The contest to lead the AI revolution is the defining technology competition of our era — with consequences that reach into every corner of the global economy

Introduction: The Technology That Changes Everything

In the history of transformative technologies, a small number stand apart not merely because they improved specific processes or industries but because they changed the fundamental conditions of economic production and human capability — gunpowder and navigation in the age of exploration, steam power and mechanisation in the industrial revolution, electricity and the combustion engine in the twentieth century. Artificial intelligence is increasingly described, and increasingly behaves, like a technology of this fundamental character — one that does not merely automate existing tasks but potentially transforms the relationship between human cognition and economic production across every domain simultaneously.

The competition to lead this transformation — to develop the most capable AI systems, to build the infrastructure on which AI runs, to capture the economic value it creates, and to shape the standards and governance frameworks that will determine who benefits and on what terms — has become the central axis of the global technology competition between the United States and China, and the defining strategic priority of every major economy and technology company that understands its implications.

This is a race with no clear finish line. Artificial intelligence is not a single technology to be developed and deployed but a rapidly evolving field in which capabilities are advancing continuously, in which the frontier shifts with each major model release, and in which the applications being enabled are expanding far faster than the governance frameworks designed to manage them. Understanding the state of the race — who is leading, in what dimensions, why it matters, and where it is heading — requires looking simultaneously at the technology, the economics, the geopolitics, and the governance challenge that AI presents.

Section I: The State of the Technology

Large Language Models and the Frontier

The most visible recent frontier of artificial intelligence development has been large language models — AI systems trained on vast quantities of text to generate human-quality language, answer questions, write code, summarise documents, and engage in sophisticated reasoning across virtually any domain. The release of ChatGPT by OpenAI in November 2022 brought this capability to mainstream public attention in a way that previous AI advances had not, demonstrating that AI systems had crossed a threshold of general-purpose utility that made them immediately applicable to an enormous range of tasks that previously required human cognitive effort.

The subsequent pace of development has been extraordinary. Within roughly three years, AI capabilities have advanced to the point where frontier models can pass professional licensing examinations in medicine, law, and accountancy; write functional code in most major programming languages; conduct multi-step reasoning across complex domains; and process not just text but images, audio, and video in integrated multimodal systems. The competitive dynamics driving this progression — with multiple well-funded organisations releasing successive generations of more capable systems at intervals of months rather than years — show no signs of slowing.

Section II: The US-China AI Competition

American Strengths

The United States maintains significant structural advantages in the global AI competition. American technology companies — OpenAI, Google DeepMind, Anthropic, Meta AI, Microsoft — have developed the world’s most capable frontier AI models and continue to define the leading edge of AI capability. American universities and research institutions remain the world’s primary training ground for AI researchers, with a brain drain dynamic that historically attracted significant global talent to American AI programmes. The American venture capital and technology investment ecosystem provides funding for AI development at a scale and speed that no other country’s innovation finance system currently matches.

The semiconductor advantage is critical: the most advanced AI chips — Nvidia’s H100 and successor architectures — are designed by American companies and manufactured primarily in Taiwan. American export controls have sought to limit China’s access to the most advanced AI chips, with the objective of maintaining a computational advantage at the training frontier. Whether those controls can be sustained against determined Chinese efforts to close the gap through domestic semiconductor development is a central question for the medium-term trajectory of the competition.

China’s AI Development

China has made AI a national strategic priority in ways that mobilise state resources, incentives, and direction at a scale that no market-driven system can replicate. Chinese AI companies — Baidu, Alibaba, Tencent, Huawei, and a large number of smaller specialists — have developed capable AI systems across multiple domains. China’s AI research output by volume — measured in publications — rivals or exceeds American output, though quality assessments are more contested. And China’s access to domestic data at scale — from its large population, its digital economy, and the data collection enabled by its technology ecosystem — provides training resources of significant value for AI development in Chinese language and cultural contexts.

The chip export controls have created a genuine constraint on China’s ability to train frontier models at the computational scale that American labs employ. Chinese AI development has responded partly by improving algorithmic efficiency — achieving competitive performance with less computation — and partly by accelerating domestic semiconductor development. The January 2025 release of DeepSeek’s R1 model, which demonstrated reasoning capabilities competitive with frontier American models at a fraction of their reported training cost, attracted global attention as evidence that the computational advantage conferred by export controls may be more limited than American policymakers had assumed.

Section III: The Economic Stakes

The economic stakes of AI leadership are extraordinary. McKinsey Global Institute estimates that AI could add approximately $13 trillion to global economic output by 2030, with the distribution of that value heavily skewed toward early leaders in AI development and adoption. Goldman Sachs has estimated that AI could automate approximately 25 to 50 percent of current work tasks across the US economy — a potential productivity transformation of historical significance, though one that also raises profound questions about labour market disruption and the distribution of AI’s economic benefits.

At the national level, the productivity gains from AI adoption compound over time. A country whose businesses, healthcare system, education sector, and government services are more effectively AI-enhanced than those of competitors will experience real productivity advantages that translate into economic growth and, over decades, into significant differences in living standards and national economic weight. This is why AI leadership is treated as a strategic national priority by governments across the world — not merely because of its military applications but because of its fundamental implications for long-term economic competitiveness.

Section IV: Governance and the Alignment Challenge

Alongside the competitive race for AI capability runs a parallel and critically important challenge: ensuring that AI systems are safe, reliable, and aligned with human values as they become more capable and more broadly deployed. The concern — taken seriously by many of the researchers building the most advanced AI systems — is that AI systems optimised to achieve goals may develop approaches to achieving those goals that are unsafe or that conflict with human interests in ways that are difficult to predict or control. Managing this risk while continuing to develop progressively more capable systems is the ‘alignment problem’ that has moved from the theoretical concerns of AI safety researchers into the strategic priorities of major AI laboratories.

International governance of AI development faces the fundamental challenge of coordinating across states with different regulatory philosophies, different risk tolerances, and in many cases competing strategic interests in AI leadership. The European Union has enacted the AI Act — the world’s first comprehensive AI regulation — establishing risk-based requirements for AI systems across different application categories. The United States has relied primarily on voluntary commitments and executive guidance rather than binding legislation. China has implemented regulations on specific AI applications, particularly generative AI and recommendation algorithms. No binding multilateral governance framework for AI development yet exists.

Conclusion: The Race Without a Finish Line

The race for AI supremacy is unique among the great technology competitions in human history in that it has no clearly defined endpoint. There is no summit to be reached, no capability threshold at which one competitor can declare victory and others acknowledge defeat. The frontier is continuously advancing, the applications being enabled are continuously expanding, and the governance frameworks are continuously lagging behind the technology they attempt to manage.

What is clear is that the countries, companies, and institutions that most effectively develop, deploy, and govern AI will have significant advantages — in economic productivity, in military capability, in scientific research, in public service delivery, and in the soft power of technological standard-setting — over those that do not. The race is not a zero-sum competition in which one country’s gain is necessarily another’s loss. Much of AI’s potential value will be realised through diffusion and adoption across the global economy. But the distribution of that value — who captures the largest share of AI’s economic benefits, who shapes its governance frameworks, and who builds the relationships of technological dependency that AI infrastructure creates — will be shaped substantially by the outcome of the competition currently underway.

Climate Risk, Early Warning Systems and Global Water Security: Michel Jarraud on theFuture of Meteorology and Multilateral Cooperation