US-China AI rivalry expands globally as countries weigh competing technology ecosystems

The competition between the United States and China for artificial intelligence leadership is rapidly becoming a battle over which country’s technology will power governments, businesses and digital infrastructure around the world. Washington is promoting complete American AI technology packages built around advanced semiconductors, cloud infrastructure and leading models, while Beijing is expanding the reach of lower-cost and often open models that can be adapted more easily by countries unable or unwilling to depend entirely on expensive American systems.

That difference in strategy is becoming increasingly important as developing economies begin deciding what their national AI infrastructure should look like. The United States retains major advantages in advanced chips and frontier models, but Chinese developers including Alibaba Group Holding, Moonshot AI and MiniMax have gained substantial international adoption through relatively inexpensive open systems. OpenRouter data cited in recent reporting indicate that Chinese models’ share of usage on the platform surged from less than 15% roughly a year ago to more than 54% last week.

The emerging landscape is not developing into a simple technological divide in which every country chooses Washington or Beijing. Governments from Southeast Asia to the Middle East are increasingly combining American hardware with Chinese models, seeking access to the strongest technology available while avoiding excessive dependence on either superpower.

Washington is exporting an entire American AI stack rather than individual technology products

The United States has moved beyond treating artificial intelligence exports primarily as commercial transactions between technology companies. The Trump administration’s American AI Exports Program explicitly encourages US companies to form consortia capable of offering foreign partners complete technology packages combining semiconductors, data centers, cloud infrastructure, models, cybersecurity systems and technical expertise.

The Department of Commerce says the program is designed to expand international adoption of secure American AI systems while reducing foreign dependence on technologies originating from countries Washington considers strategic competitors. The White House has simultaneously presented AI infrastructure as a diplomatic tool, arguing that countries can retain control of their own data and domestic applications while building on American technology.

Washington is also attempting to address one of the largest barriers facing developing countries: cost. New financing initiatives through American development agencies and international institutions are intended to help partner governments afford AI infrastructure that might otherwise remain financially inaccessible, while a technology-focused Peace Corps initiative is designed to provide technical support during deployment.

The strategy reflects recognition that technological leadership is no longer measured solely by who develops the most advanced model. Global influence will increasingly depend on which systems become embedded inside national governments, healthcare networks, financial systems, universities, factories and telecommunications infrastructure.

Once a country builds data centers, trains developers and restructures digital services around a particular technology ecosystem, switching to a different system can become costly and disruptive. That gives early AI adoption decisions long-term strategic consequences similar to earlier battles over telecommunications networks, operating systems and internet infrastructure.

China’s open-source AI strategy is expanding adoption despite US chip restrictions

China is pursuing a different route to global influence. Rather than relying primarily on proprietary frontier systems, major Chinese developers have released powerful open or open-weight models that companies can download, modify and deploy on their own infrastructure.

That approach has helped Chinese models gain adoption even while Washington restricts China’s access to the most advanced American-designed AI chips. A US congressional advisory body warned earlier this year that Chinese open-source AI was creating a self-reinforcing competitive advantage, with models from companies including Alibaba Group Holding, Moonshot AI and MiniMax rising rapidly across global usage rankings.

Chinese President Xi Jinping has explicitly connected that technological model with Beijing’s international strategy. At the World Artificial Intelligence Conference in Shanghai, Xi promoted open-source AI as a way to broaden access to advanced technology and pledged training and technical cooperation for developing nations, while China launched the World AI Cooperation Organisation as a platform for countries seeking a greater role in AI governance.

Lower costs are a major advantage. Governments attempting to introduce AI into education, agriculture, healthcare or public services may lack the resources required to purchase large quantities of expensive computing capacity or pay recurring fees for proprietary models.

Chinese models can therefore appeal to markets where affordability, customization and local control matter as much as absolute frontier performance. That does not eliminate concerns about security, censorship, data governance or Chinese state influence, but it helps explain why US export controls alone have not prevented Chinese AI from spreading internationally.

Many countries are combining American chips with Chinese models instead of choosing one side

Perhaps the most important development is that governments are finding ways to mix technologies from both ecosystems. Saudi Arabia provides a clear example: its state-backed AI company has developed an Arabic-language model using technology from Chinese startup MiniMax while simultaneously signing major infrastructure agreements involving US companies such as NVIDIA Corporation.

Similar patterns are appearing across Latin America, Southeast Asia and Central Asia. Countries may prefer American semiconductors and networking technology for data centers while selecting Chinese open models for particular applications because those systems are cheaper or easier to customize locally.

This approach complicates Washington’s attempt to establish a largely American technology stack among partner countries. Even governments with strong political or security relationships with the United States may resist becoming dependent on a single supplier if competing Chinese technology offers useful capabilities at lower cost.

American companies themselves have also experimented with Chinese models. Businesses ranging from technology firms to consumer platforms have evaluated open Chinese systems because developers can modify them internally instead of relying entirely on externally hosted proprietary services.

That does not mean countries view American and Chinese technology as interchangeable. Washington maintains strong advantages in advanced semiconductors, high-end networking equipment and frontier cloud infrastructure, while Chinese providers can face greater scrutiny over data security and government influence.

The emerging pattern is instead one of selective dependence. Countries may purchase American computing infrastructure, Chinese models and locally developed applications simultaneously, creating national AI ecosystems that cannot easily be classified as belonging entirely to one geopolitical camp.

Alibaba, Moonshot and other Chinese developers are narrowing parts of the technology gap

China’s AI expansion is also being supported by substantial domestic investment. Alibaba Group Holding recently unveiled its Zhenwu V900 AI chip and announced plans for a next-generation model containing between 5 trillion and 10 trillion parameters, while targeting more than 20 gigawatts of Alibaba Cloud data-center capacity by 2032.

