The Battle for the AI Age: The US, Europe, and China Reshape the World Economy

The geopolitical contest over artificial intelligence has, by 2026, definitively moved out of technology departments and into foreign, economic, and security policy. Washington, Brussels, and Beijing are pursuing three distinct strategies. The United States is betting above all on technological leadership, computing capacity, and control of strategic semiconductors. China is combining a state-directed expansion of its domestic AI industry with open models and an international governance offensive. The European Union is trying to reconcile economic competitiveness, technological sovereignty, and binding regulation.

This is no longer simply about which company builds the most capable language model. What matters now is who controls the infrastructure of the AI age: semiconductors, data centers, power supply, cloud platforms, data, models, standards, and ultimately the rules under which these systems are deployed worldwide. The resulting order resembles less a classic technology race than a struggle for industrial power. AI is simultaneously a productivity technology, a military resource, an export good, and strategic infrastructure.

Washington sees AI as an instrument of power

The American strategy is comparatively clear-cut. The US wants not merely to preserve its technological lead but to translate it into economic and geopolitical power. The “America’s AI Action Plan,” unveiled in 2025, focuses on three areas: accelerating innovation, expanding American AI infrastructure, and securing US international leadership. Control over high-performance chips and the technologies needed to manufacture them is especially central. The plan explicitly designates advanced AI computing power as a strategic resource and calls for stricter enforcement of export controls. At the same time, Washington is demanding that allies not undermine American restrictions through their own supply chains.

The scale of the American lead explains this policy. The White House’s 2026 economic report cites estimates that, as of May 2025, roughly 74 percent of the world’s recorded AI computing capacity was located in the United States. The underlying data only covers part of global GPU capacity, however, so the figure should be read as an indicator of concentration rather than a complete inventory of global infrastructure.

Washington increasingly treats this infrastructure as a matter of national security. President Donald Trump signed a memorandum in June 2026 on the use of AI within the national security apparatus. The administration wants to make advanced commercial and open-source systems available more quickly for defense and intelligence purposes, while simultaneously building especially secure computing infrastructure. According to the White House, deployed systems are meant to remain “robust, steerable, controllable” and to operate within clear lines of accountability.

This is also changing the role of America’s technology giants. The major AI companies are no longer purely private providers of digital products. Their models, data centers, and research programs are becoming components of a national strategic infrastructure. The consequence for Europe is significant. Whoever uses American models and cloud infrastructure gains access to a technologically leading ecosystem — but also grows more dependent on decisions made in Washington and in American corporate headquarters.

China answers the American strategy with openness — on its own terms

Beijing is pursuing a different approach.

China cannot fully neutralize the American lead in the most advanced chips in the short term. Instead, the country is trying to expand its position through scale, low-cost models, open source, and international distribution. This strategy could be especially effective in emerging and developing countries, where the price of an AI application often matters more than access to the single most capable model available. Beijing further expanded this international dimension in 2026. At the World Artificial Intelligence Conference in Shanghai, China published an action plan for international AI cooperation in July. The document covers data, computing power, industrial applications, talent, standards, governance, and ethics. Among other things, Beijing calls for broader access to computing capacity and wider dissemination of open AI ecosystems.

In parallel, China published an action plan for international AI ethics. It calls for risk-based governance, international coordination, and rules covering the entire life cycle of AI systems. At the same time, the Chinese government emphasizes accounting for national circumstances and deepening cooperation with developing countries. This language carries geopolitical weight. China increasingly presents AI as an international development project, offering a counter-model to a technological order shaped by American companies and export controls.

At the Shanghai conference, the founding of the World Artificial Intelligence Cooperation Organization was also agreed. According to China, the organization will be headquartered in Shanghai and will help, in particular, countries of the Global South build their own AI capacities. Beijing explicitly links this to calls for a stronger role for the United Nations and for international cooperation on AI supply chains. The political core of this strategy is: China does not merely want to develop AI. It wants a say in the rules by which AI is distributed and regulated internationally.

Europe is caught between regulation and dependency

The situation is more complicated for Europe.

