Caixin Global reported on September 11 that a growing number of Chinese AI developers are accelerating their shift toward overseas markets in order to hit profitability targets. The report cited an analysis by Jiang Xiaojuan, a professor at the University of Chinese Academy of Social Sciences and a former deputy secretary-general of China's State Council, who said that because of the limited size of the domestic market, many Chinese AI firms find it hard to achieve commercial profitability relying on domestic customers alone, and are being pushed — or are choosing — to expand into global markets for new customers and revenue. Industry observers say the shift reflects a new realignment in the global division of labor in the tech industry, and also underscores the continued rise of China's technological innovation capacity and industrial competitiveness in AI.

Jiang Xiaojuan made these remarks in a speech on September 9 at the opening of the 2026 Inclusion·Bund Summit in Shanghai's Houtan World Expo area. Held under the theme "Co-creating the New AI Economy," the forum drew more than 600 scholars, scientists and entrepreneurs from around the world. In her speech, Jiang said, "In the AI era, all innovative products and services are born global" — meaning that for Chinese AI companies, technology development, product design, business models and even compliance systems must face global markets from the moment they are created, rather than following the traditional path of "fully developing the domestic market first, then considering going global."

Jiang argued that while training a large AI model is costly, once training is complete, the marginal cost of applying it is extremely low — giving AI products a built-in economic logic for global expansion. At the same time, she said, the spread of open-source models has further leveled the competitive starting point among different market participants. She stressed in particular: "A model, once trained, must be used globally — only then can its business model possibly become profitable." In other words, relying on a limited domestic market alone makes it very difficult to spread out the high training and computing costs of a large model; companies need a global user base to break even or turn a profit.

Jiang cited data in her speech showing that this year, downloads of Chinese open-source large models on major global open-source communities surpassed those from the United States for the first time, and that the top six models on the usage-ranking charts for July and August were all Chinese-developed open-source models. She said this shows that China's open-source large models are becoming a global "technology public good," giving Chinese AI companies a favorable channel for reaching overseas developers and enterprise customers quickly through the open-source ecosystem.

At the same time, fierce price wars and homogeneous competition in China's domestic large-model market have objectively made it harder for companies to turn a profit. According to industry observers, the so-called "Four Little Dragons" of Chinese AI — DeepSeek, Zhipu AI, MiniMax and Moonshot AI — have shown starkly different paths to commercialization in 2026. Zhipu listed on the Hong Kong Stock Exchange in January, becoming the first large-model company to go public, and its valuation surged sharply after listing, though the company still faces continued losses; MiniMax was the first among the group to become self-sustaining, with steadily improving gross margins; Moonshot AI's annualized recurring revenue (ARR) doubled within two months; and DeepSeek stands out for its highly competitive cost structure but has yet to establish a stable, closed-loop business model. Analysts say it is precisely this intense domestic competition and limited willingness among domestic users to pay that is pushing these companies to more actively seek new growth space overseas.

In practice, DeepSeek, leveraging its highly cost-effective models, has at times climbed to as high as second place on global AI app download charts, and formally entered the Middle Eastern market through cooperation with a Saudi Aramco digital data center. MiniMax's social companion app Talkie has accumulated more than ten million downloads in markets outside mainland China since its launch, making it one of the representative cases of Chinese AI apps going global, while Zero One Everything's productivity app PopAI has likewise built up several million overseas users. Industry reports also note that the destinations for Chinese AI companies' overseas expansion are extending beyond the traditional European and American markets into Southeast Asia, the Middle East and other emerging markets, where they compete with Western rivals on different terms — lower API pricing, a rich accumulation of application scenarios, and fast iteration.

Addressing the new challenges that come with Chinese AI firms' accelerating globalization, Jiang Xiaojuan said that in this era of "native globalization," companies need to build a global compliance system from the moment they are founded, covering data compliance, business licensing, global taxation and other dimensions, to ensure that products meet the high compliance standards of major markets from birth. She also recommended that relevant authorities push forward institutional reforms and adopt more flexible fiscal subsidy arrangements, to give companies more adaptive policy support as they expand abroad — helping Chinese AI firms navigate different markets' regulatory and compliance challenges while preserving their technological and cost advantages. She added that for the new generation of AI companies, compliance with multilingual, data, algorithm and content rules, along with regulatory requirements that differ from market to market, will likely need to be built into product design from the outset, rather than fixed after the fact.

Several industry analysts say that the accelerating global push by Chinese AI companies essentially reflects a new shift in the global division of labor in the AI industry: as China continues to make progress in basic research capability, industrial support systems and the accumulation of application scenarios, Chinese AI companies are now capable of competing head-to-head with the world's leading firms, rather than merely absorbing and integrating Western technology. This trend is also seen as an important signal of the rise in China's technological innovation capacity and industrial competitiveness, marking a shift for China's AI industry from being a "follower" to becoming, in some areas, a peer competitor or even a leader.

That said, analysts note that the path to globalization for Chinese AI firms is not without resistance, including continued tightening of export controls on Chinese AI chips, computing power and related technologies by the United States and other Western countries, stricter scrutiny of data security and algorithm compliance in some markets, and the lack of a unified global AI regulatory framework — factors that could still constrain the pace of Chinese AI firms' overseas expansion. Looking ahead, industry observers broadly agree that whether Chinese AI companies can strike a balance between global compliance, technological iteration and business-model innovation will determine whether they can truly convert the scale advantages of overseas markets into sustainable profitability.