The company says it intends to build across the entire technology stack rather than relying indefinitely on foreign suppliers. Such investments are particularly important because American export restrictions have limited Chinese access to the most advanced NVIDIA Corporation chips.

Chinese developers have responded partly by emphasizing efficiency. If models can deliver competitive performance while using fewer high-end chips, restrictions on computing hardware become less effective as a mechanism for slowing adoption.

American companies have noticed the shift. Reuters reported that US technology groups are renewing investment in open-weight models partly because Chinese systems have demonstrated that inexpensive and customizable AI can attract developers even when proprietary American systems remain stronger in some frontier tasks.

That competitive response could ultimately benefit users by increasing the number of available models and pushing providers to reduce costs. It could also accelerate AI development further as both countries attempt to make their ecosystems more attractive internationally.

AI safety is becoming one area where Washington and Beijing may still cooperate

The commercial rivalry is intensifying at the same time that both governments are acknowledging risks created by increasingly capable AI systems. US Treasury Secretary Scott Bessent recently proposed a mechanism through which Washington and Beijing could notify one another about major AI safety incidents and discuss threats involving critical infrastructure or national security.

The proposal was discussed ahead of Donald Trump’s meetings with Xi Jinping and reflects concern that neither country benefits from uncontrolled AI systems capable of creating major cybersecurity, biological or infrastructure risks. China has developed its own regulatory approach focused on technical standards, state oversight and keeping advanced systems under human control.

AI safety therefore presents an unusual combination of competition and shared interest. Washington and Beijing want their domestic industries to lead global development, but both have reasons to prevent dangerous models from triggering incidents that could destabilize financial systems, communications infrastructure or military decision-making.

Building meaningful cooperation will remain difficult because the same technologies that create safety concerns also have enormous strategic value. Governments may hesitate to share information that reveals the capabilities or weaknesses of their most advanced systems.

Even limited communication could nevertheless become increasingly important as AI becomes integrated into cybersecurity, military operations and national infrastructure. The United States and China competed intensely during their recent summit but still discussed potential mechanisms for managing extreme AI risks, demonstrating that technological rivalry does not eliminate every area of potential coordination.

NVIDIA and Alibaba shares show investors are pricing different parts of the AI race

Financial markets continue to assign substantial value to companies positioned on both sides of the competition, although recent stock movements show mixed short-term sentiment. NVIDIA Corporation closed at $225.07 on September 25, rising 0.22% for the session and approximately 1.3% from its September 18 close as investors continued to value the company’s dominant position in advanced AI computing.

US-listed shares of Alibaba Group Holding closed at $109.74, down 0.8% for the session and roughly 3.1% from September 18 despite renewed investor attention around the company’s AI chip and model announcements. Alibaba shares had received a separate boost earlier in the week following the unveiling of its expanded AI strategy, illustrating how investors continue to respond positively to signs that the company can build more of its own technology stack.

The differing performance does not establish which country is leading the AI race. NVIDIA remains positioned around the scarce computing infrastructure needed to train and run advanced models, while Alibaba combines cloud services, models, chips and consumer applications inside China’s broader technology ecosystem.

Investor sentiment therefore reflects different business models as much as geopolitics. The long-term commercial question is whether global AI adoption remains concentrated around expensive frontier infrastructure or increasingly shifts toward cheaper, customizable models that can run across a wider range of computing environments.

Global AI leadership may ultimately depend more on adoption than headline model rankings

The next stage of the US-China AI competition will increasingly be decided outside Silicon Valley and Beijing. Governments across Africa, Latin America, Southeast Asia and the Middle East must determine how quickly they adopt AI, what infrastructure they can afford and how much technological dependence they are willing to accept.

The United States offers powerful chips, leading frontier models, established cloud providers and a large network of technology companies. China is countering with lower-cost open models, expanding domestic hardware and an international strategy emphasizing accessibility and customization.

Neither approach guarantees dominance. Security concerns could restrict Chinese technology in some markets, while cost and licensing structures could limit the reach of American systems elsewhere.

The more likely outcome is a fragmented global AI landscape in which countries combine technologies according to their own economic and strategic priorities. That would make AI competition less about forcing governments to choose one superpower and more about which companies can become indispensable across multiple layers of the technology stack.

For Washington and Beijing, that distinction matters enormously. Building the strongest model may generate headlines, but embedding technology inside another country’s schools, hospitals, businesses and government systems can create influence lasting much longer than any individual product cycle.

Key takeaways from the widening US-China global AI competition

• The United States is expanding an AI export strategy that packages American chips, cloud infrastructure, models and technical services for international partners.
• China is competing through lower-cost and often open AI models that governments and businesses can customize without relying entirely on proprietary services.
• OpenRouter data cited in recent reporting show Chinese models rising from less than 15% to more than 54% of model usage on the platform within roughly a year.
• Countries including Saudi Arabia are already combining American computing infrastructure with Chinese models rather than committing exclusively to one technology ecosystem.
• Alibaba Group Holding, Moonshot AI and MiniMax are among the Chinese developers expanding rapidly despite US restrictions on access to advanced semiconductor technology.
• Washington retains important advantages in high-end chips and cloud infrastructure, while China’s cost and open-model strategy can be particularly attractive across developing markets.
• The two governments remain intense technology competitors but are also exploring limited cooperation around major AI safety incidents and risks to critical infrastructure.
• Global AI leadership may increasingly depend on widespread adoption and infrastructure deployment rather than simply which country produces the highest-performing frontier model.