China's open-source AI models are rapidly gaining ground in developing nations, capturing over 54% of global model usage as of late September 2026, up from less than 15% just a year ago. This shift reflects a fundamental tension in global AI governance: while the United States maintains technological superiority through advanced chips and frontier models, China is winning hearts and market share by offering cheaper, more accessible alternatives that don't require expensive subscriptions or reliance on foreign cloud providers.

The contrast between the two superpowers' strategies became starkly apparent during the recent Trump-Xi summit in Washington. While US officials emphasized maintaining America's AI lead, Chinese President Xi Jinping called for "drawing on each other's strengths, not guarding against each other," a direct reference to Beijing's frustration with US technology export controls. Notably, chip export restrictions were not a major topic of discussion at the summit, and neither Beijing nor Washington issued public statements about lifting or tightening semiconductor controls.

What's Driving China's AI Accessibility Strategy?

China's embrace of open-source AI models wasn't initially a top-down government strategy. Instead, it emerged from necessity. US export controls on advanced semiconductors, coupled with China's smaller capital markets, pushed Chinese developers toward open models as a way to compete with leading US proprietary systems like those from OpenAI and Anthropic. Companies like DeepSeek and Moonshot capitalized on this opportunity, offering models that users can freely download, customize, and run without paying hefty subscription fees.

The appeal is straightforward for developers in the Global South. As Eric Olander, editor in chief of The China-Global South Project, explained, "For developers in the Global South, an inexpensive Chinese model from DeepSeek or Moonshot may be more useful than a slightly more capable system requiring an expensive subscription and access to a foreign cloud provider". Even major American companies, including Airbnb, DoorDash, and Shopify, have begun adopting Chinese models to take advantage of lower costs and greater flexibility for customization.

How Are Global Coalitions Reshaping AI Access?

The geopolitical competition over AI has crystallized into two competing frameworks. The US-led Pax Silica alliance, launched in December 2025, aims to secure critical supply chains free from Chinese influence, with over two dozen countries and the European Union signed on. In response, Chinese leader Xi Jinping launched the World Artificial Intelligence Cooperation Organization (WAICO) in July 2026, which has recruited 29 countries, including Russia, Indonesia, and Pakistan.

These competing coalitions represent fundamentally different visions for AI governance:

• Pax Silica Approach: Gates frontier AI behind trusted supply chains, emphasizing mutual prosperity and technological progress while reducing dependencies on Chinese technology and countering unfair market practices.
• WAICO Approach: Organizes access around open-weight models for the Global South, promoting open-source innovation and lower-cost alternatives that don't require expensive subscriptions or foreign cloud infrastructure.
• Strategic Asset Framing: Both frameworks now treat AI as a strategic asset subject to supply control rather than a commercial product, making access architecture a central concern for any nation dependent on imported AI technology.

The stakes of this competition became clear during a dramatic incident in mid-2026 involving Anthropic's Mythos model, a frontier AI system with advanced cybersecurity capabilities. In June 2026, the US Department of Commerce prohibited access to Mythos for all foreign nationals, marking the first time export controls were applied to a deployed AI model rather than just the underlying semiconductors. The model was disabled globally within weeks, alarming US allies who had assumed guaranteed access. French President Emmanuel Macron warned that US allies "will not buy any model made by [US AI] companies if from one day to the next you can just turn off the switch".

Emmanuel Macron

Will China Keep Its Models Open?

The critical question facing the global AI ecosystem is whether China will maintain open access to its increasingly capable models. As Chinese AI systems approach the performance levels of Anthropic and OpenAI's most advanced offerings, Beijing faces mounting pressure to restrict their deployment for national security reasons. Alex Colville, an analyst focusing on tech and security at the Australian Strategic Policy Institute, noted that "Beijing wants to export the country's technology to bolster its international influence, but national security always comes first".

"Beijing wants to export the country's technology to bolster its international influence, but national security always comes first," said Alex Colville, analyst at the Australian Strategic Policy Institute.

Alex Colville, Analyst at Australian Strategic Policy Institute

Rather than imposing a blanket closure, Beijing is likely to implement licensing rules that restrict where certain models can be used, Colville suggested. This mixed approach would allow China to maintain its appeal to developing nations while protecting sensitive capabilities from potential adversaries.

Reva Goujon, a director focusing on China at Rhodium Group, emphasized the fragility of China's current advantage: "China's strategy is only more compelling to the world so long as the models remain available and open so that they can be adopted and fine-tuned for domestic and enterprise needs". Once restrictions are imposed, the appeal of Chinese models may diminish significantly, potentially driving countries back toward US alternatives or spurring efforts to build sovereign AI capabilities.

"China's strategy is only more compelling to the world so long as the models remain available and open so that they can be adopted and fine-tuned for domestic and enterprise needs," explained Reva Goujon, director at Rhodium Group.

Reva Goujon, Director at Rhodium Group

What Does Winning the AI Race Actually Mean?

The definition of victory in the AI competition remains contested. Different metrics, from profitability to market share, could yield entirely different answers about who is actually winning. Chucheng Feng, founding partner of advisory firm Hutong Research, framed the choice facing companies and nations: "Do I want a model that costs 10 times as much but is six months ahead, or one that costs one-tenth as much but is only six months behind?".

As AI capabilities continue to advance, the performance gap between frontier US models and slightly less capable Chinese systems may become less noticeable to most people and businesses. Some companies may require cutting-edge US models for specialized tasks, while others may find Chinese models efficient enough for their needs. The likely outcome, according to experts, may mirror the mobile phone market, where Apple captures 90% of profits despite having a smaller user base than competitors.

Meanwhile, China's semiconductor industry is making progress despite US export controls. According to forecasts from a Taipei-based research company, the share of Nvidia, AMD, and other foreign suppliers is expected to decline to 21% in 2026 from 34% the previous year. Paul Triolo, a partner at DGA-Albright Stonebridge Group, believes Beijing is satisfied with the progress of its semiconductor industry despite US restrictions.

However, Triolo warned that the current pause in escalating chip export controls may be temporary. Legislative initiatives in the US Congress, including the Match Act, are intended to encourage Washington's allies, including Japan and the Netherlands, to coordinate restrictions on sales of advanced chipmaking equipment to China. If new restrictions are imposed, Chinese officials may view them as violations of agreements reached in Busan in October 2025 and could respond by expanding controls over rare earth metals and equipment for their production.

The bifurcation of global AI access infrastructure reflects a deeper reality: frontier AI is no longer treated as a commercial product but as a strategic asset subject to geopolitical competition. For nations caught between the two superpowers, the challenge is navigating a landscape where access is conditional, revocable, and increasingly tied to political alignment rather than technical merit alone.