The European Union has a large single market, substantial industrial data resources, top-tier research, and key companies within the semiconductor supply chain. But when it comes to the major cloud platforms, leading AI models, and especially available high-performance computing capacity, Europe still lags behind the United States. This is the source of Europe’s central dilemma: it wants to achieve digital sovereignty, but at least in the short term needs technology and infrastructure from the United States to do so. While Washington acts primarily through innovation and security policy, Brussels has so far built a regulatory foundation.

Since August 2, 2026, key parts of the European AI Act have applied, and the responsible authorities now hold corresponding enforcement powers. These include rules for general-purpose AI models and transparency requirements. Rules for certain high-risk applications will follow later under the revised timetable: for the areas covered by Annex III in December 2027, and for AI components of certain regulated products in August 2028. The European AI Office can demand technical information from general-purpose models, have models evaluated, and order corrective measures. For certain violations, the legal framework provides for substantial fines. For prohibited AI practices, these can reach up to €35 million or 7 percent of global annual turnover, according to the European Commission, whichever is higher.

Europe has thereby created something neither the US nor China possesses in the same form: a horizontal, legally binding regulatory framework for AI across a single market of roughly 450 million people. But regulation alone does not build data centers. This is precisely Europe’s strategic problem. A recent Bruegel analysis describes Europe’s gap in AI computing infrastructure as urgent. The challenge, it argues, cannot be solved through capital alone; permitting procedures, grid access, and demand for European computing capacity all play a central role as well. Europe’s competitiveness in AI will therefore be decided less by how many rules Brussels passes than by whether Europe can simultaneously build a competitive technological base.

Chips are the new strategic currency

At the center of this contest sits an unassuming piece of hardware: the high-performance processor. Without modern accelerator chips, the largest AI systems can neither be trained economically nor operated at scale. At the same time, the production chain is extremely concentrated. Chip design, manufacturing, lithography, memory technology, packaging, and data centers are distributed across a small number of companies and countries. This creates an unusual form of mutual dependency.

The US has leading chip designers and cloud companies. Taiwan is central to state-of-the-art manufacturing. Europe holds indispensable parts of the production chain through its semiconductor equipment industry. China, in turn, controls important industrial capacities and raw-material supply chains, and is investing heavily in building an independent semiconductor industry. AI turns this complex division of labor into a geopolitical security problem.

Washington is therefore trying to limit China’s access to especially powerful processors and manufacturing technologies. The American AI Action Plan explicitly calls for closing loopholes in existing semiconductor controls and binding international partners more tightly to American restrictions.

This puts Europe in a difficult position. A transatlantic analysis by the Atlantic Council describes the likely outcome as “managed interdependence”: Europe and the United States remain closely linked, but hold different levers within that relationship. American export controls against China can therefore directly affect European economic interests. This may be one of the decisive conflicts of the coming years: how closely will Europe align its China policy in the technology sector with Washington’s?

The transatlantic conflict does not run between friends and rivals

Europe and the United States remain closely bound together technologically, economically, and in security terms. Yet they do not automatically share the same interests on AI. Washington wants to combine speed, capital, infrastructure, and technological leadership. Brussels wants to enable innovation while also safeguarding fundamental rights, competition, data protection, and institutional oversight. These differences are not merely philosophical. They carry economic consequences.

When European rules create additional compliance costs for international AI providers, Washington may see this as an obstacle for American companies. When the US restricts access to models, chips, or cloud infrastructure on national-security grounds, Europe may see this as a sovereignty risk. At the same time, neither side can view China in technological isolation from the other. Many of the most effective export controls only work if multiple states cooperate along the semiconductor supply chain. The real transatlantic AI question, then, is not whether Europe prefers American or Chinese technology. It is how much independent capacity to act Europe can build within an increasingly American-Chinese-dominated system.

The competition is shifting from the model to the infrastructure

Public debate often focuses on new generations of models. Strategically, however, the underlying infrastructure may matter more. An AI system needs chips. Chips need data centers. Data centers need power, grid connections, cooling systems, and capital. Applications, in turn, need data, cloud infrastructure, and skilled labor. Technology policy is thus merging with energy, industrial, and trade policy.

