Artificial intelligence becomes a foreign-policy system
The rivalry between the United States and China is increasingly dividing artificial-intelligence development into competing political and technological blocs. Washington’s Pax Silica initiative emphasizes advanced chips, trusted supply chains and partnerships among aligned countries. Beijing’s World AI Cooperation Organization promotes more affordable systems, open models and participation by developing economies.
For many governments, the choice is not ideological. They want access to American computing hardware and investment while also using Chinese models that may be cheaper and easier to adapt. As security restrictions expand, however, participation in both systems may become harder.
Kazakhstan illustrates the balancing act
Kazakhstan joined the U.S.-led supply-chain group and then became a founding member of China’s organization. The Central Asian country has critical minerals, relatively inexpensive electricity and ambitions to build data centers. Its government signed a potential investment agreement worth up to $10 billion involving Nvidia, while Chinese direct investment reached a reported $2.8 billion in 2025.
The strategy gives Kazakhstan access to capital and a voice in multiple forums, but it also creates vulnerability. U.S. officials have signaled that trusted partners cannot meaningfully support a competing vision for advanced AI. China, meanwhile, presents its organization as inclusive and criticizes American restrictions on high-end chips and chipmaking equipment.
Other countries are making similar calculations
Indonesia has joined China’s organization while seeking technology and investment from both powers. Singapore is considering participation in the Chinese group even as American company Anthropic establishes an Asian office there. Saudi Arabia has paired a platform built on a Chinese open model with plans to expand computing capacity using Nvidia chips.
Japan remains closely aligned with the United States on strategic technology and semiconductor security, but researchers say it may still use lower-cost Chinese models for applications considered less sensitive. France is trying to develop more independent capacity through companies such as Mistral, though building frontier models requires enormous computing resources, specialized chips, energy and capital.
Standards and infrastructure carry influence
AI alignment is about more than software. It includes data centers, cloud services, cybersecurity rules, technical standards and access to semiconductor supply chains. Once a country builds critical systems around one technology stack, switching can become expensive and politically difficult.
China has financed AI-related projects through grants and loans in at least 77 low- and middle-income countries, according to research cited by the Associated Press. The United States relies on its lead in advanced chips, frontier models and a large network of allies. Each approach gives partner countries benefits while creating different dependencies.
Sovereignty may require practical safeguards
Most countries cannot build the largest models themselves, but they can still protect room to make decisions. Diversified suppliers, interoperable standards, local technical training and clear rules for government data can reduce dependence on any single provider. Regional cooperation can also give smaller markets more negotiating power.
The emerging blocs do not yet amount to a complete technological separation. Governments continue to hedge, and companies operate across borders where rules permit. The direction, however, is clear: access to AI is becoming tied to strategic trust. Countries that hope to work with both Washington and Beijing may find that the space between them is narrowing.
Procurement decisions made now may last for years. Governments therefore need to examine where data is stored, who can inspect models, how systems are updated and whether critical services can continue if political rules change. Price and performance alone are no longer sufficient criteria.
Source: Associated Press report. Reporting reviewed October 25, 2026.