US China AI Race: Why China’s Open AI Strategy Matters

The global Artificial Intelligence race is no longer simply a contest between American and Chinese Technology companies. It is increasingly becoming a struggle over how AI should be built, distributed, regulated and accessed around the world. Washington is pushing countries to align with a US-led technology ecosystem, while Beijing is promoting a model that places greater emphasis on open systems, lower costs and broader access.

The competition has become more urgent as Chinese AI developers rapidly narrow the gap with their American counterparts. The United States still has major advantages, including leading AI research laboratories, advanced semiconductor technology, enormous pools of private capital and some of the world’s most capable frontier models. But China‘s progress over the past year has challenged the assumption that American technological leadership will remain uncontested.

US President Donald Trump has described the stakes in unusually stark terms, saying earlier this month that the United States is leading China in AI and that he wants to maintain that position because, in his words, “whoever wins AI wins.” His comments reflect how closely Washington increasingly links artificial intelligence with economic power, technological influence and national security.

Yet the definition of winning the AI race is becoming increasingly complicated. A country can lead in the development of the most sophisticated models without necessarily dominating global usage. Another country can produce cheaper systems that become widely adopted across developing economies and commercial markets.

That distinction is at the heart of China’s emerging AI strategy.

China Is Offering the World an Alternative AI Model

China’s pitch to international users is built around accessibility. While Chinese companies still trail the United States in some areas of frontier AI performance, developers in China have increasingly embraced open models that can be downloaded, customized and deployed without the same dependence on expensive subscriptions or foreign cloud infrastructure.

Companies such as DeepSeek and Moonshot AI have become important examples of this approach. For developers in emerging markets, the practical advantages can be significant. A model that is slightly less capable but dramatically cheaper may be more useful than a cutting-edge system that requires a costly subscription, restricted access or a cloud provider controlled outside the country.

Eric Olander, editor in chief of The China-Global South Project, has argued that this difference could matter particularly in the Global South, where affordability and accessibility can outweigh small differences in benchmark performance.

That creates a fundamentally different proposition from simply asking countries to choose the most advanced AI model available. For many governments, universities, startups and businesses, the question is not whether one model is technically superior in every category. It is whether the technology can be deployed at a scale they can afford and adapted to their own requirements.

China is also attempting to institutionalize its approach internationally.

In July, Chinese leader Xi Jinping launched the World Artificial Intelligence Cooperation Organization. The initiative represents an alternative to the US-backed Pax Silica alliance, which was established to reduce dependence on China in critical AI supply chains.

More than two dozen countries and the European Union signed up to Pax Silica, while Xi has recruited 29 countries, including Russia, Indonesia and Pakistan, around his alternative vision of AI cooperation.

The competing initiatives show that the AI race is expanding beyond laboratories and technology companies. It is also becoming a contest over international partnerships, supply chains, standards and technological influence.

Open AI Models Are Changing the Competitive Equation

China’s embrace of open AI systems was not necessarily the result of a simple top-down strategy. US restrictions on access to the most advanced chips created serious obstacles for Chinese developers, while China’s smaller capital markets limited some of the advantages available to the largest American AI companies.

Those constraints encouraged Chinese developers to search for different ways to compete.

Open models became one of those routes. Instead of relying entirely on massive proprietary systems and enormous computing budgets, Chinese developers could focus on efficiency, model optimization and architectures that could be distributed more widely.

The approach appears to have gained significant traction.

According to AI leaderboard data cited in the supplied analysis from OpenRouter, Chinese models accounted for more than 54% of global usage last week, compared with less than 15% a year earlier. DeepSeek has been a major driver of that expansion.

The growing use of Chinese models has not been confined to Chinese businesses. American companies, including Airbnb, DoorDash and Shopify, have also explored Chinese AI models, attracted by their lower costs and flexibility.

That behavior highlights a major challenge for Washington’s strategy of dividing the AI ecosystem into competing technological camps. Businesses do not necessarily make technology decisions according to geopolitical alliances. They often choose multiple suppliers based on price, performance, availability and the ability to customize systems.

