The debate on artificial intelligence continues to be framed around the wrong question: who has the best model? It is a useful question for the market, for investors and for corporate communications, but it is not enough to understand the geopolitical dimension of the competition between the US and China by Andrea Monti – Initially published in Italian by Formiche.net
The news of the order by which the US Department of Commerce has banned the provision of Anthropic’s two latest AI models to non-US entities provides an opportunity to analyse the US strategy as a whole.
This appears to be based on a fundamental premise: control of the supply chain—from technology to the market—that underpins AI systems. Consequently, the approach to countering rival strategies essentially involves denying competitors access to cutting-edge technologies and infrastructure, limiting the ability of other (groups of) countries to develop autonomous technological capabilities, and forcing the non-US market to operate within its own systems and technological framework.
The tactical implementation of these strategic lines translates, vis-à-vis China and the EU, into technology export bans (as in the case of NVIDIA GPUs in China), exclusion from access to services (namely, the exclusion of non-US entities from using Anthropic’s two flagship models), and confining partners to the role of users, through the widespread deployment of US services at every level, thereby making the development of autonomous and sovereign AI capabilities uneconomical.
The instrument implementing these tactics is Big Tech, which is a fundamental part of the operational apparatus of American technological power. Cloud, processors, foundational models, AI services, development platforms and cybersecurity tools are the layers through which the United States makes foreign governments, universities and businesses dependent on American decisions. In this way, formal sovereignty remains intact, but operational autonomy is compromised.
The six elements of the US strategy
The pursuit of these objectives appears to be structured around six elements.
The first is control over the most critical links in the AI production chain and their de facto assimilation into equivalents of defence contractors — companies that are formally private but structurally integrated into the US administrative machinery. Big Tech, albeit with a few exceptions and some resistance, is tasked with limiting the potential for partner countries and non-friendly nations to develop autonomous digital sovereignty.
The second element is pressure on China. Washington has already used export restrictions, technology bans and controls on dual-use goods to limit China’s access to advanced chips and the manufacturing capabilities needed to compete. However, this policy has produced mixed results. It has certainly slowed Beijing down in some sectors, but it has also accelerated the development of domestic supply chains. From the American perspective, the most effective strategy is not absolute decoupling, but controlled interdependence: denying Beijing the resources that would allow it to achieve strategic parity, whilst retaining sufficient leverage to keep it exposed to American pressure.
The third element of the American strategy is the talent market. In AI, people matter at least as much as infrastructure, which transforms human capital into geopolitical assets, and the United States can attract these assets better than anyone else. The natural pool for sourcing talent is undoubtedly Europe, from which to poach excellent researchers who, however, lack access to infrastructure on a par with that of the US. China is not (yet) a market to be considered by Western start-ups, but its academic appeal to European researchers is beginning to be significant.
This leads to the fourth element of the US strategy: the creation and maintenance of European subordination. The European Union regulates technologies developed elsewhere, uses foreign platforms, depends on non-European cloud infrastructure and lacks an autonomous AI capability for defence comparable to that of the US. In other words, the EU is a regulator without infrastructure that can impose rules but cannot exercise power. Directives and regulations serve little purpose if businesses, universities and public administrations require American computing power, American cybersecurity, American processors and American models. The reality is that Europe is not an AI hub, but a hunting ground for Big Tech and, in the US view, must remain so.
The fifth element is the exploitation of the boomerang effect of regulations and directives against the EU itself. For some time now, EU regulations have changed their role: from being a tool for harmonising the internal legal systems of Member States, they have become part of the geopolitical arsenal of an entity that lacks real autonomy and sovereignty. This disconnect between the EU and the interests of the states that have already invested strategically and tangibly in AI allows the US to sit on the sidelines and wait for the inevitable rifts to emerge.
The sixth element is the military dimension, which Europe continues to treat with ambiguity. In strategic terms, a technology that cannot be used for deterrence, intelligence, surveillance, cyber operations, military planning and defence does not generate strength nor allow it to be used as a deterrent. The United States has understood this and is integrating AI into national security and defence. Not so Europe, which, at best, will only receive the versions that Washington decides to make available.
The Chinese reaction
An analysis of Chinese strategies suggests that Beijing is reacting to US policies by attempting to shift the competition into an arena where it dictates the rules, exploiting the self-imposed rigidities within the American system. This strategy can be summarised by the concept of autonomy without isolation. Beijing may accept interaction when it is useful, but rejects dependence when it becomes strategic vulnerability.
US AI, in fact, is based on gigantism: it relies on enormous data centres, massive energy consumption, high-end GPUs and the concentration of computing power in ever-larger facilities. But the US cannot continue to grow indefinitely and will sooner or later have to accept that the growth of infrastructure does not correspond to a proportional or even greater-than-proportional increase in performance. In other words, sooner or later the performance curve of ever-larger systems will flatten out, and every increase in capacity will cost more than it yields.
Probably aware of this structural weakness, the Chinese response therefore consists of making the infrastructure smaller, cheaper, more distributed and less dependent on hardware controlled by the West. This is why frugal AI, optimised models, domestic processors, analogue architectures and robotics are technical options, but organised with a geopolitical function. By reducing the centrality of American hardware, Beijing is evading, at least partially, the logic of containment constructed by Washington.
The European dead end
Assessing the European response is more difficult, because Europe has not yet decided what role it wishes to play. This indecision is clearly evident in the way the EU confuses strategy with regulation, thinking in Hegelian terms that the enactment of an abstract rule can serve a strategic function regardless of how it stands up to reality.
But reality dictates that equipping itself (assuming this is still possible) with sovereign computing, European datasets, foundational models, cybersecurity tools, deployment platforms and defence-specific AI requires a level of political, economic and technological integration that the EU is unable to achieve.
The EU might be tempted to look to the East to escape the paralysis into which it has confined itself. However, regardless of the fact that the US would not allow it, without a clear strategic plan that goes beyond simply ‘funding AI’, the effect of such a choice would merely be to replace one form of subordination with another.
Conclusions
To summarise the geopolitical positions of the three actors in the most extreme terms, the United States can maintain its advantage if China’s autonomy remains incomplete and European sovereignty remains merely rhetorical. China can close the gap by building an alternative technological sphere. Europe can carve out a leading role only by ceasing to be a market to be colonised and becoming a power in its own right.