U.S.-China AI cold war: Where is the world heading?

The artificial intelligence (AI) has become much more than a new technological industry. From Silicon Valley to Beijing’s Zhongguancun, the struggle over AI is becoming a struggle over power, security, knowledge, and the future of the world.

Washington and Beijing are intensifying technological competition while also recognising that neither country can manage all the risks advanced AI creates alone. The present confrontation is therefore better understood as an emerging form of AI geopolitical competition.

Developing countries should not approach AI geopolitics merely as a question of choosing between Washington and Beijing. The more fundamental issue is whether countries can acquire technological capabilities while retaining sufficient autonomy over their data, institutions, knowledge and cultural priorities.

The future global AI order will also be shaped by countries that are neither Silicon Valley nor Beijing. India, Southeast Asia, the Gulf states, Africa, Latin America and smaller countries such as Sri Lanka will become increasingly important. The next stage of AI geopolitics may move toward a much broader international negotiation over who gets to shape the technological future.

Photo/Baidu AI

How Did the US-China AI Confrontation Get Here?

The roots of the current confrontation extend well beyond ChatGPT, DeepSeek and today's generative-AI revolution. For decades, the United States developed substantial advantages in advanced semiconductor design, computing infrastructure, research universities, venture capital, and globally influential technology companies. China, meanwhile, developed an enormous manufacturing ecosystem, a huge domestic digital market, extensive infrastructure and strong state-supported investment in strategic technologies. The relationship changed significantly when Washington increasingly treated advanced semiconductors and AI-related computing capabilities as matters of national security.

Restrictions on the export of advanced chips and semiconductor-manufacturing technologies to China were designed partly to constrain China's access to the computing power required for frontier artificial intelligence. Beijing responded by accelerating technological indigenisation, investing in domestic semiconductor capabilities, developing more efficient algorithms and encouraging Chinese companies to reduce dependence on foreign technology. The consequence has been a technological feedback loop: American restrictions have encouraged greater Chinese efforts toward self-reliance, while China's rapid technological progress has reinforced American concerns about strategic dependence and technological security.

By 2026, this process had produced a significant transformation in perceptions of the AI balance. Taylar Rajic and Lauryn Williams of the Center for Strategic and International Studies reported in September 2026 that some leading academics had effectively begun treating the previous US–China model-performance gap as closed in important areas. Their analysis identifies technological indigenisation and China's open-weight strategy as important components of this development, while also noting that the United States retains major advantages in frontier AI capabilities. The significance of this development is not that China has simply “defeated” American AI, nor that the United States has lost its technological leadership. Rather, the earlier assumption of a permanent and overwhelming technological gap has become much more difficult to sustain.

Artificial intelligence matters to Washington and Beijing because it increasingly affects almost every major dimension of national power. AI has applications in cybersecurity, intelligence, military planning, autonomous systems, scientific research, medicine, financial services, logistics, education, public administration and industrial production. Consequently, technological leadership in AI may create advantages across multiple sectors simultaneously. This explains why both governments increasingly regard AI as a strategic technology rather than simply another commercial innovation. Yet there is another dimension that deserves much greater attention from communication scholars: AI is also becoming a technology through which knowledge itself is produced, interpreted and communicated.

Estimates for 2026 indicate an extraordinary expansion of AI-related infrastructure investment. Chinese AI infrastructure investment has been estimated at approximately US$139 billion for the year, while US data-centre investment has been estimated at roughly US$800 billion. These figures do not represent identical categories of expenditure and therefore should not be interpreted as a simple one-to-one comparison. Nevertheless, they demonstrate two important realities simultaneously: the enormous scale of American capital and infrastructure investment and the equally significant mobilisation of Chinese technological resources.

Will the World Divide into Two AI Systems?

It is therefore too simplistic to describe the confrontation as merely Silicon Valley versus Beijing. The competition involves entire technological ecosystems that Silicon Valley and Beijing are building. The American ecosystem includes major AI companies, cloud providers, semiconductor firms, universities, venture-capital networks and research laboratories. China's ecosystem includes technology corporations, AI start-ups, semiconductor companies, universities, state-supported research institutions and an enormous domestic market. Both systems possess different institutional structures and development strategies, but neither operates completely independently from the global economy.

This is one reason the current confrontation differs from the twentieth-century Cold War. Washington and Beijing remain deeply connected through trade, manufacturing, scientific knowledge, international markets and global supply chains. Their competition is therefore occurring inside a system of continuing interdependence. The central challenge is not simply how to defeat the technological competitor, but how to compete without creating an uncontrolled process of technological fragmentation that could damage the global economy itself.

One of the most consequential questions is whether the world will eventually develop two substantially separate AI ecosystems. A complete technological separation would mean increasingly different semiconductor supply chains, cloud infrastructure, AI models, technical standards, governance systems, and digital platforms operating around American and Chinese spheres of influence. Such fragmentation would have consequences far beyond Washington and Beijing because countries in Asia, Africa, Latin America and the Middle East could find themselves under pressure to align with technological ecosystems.

Recent scholarship also suggests that the Global South may experience these relationships differently. An article in International Affairs in 2026 characterises the US approach as a “leader” model associated with systematic technology transfer but comparatively limited localisation, while describing China's approach as a “collaborator” model with greater emphasis on localisation but weaker regulatory integration. These differences suggest that developing countries should not approach AI geopolitics merely as a question of choosing between Washington and Beijing. The more fundamental issue is whether countries can acquire technological capabilities while retaining sufficient autonomy over their data, institutions, knowledge and cultural priorities.

What Does China's Approach Mean for Developing Countries?

