The United States Wins the Artificial Intelligence Race — But for How Long?
Silviane Duranton: Global Leader at BCG X
Nicolas Lang: Global Leader at BCG Henderson
It is tempting to view artificial intelligence as merely another technological advancement, but that is akin to saying the steam engine was just a faster way to pull a cart. The reality is that AI is rapidly transforming the geopolitical landscape. While global power used to rely on armies, oil, and control over land and air, it now also hinges on data, talent, computational infrastructure, and the legal frameworks required to develop and deploy advanced models.
Currently, the United States is clearly in the lead. Since 2019, American tech companies have produced more than half of the leading AI models. Between 2023 and 2024, six major companies invested $212 billion in research and development, while generative AI startups raised nearly $90 billion in 2024. The U.S. boasts the largest pool of AI specialists — about 500,000 — and the highest data center capacity in the world, at 45 gigawatts (GW).
However, as other nations move quickly to close the gap, America’s position is no longer guaranteed. At first glance, the race might resemble a Cold War, with two great powers — the U.S. and China — vying for technological dominance while the rest of the world watches. In reality, countries around the globe are heavily invested in the game. In a contest without fixed rules or trustworthy referees, that creates complications.
The United States and China adopt drastically different approaches to AI. The U.S. relies on its dynamic private sector, bolstered by a system of vibrant startups and tech giants like Google, Microsoft, and OpenAI. Conversely, China leans on central planning — massive government funding, military-industrial integration, and global trade initiatives — to enhance its thriving landscape of AI startups.
China’s strategy is already bearing fruit. The government has injected around $132 billion into the AI sector since 2019 and invests $60 billion annually in R&D for companies. As a result, China possesses 15 of the 40 leading AI models and employs 18% of the top researchers in the field, generating more STEM graduates than any other country. China’s data center capacity stands at 20 gigawatts, the second-largest in the world.
Despite U.S. export controls limiting China’s access to advanced semiconductors, Chinese engineers have managed to extract increasingly capable devices through efficiency breakthroughs. This has allowed the relatively unknown startup DeepSeek to make waves by delivering performance on par with the best U.S. models while using a fraction of the computing power.
The European Union finds itself in a tight spot. Lacking the computational power of the United States and the vast data resources of China, it ranks third in developing AI with around 275,000 specialists and an annual R&D expenditure of €381.4 billion ($449 billion), along with about eight gigawatts of data center capacity.
That doesn’t mean Europe is short on strategic advantages. While only three of the 40 leading AI models originated on the continent, the EU accounts for 19% of the most-cited academic research in the field. It is also home to ASML, the Dutch company that provides most of the advanced lithography machines used in semiconductor manufacturing.
Recognizing its inability to compete with the U.S. and China in creativity or scale, the EU has focused on setting the rules of the game. The Artificial Intelligence Act — the most comprehensive cross-border regulatory framework for AI to date — is more than just a set of rules; it is a step toward soft power under the guise of regulation.
The EU has also earmarked over €200 billion for achieving the Digital Decade goals, investing in infrastructure, training, and digital transformation to carve out a strategic space for the AI industry. With sufficient resources, momentum, and political will, Europe can certainly become a global AI power in its own right.
Meanwhile, the U.S.-China competition intensifies. On January 20, DeepSeek unveiled its ultra-efficient model, R1, demonstrating that U.S. sanctions would not stifle China’s ambitions in AI. The very next day, U.S. President Donald Trump launched the Stargate Initiative — a $500 billion plan to establish next-generation data centers and secure the supply of chips for America. The timing of these announcements was not coincidental; it marked the starting gun in a war for technological self-sufficiency.
This race is not merely about hardware and software; it is also about values. Who decides what is fair when it comes to emerging technologies? Whose ethics will be embedded in the algorithms increasingly governing essential services like healthcare and education? Answers to these questions could shape how AI is used and which interests it serves, making the export of AI ethics as critical as the export of the models themselves.
Furthermore, the struggle for the future of AI is no longer limited to China, the United States, and Europe. Until now, the Global South has been a passive recipient of systems it did not design, trained on data it did not produce, governed by rules it did not help shape. But this situation is changing. Across Africa, Latin America, and Asia, national and regional initiatives are developing models tailored to meet local priorities and needs, thus enhancing linguistic and digital sovereignty.
The Middle East is a prime example. In its quest to carve out a place in the emerging technological landscape, Saudi Arabia, Qatar, and the United Arab Emirates are pouring their oil wealth into world-class data centers and homegrown AI models.
This underscores an urgent need for accountability mechanisms that match the scale and speed of technological innovation, supported by a shared global vision for AI governance. Although the Organisation for Economic Co-operation and Development and the G7 have made gestures toward regulatory coordination, implementation is nearly nonexistent. Similarly, the United Nations’ proposal to establish a global advisory body for AI has yet to translate into actions, despite being promising.
The risks are immense. Without coordinated international action, AI could concentrate control over knowledge, deepen digital divides, and reshape cultures and societies in ways that escape public scrutiny.
While this outcome may seem inevitable, it is not. The window of opportunity is narrow—and closing quickly—for choosing a different path. This choice will be the defining challenge of our era.