Whoever has enough energy and capital can build large data centers. Whoever controls advanced chips can influence where the most capable systems are developed. Whoever operates dominant cloud platforms controls a key gateway to the AI economy. And whoever sets technical standards can spread rules internationally without formally exercising political power.

The geopolitical significance of AI, then, does not arise from the intelligence of the models alone. It arises from the combination of software and physical infrastructure. That is what distinguishes the current AI revolution from many earlier phases of digitization.

At the same time, the safety question is growing

The race for technological leadership coincides with a second development: the discussion of the risks posed by powerful AI systems is becoming more concrete.

According to the latest reporting by the Associated Press, the International AI Safety Report 2026 reaches a cautious assessment. Current systems, it finds, show early capabilities that could be relevant to loss of control, but these are not yet at a level that would enable such a loss of control. The likelihood, nature, and timeline of such risks remain unusually uncertain. That very uncertainty makes international cooperation difficult.

The US wants to avoid safety rules that would narrow its technological lead over China. China, in turn, accuses Washington of partly using safety arguments as a means of technological containment against Chinese companies. Even so, a limited space for dialogue is emerging. US Treasury Secretary Scott Bessent said this week that Washington is willing to talk with China about the shared risks of AI. The issue is also likely to feature in upcoming US-China talks. This produces a remarkable paradox: the two states are competing for control of the defining technology of their era while simultaneously needing a minimum of cooperation to limit that technology’s risks.

Three models for the same technological revolution

The global AI order is thus developing along three distinct models of power. The American model rests chiefly on capital, private technology companies, computing capacity, cutting-edge chips, and the international reach of American platforms. The state increasingly supplements this industrial strength with export controls and national security policy.

The Chinese model combines state industrial policy with a huge domestic market, rapid implementation, open models, and a growing international governance strategy. Especially toward countries of the Global South, Beijing presents AI as development infrastructure meant to be as widely available as possible.

The European model, by contrast, rests on market size, the rule of law, standards, and regulation. Europe’s challenge is to complement this regulatory influence with its own industrial and technological capacity.

None of these strategies exists in isolation. American companies need international markets. China, despite major progress, still needs access to parts of global technology and supply chains. Europe needs American computing power and technology, while American companies need access to the European single market. The result is not full technological decoupling, but a system of managed dependencies.

The decisive question for Europe

For Europe, AI could therefore become an economic-policy fork in the road. The continent can try to be primarily a particularly safe and regulated market for technologies developed elsewhere. Or it can pair regulation with a far more aggressive infrastructure, energy, capital-markets, and industrial policy. This does not necessarily mean fully replacing American or Chinese systems. Full technological autarky would hardly be realistic given global supply chains.

What matters is whether Europe has enough of its own capacity to make political decisions without its access to core technology depending on a single external actor. That includes European data centers as much as semiconductor expertise, cloud infrastructure, models, energy supply, and capital for fast-growing technology companies. The AI Act answers the question of the conditions under which AI may be deployed in Europe. It does not answer the second, equally important geopolitical question: who will own and operate this AI?

The new world order is being built in data centers

The competition among the US, China, and Europe therefore reflects a larger shift in international politics. In the industrial age, steel, oil, machinery, and production capacity were the central foundations of state power. In the digital age, semiconductors, telecommunications networks, and software platforms were added. In the AI age, computing power, energy, data, and models are becoming strategic resources.

The United States currently holds the strongest combination of capital, leading AI companies, and computing infrastructure. China has industrial scale, a vast domestic market, and the ability to combine technological ecosystems with state strategy. Europe has a large market, significant research, specialized industrial expertise, and regulatory influence — but markedly less control over key parts of AI infrastructure.

At the same time, the world outside these three power centers is becoming more important. Countries across Asia, Latin America, Africa, and the Middle East must decide which infrastructure to adopt, which standards to follow, and which technology providers to depend on. Washington and Beijing have long since begun competing for these international technology partnerships. Europe must therefore decide not only how to regulate AI. It must decide what role it wants to play in the emerging AI economy.

Because the decisive geopolitical question of the coming years may not be which country owns the smartest model. It may be who controls the infrastructure on which the world’s smartest models run.