Leaders in Beijing and Washington believe that whichever country achieves dominance in AI will enjoy a decisive economic, strategic, and military advantage. If there is one idea that leaders in both countries share, it is that they cannot afford to lose the “AI race” to the other, even though there is very little clarity on where the finish line is or what it would represent. Given that both China and the US are courting Southeast Asia, countries in the region will need to navigate deftly to avoid being caught exclusively in the technology systems of either one of them. In short, Southeast Asia’s oft-repeated adage — that it does not want to “choose” between Beijing and Washington — has now moved into the AI domain.

America’s strategy is built on a bet that “whoever has the largest AI ecosystem will set global AI standards and reap broad economic and military benefits.” Thus, American policymakers are bent on avoiding and removing regulatory obstacles. This will enable the rapid buildout of computing power and prevent China from stealing the march from America’s leading technology companies in innovation and standard-setting. Vice President J.D. Vance has declared that “the AI future is not going to be won by hand-wringing about safety. It will be won by building.”

In that spirit, America’s AI Action Plan directs the US to “export its full AI technology stack — hardware, models, software, applications, and standards — to all countries willing to join America’s AI alliance.” What is left unsaid is that many parts of the “US AI technology stack” are sourced from Taiwan, South Korea, Japan, and China. Through the transfer of its full stack, so the thinking goes, the US will lock other countries into its technology and standards, thereby securing its dominant position.

China’s strategy is, in many ways, a study in contrast with the Trump administration’s approach. At the World Artificial Intelligence Conference in July, Premier Li Qiang emphasised that China aspires not only to lead on AI safety issues, but also to create a World AI Cooperation Organisation based in Shanghai. Whereas the Trump administration instinctively opposes the United Nations taking on a larger role in AI governance, China — at least rhetorically — is supportive. While America’s leading technology companies have poured billions of dollars into building leading-edge, proprietary, closed AI systems (Meta’s Llama and OpenAI’s gpt-oss being exceptions), Chinese AI firms are building open models to appeal to global users, particularly in the developing world.

Given the existential stakes Washington has placed upon “winning the AI race”, the US will likely grow more aggressive in seeking to slow China’s progress as Chinese firms continue to make technological strides and diffuse Chinese open-source models around the world. This will have spillover effects in Southeast Asia, a region that both the US and China are seeking to court.

The US has a conflicted approach to Southeast Asia on AI. On one hand, Washington will likely ratchet up pressure on Southeast Asian countries to adhere to American export controls and regulatory restrictions on technology transfer to China. There is a widely held view in Washington that China’s recent AI advances have been aided by smuggling and illicit transshipments of restricted semiconductor chips from Southeast Asia to China. Washington will likely step up law enforcement actions, sanctions, and technology transfer restrictions to cut off China’s access to leading-edge chips and related technologies.

To maximise benefits while minimising risks, leaders in the region will need to be clear, consistent, and firm with Washington and Beijing about their goals and concerns. They will need to disavow values or ideology as a basis for technology adoption.

On the other hand, US officials want Southeast Asian countries to adopt America’s AI tech stack. The challenge for the US is that if it applies too heavy a hand in pushing to limit the adoption of Chinese technology or pressuring countries to limit technology cooperation with Chinese firms, then it could well achieve the opposite of its intended result. An overbearing American approach could push countries in the region to turn away from American AI in favour of China’s offerings.

In this sense, technology diffusion will likely be an epicentre of US-China strategic competition in Southeast Asia. Pressure will grow on countries and companies in the region to choose sides. Southeast Asian countries naturally seek to benefit from both powers without being forced to adopt one AI tech stack or the other monogamously.

Amid these challenges, the US-China AI competition will also create opportunities for the region. For example, major multinational technology firms will likely continue to grow their investments in local talent and innovation ecosystems in Southeast Asia. Global producers will see benefits in diversifying value chains and gaining insulation from geopolitical uncertainty by producing in Southeast Asia. This can already be seen in the buildout of AI data centres in Malaysia, Thailand and Singapore, where the US and China have leading companies with significant presence. These investments will generate jobs and develop infrastructure and energy capacity.

To maximise benefits while minimising risks, leaders in the region will need to be clear, consistent, and firm with Washington and Beijing about their goals and concerns. They will need to disavow values or ideology as a basis for technology adoption. They also need to warn that onerous regulatory burdens or capricious policy decisions will disincentivise long-term investments in either side’s technology stacks.

Striking such a balance between the US and China will not be easy, but it will be necessary for the region to avoid a complete bifurcation between US and Chinese technology ecosystems. As uncomfortable as it will be for countries and companies to be caught in the middle, it will still serve as a better option for the region than becoming viewed as being aligned exclusively with one power in opposition to the other. That outcome would not only result in less optionality; it would also be dangerous.

2025/280