2024 ranked eighth globally but is only about one-tenth the value of its goods exports—the lowest ratio among the world's top 10 services exporters. Germany, the next lowest, exported services that equaled 28 percent of its goods exports.
The biggest shortfall for China lies in software and IT services, which make up just 18 percent of its total services exports. That compares with 30 percent for the United States and 44 percent for Germany.
Analysts chalk up China's underperformance to two factors: limited English-language skills among Chinese developers, and, more debatably, a long-standing engineering culture that favors hardware over software.
The biggest shortfall for China lies in software and IT services, which make up just 18 percent of its total services exports.
AI Could Be the Great Equalizer
That inequality may be about to change because of AI, although it is still over the horizon. Historically, new hardware—whether personal computers or smartphones—has triggered industry cycles by introducing new capabilities, while software has determined how effectively those capabilities are used. Software innovation ultimately drives the industry.
At this early stage of the AI industry cycle, hardware still dominates. In 2023, global data center hardware revenues reached $230 billion, with the United States taking 41 percent and China 28 percent. AI software and services, by contrast, brought in $160 billion. Over the next decade, both markets will grow but software will grow faster. Hardware revenues are forecast to exceed $1 trillion by 2033 (growing 15.8 percent annually on average), while software and services could soar to $2.5 trillion—an annual growth rate of 31.7 percent. The United States and China will continue to be the two largest providers of hardware, together holding about half the global market.
The surprise lies in where software will be developed. In 2023, China's AI software and IT services output was a modest $5.4 billion—just 3 percent of the global market. By 2033, it is expected to leap to $327 billion, giving China the largest share of the global market—about 13 percent. That's a 50 percent annual growth rate, compared to the United States' projected increase from $55 billion to $320 billion over the same period (a 19 percent annual growth rate).
The Machinery Behind the Rise
China's rise in AI is anchored by a vast, state-engineered network—a web of venture funds, data centers, startups, universities, and tech giants—fueled by roughly $210 billion in state capital over the past decade. Both central and local governments have poured money into AI infrastructure and research, following an industrial policy model that has powered earlier leaps in solar panels, telecoms, and electric vehicles.
That model has coalesced into the ambitious “AI+” initiative, announced by the State Council, which aims to weave artificial intelligence into nearly every major branch of China's economy—manufacturing, health care, transportation, finance, and energy—by 2027. Provincial governments, such as Shanghai, are translating this mandate into practice through large-scale subsidies for computing clusters, domestic model training, and start-up incubation. State-owned enterprises are piloting dozens of real-world uses—from fatigue-damage detection on jetliners and power-grid inspections to medical-AI platforms in hospitals. Software is a natural focus of this initiative.
Government support is a well-tested approach in China that can count many hardware successes—such as solar panels and electric vehicles, but no notable software successes. Will it work in AI software development? The record of software development in China is, on the whole, poor. This is not for want of trying. For example, in the year 2000, having missed the Y2K resolution business, Chinese government money poured into software startups, most of which failed.
English is no longer the gatekeeper of software innovation. With large language models now fluent in nearly all major world languages, coding and application development can thrive in Chinese as easily as in English.
There are three reasons to believe that a different outcome is probable in AI. First and most important, English is no longer the gatekeeper of software innovation. With large language models now fluent in nearly all major world languages, coding and application development can thrive in Chinese as easily as in English. Second, China's enormous domestic market generates an abundance of data that give its AI developers an unmatched reservoir of training material drawn from hundreds of millions of users and pervasive digital transactions.
Third, China has global ambitions for AI to be achieved through its Belt and Road Initiative. Since 2017, AI has been central to the BRI's digital component, termed the Digital Silk Road initiative. Chinese firms are now building data centers across Southeast Asia and Africa, providing the backbone for both business and government systems. For countries like Malaysia—where capital is scarce but ambitions for AI run high—these data centers help power new applications using Chinese LLMs for local industries. In the process they deepen reliance on Chinese software, systems, infrastructure, and standards.
Of course, even as China forges ahead, the rest of the world will not stand still. All developed countries and some fast-growing developing countries such as India are implementing national AI strategies. Yet, China's early start, massive state investment, and the three drivers noted above give it a unique advantage. China's dominance of the factory floor may soon extend to the office floor.