“Rather than picking a side, companies – American or not – generally prefer to diversify their technology suppliers,” Olander said.

That diversification could give Chinese AI companies opportunities to expand through networks that already exist in developing markets. Chinese telecommunications, cloud and hardware companies have established significant distribution channels in many emerging economies, creating infrastructure that could help AI services spread alongside existing technology products.

The AI Race May Not Have One Winner

Another complication is that there is no universally accepted definition of what it means to win the AI race.

One measurement might focus on which country produces the most capable frontier model. Another could measure the number of users, the amount of computing infrastructure, commercial revenue, enterprise adoption or global Market Share.

Those measures could produce very different results.

Olander compared the situation to the smartphone industry, where technological leadership and profit leadership do not necessarily belong to the same company or country. Apple may have a smaller share of global smartphone users than some competitors while capturing a disproportionately large share of industry profits.

A similar pattern could emerge in artificial intelligence. The United States could maintain an advantage in the most advanced and profitable AI systems while Chinese companies accumulate a much larger global user base through inexpensive and accessible models.

The distinction could become even more important as AI models converge in practical performance.

For highly specialized applications, a frontier model that offers the highest possible reasoning, coding or scientific capabilities may remain essential. But many everyday users and businesses may not need the absolute best model. If a cheaper system can perform a task adequately, the performance gap may become less meaningful.

Chucheng Feng, founding partner of advisory firm Hutong Research, described the eventual decision as a Trade-off between price and performance: whether users would prefer a model costing many times more while being several months ahead, or a much cheaper model that trails by a similar period.

That is a particularly important question for developing countries. Limited technology budgets can make the difference between an AI system that is theoretically superior and one that can actually be deployed across schools, businesses, public services and local startups.

China’s Open AI Strategy Has Limits

China’s open-model advantage, however, depends on how open its AI ecosystem remains as its technology becomes more powerful.

At China’s major AI conference in July, Xi called for greater openness and cooperation while also emphasizing regulation and the need to keep AI under human control. Those two goals could eventually create tensions.

The more capable Chinese AI models become, the greater the strategic value of controlling how they are distributed and where they can be deployed.

Alex Colville, an analyst focusing on technology and security at the Australian Strategic Policy Institute, said Beijing wants to export Chinese technology to strengthen international influence, but national security considerations remain central.

That means China’s AI ecosystem may not evolve toward unlimited openness. Instead, the country could end up with a mixture of open and restricted systems.

For example, highly capable models with advanced cybersecurity or other sensitive capabilities could face tighter controls. Licensing rules could potentially be used to determine where particular systems can be deployed and by whom.

Such restrictions could weaken one of China’s strongest selling points.

Reva Goujon, a director focusing on China at the Rhodium Group, said the international appeal of Chinese AI depends heavily on continued access to models that can be adopted and fine-tuned for domestic and enterprise applications.

This creates an unusual strategic dilemma for Beijing. Keeping powerful models widely accessible can accelerate their international adoption and strengthen China’s technological influence. Restricting them can reduce security risks but potentially make the ecosystem less attractive to foreign developers.

Washington Is Increasing Pressure on Chinese AI

The competition has also intensified because of growing accusations that Chinese AI developers have benefited from American technology.

Anthropic and OpenAI have published reports alleging that Chinese AI laboratories used outputs from American models to improve their own systems, a process known as distillation. Anthropic has specifically described allegations involving DeepSeek and Moonshot, including claims that some customer requests were sent to Claude for purposes connected to training and improving competing systems.

Beijing has rejected the accusations, while DeepSeek and Moonshot did not respond to the request for comment described in the supplied material.

The dispute adds another layer to an already complicated technology relationship. Washington has imposed restrictions on advanced semiconductor exports to China and has considered additional measures affecting AI systems and related technologies.

At the same time, American technology companies do not necessarily support every restriction being discussed in Washington.

Nvidia, Meta, Microsoft and dozens of other companies warned against broad restrictions on open AI models in an open letter in July. Their position illustrates a growing tension between national-security policy and commercial interests.