This is particularly important for Sri Lanka. The danger for developing countries is that they become permanent consumers of AI technologies designed elsewhere. A 2025 Nature Machine Intelligence discussion of localising AI in the Global South highlighted the limitations of many large language models trained predominantly on English-language and Western cultural data. Such systems may perform less effectively when confronted with non-Western languages, cultural contexts and social realities. The issue is therefore not simply whether a country has access to artificial intelligence. It is whether artificial intelligence can meaningfully understand the people and societies it is designed to serve.

This distinction is crucial. If Sinhala and Tamil communities remain poorly represented in AI datasets, then technological access alone does not necessarily produce genuine inclusion. The question becomes one of knowledge sovereignty: who creates the datasets, who defines the categories, whose cultural experiences become visible and whose knowledge remains peripheral? Sri Lanka's emerging AI strategy, digital infrastructure initiatives and investments in local-language datasets therefore have significance beyond technology policy. They represent an opportunity to participate in shaping the knowledge architecture of AI rather than merely importing finished systems.

For Sri Lanka, the US–China AI confrontation should not be reduced to a binary choice between America and China. Sri Lanka can maintain relationships with American universities, technology companies and research networks while simultaneously engaging Chinese technological and infrastructure partnerships. The more strategic objective should be the development of digital and AI capabilities that allow Sri Lanka to cooperate with both while retaining sufficient autonomy over its data, knowledge, languages, institutions and public interests.

This requires investment not only in data centres and connectivity but also in universities, researchers, local-language datasets, AI literacy, ethical governance and international research collaboration. Sri Lanka's emerging AI policy environment provides an important foundation, but infrastructure alone cannot create technological sovereignty. A data centre is infrastructure. A model is technology. A dataset is knowledge. But the ability to decide what knowledge should be represented, how communities should be understood and whose interests AI should serve is ultimately a question of cultural and communicative sovereignty.

How Will the US-China AI Confrontation End?

The eventual outcome is unlikely to resemble a conventional war in which one side simply defeats the other. The technological, economic, and institutional interdependence between the United States and China is too extensive for such a straightforward conclusion. Several processes are more likely to develop simultaneously. Competition will probably continue in advanced chips, computing infrastructure, AI models, military applications, research, and international standards. Some critical technologies may become increasingly separated into American and Chinese ecosystems, creating partial technological bifurcation. At the same time, practical cooperation may develop around areas where uncontrolled AI risks threaten both countries.

The most significant recent indication of this possibility is the emergence of direct AI-safety communication between Washington and Beijing. Reuters reported on 21 September 2026 that US and Chinese officials had agreed to formalise an AI dialogue and discussed an “incident line” for serious AI-safety incidents, including concerns involving uncontrollable AI agents and cyber activity by non-state actors. Another meeting is expected to take place in Shenzhen in approximately two months. This does not indicate that the broader geopolitical rivalry is ending. Instead, it demonstrates an important principle: intense strategic competitors may still require communication mechanisms precisely because their technologies are becoming sufficiently powerful to create risks that cannot be managed unilaterally.

The most plausible long-term direction may therefore be neither complete confrontation nor complete cooperation, but managed competition. The United States and China may continue to compete aggressively over technological leadership while simultaneously developing mechanisms to prevent competition from producing catastrophic consequences. This would resemble a new form of technological coexistence in which strategic rivalry remains but certain boundaries are negotiated.

Such an arrangement would require more than diplomatic statements. It would require crisis communication channels, technical exchanges, transparency mechanisms, standards negotiations and agreed procedures for responding to dangerous AI incidents. Artificial intelligence introduces a new problem into international relations because an AI-generated cyberattack, autonomous system failure or dangerous biological application could create consequences that move faster than traditional diplomatic communication.

This is why international communication and public relations should no longer be considered peripheral to AI geopolitics. The future relationship between technological powers will depend partly on their ability to communicate during moments of uncertainty. The question will not always be whether Washington and Beijing agree. The more important question may be whether they can disagree without allowing technological competition to become uncontrolled escalation.

The Future May Not Belong to the Winner

The deepest lesson of the US–China AI confrontation may therefore be that the future cannot be understood simply through the question of who wins the AI race. The United States possesses extraordinary strengths in frontier AI, capital, computing infrastructure and global technology companies. China has demonstrated remarkable capacity for technological adaptation, domestic industrial mobilisation, open-weight development and large-scale deployment. Both countries are likely to continue competing intensely.

Yet the future global AI order will also be shaped by countries that are neither Silicon Valley nor Beijing. India, Southeast Asia, the Gulf states, Africa, Latin America and smaller countries such as Sri Lanka possess markets, languages, data, talent, cultural knowledge and political agency that will become increasingly important. The next stage of AI geopolitics may therefore move beyond a simple US–China race toward a much broader international negotiation over who gets to shape the technological future.

The most sustainable outcome does not necessarily require the disappearance of competition. Competition can stimulate innovation and investment. The greater challenge is to prevent technological rivalry from becoming permanent fragmentation, military escalation or a new form of digital dependency.

Artificial intelligence may eventually become the most powerful communication technology humanity has ever created. If that happens, its future cannot be determined only by engineers, investors and geopolitical strategists. It must also be shaped by communication scholars, ethicists, governments, citizens and societies asking a deceptively simple question: When machines increasingly mediate how humanity understands the world, who will teach the machines how to understand humanity?

Writing by Manoj Jinadasa (PhD in Digital Critical Media Studies, Newcastle, UK and Senior Lecturer and Head, Department of Mass Communication, University of Kelaniya, Sri Lanka); Editing by Wang Shixue