US policymakers may view restrictions on Chinese AI as necessary to protect technological advantages and prevent sensitive capabilities from reaching strategic competitors. Technology companies, meanwhile, have commercial reasons to favor a more open ecosystem in which developers can use and build on different models.

Those competing interests make it difficult to establish a single US approach to open-source AI.

Trump and Xi Find Limited Common Ground on AI

Artificial intelligence was also a major subject during the Trump-Xi summit this week, against the backdrop of growing concerns about the risks associated with increasingly powerful AI systems.

Xi argued that the United States and China should draw on each other’s strengths rather than guard against one another. He also called for continued discussions about the risks and benefits of AI and cooperation to prevent misuse and abuse.

The comments followed preliminary discussions the previous weekend, when the two sides proposed an AI safety notification system and agreed to establish a formal dialogue on artificial intelligence.

But those developments have not produced a major breakthrough. The two governments remain divided over technology restrictions, national security and the broader direction of their relationship.

George Chen, chair of digital practice at The Asia Group consultancy, said low levels of trust continue to limit cooperation between the two countries. Beijing continues to view Washington’s technology policies as part of a broader effort to contain China’s rise, a perception that is likely to influence future AI engagement.

The result is a paradox. The United States and China both have strong incentives to discuss AI safety because neither country can completely isolate itself from the consequences of rapidly advancing technology. Yet the same technology has become a central element of their strategic competition.

Why Many Countries May Resist Choosing One Side

For countries outside the US-China rivalry, the choice may be less straightforward than Washington’s framing suggests.

A government seeking to build domestic AI capabilities may want access to American frontier models because of their performance and established developer ecosystems. At the same time, it may want inexpensive Chinese models that can be customized locally without creating permanent dependence on an expensive foreign platform.

Some governments may also prefer to use multiple suppliers rather than allow either the United States or China to dominate their AI infrastructure.

That approach could support the development of sovereign AI capabilities, allowing countries to adapt models to local languages, industries and regulatory requirements.

Chinese open models can therefore offer something strategically valuable beyond low prices: the possibility of greater control over how AI is deployed locally.

But countries adopting them will also have to consider the long-term availability of the technology, possible licensing restrictions, data governance concerns and geopolitical pressures from both sides.

The AI divide may consequently become less about choosing a permanent technological allegiance and more about managing a portfolio of systems from competing ecosystems.

The Global AI Race Is Becoming a Battle Over Access

The United States retains major strengths in frontier AI, advanced chips, investment and research. China’s rapid progress, meanwhile, demonstrates that technological leadership can be challenged through different strategies, including efficiency, open models and broad distribution.

The most consequential question may therefore not be which country develops the single most powerful AI model.

It may be which ecosystem becomes embedded most deeply into the everyday technology infrastructure of the world.

If American companies continue to dominate the most advanced models and capture the largest profits, Washington could retain substantial economic influence even if Chinese systems become more widely used. If Chinese models become the default choice for developers in emerging markets because they are cheaper, customizable and easier to deploy, Beijing could accumulate a different kind of influence through widespread technological adoption.

Much will depend on how both sides balance openness with security.

The United States faces a choice between restricting competing technology and preserving an open ecosystem that American companies themselves benefit from. China faces a similar tension between exporting accessible AI and restricting increasingly powerful capabilities for national-security reasons.

For the rest of the world, the outcome could create more options rather than fewer. Countries may use American frontier models for applications where maximum performance matters, Chinese open models where affordability and customization matter, and domestic systems where technological sovereignty is the priority.

The US-China AI competition is therefore unlikely to be decided simply by one dramatic breakthrough. It will be shaped by cost, capability, access, infrastructure, regulation, trust and the willingness of businesses and governments to adopt one ecosystem over another.

For now, Washington is asking the world to recognize AI as a strategic contest and align accordingly. Beijing is offering a different proposition: an AI ecosystem that emphasizes access, affordability and openness, while retaining the ability to impose controls when national interests demand them.

Which approach gains the widest international adoption may ultimately depend less on political declarations than on a practical question for users everywhere: which technology delivers enough capability, at a price and level of access they can actually use.