Could the AI Race be a Geoeconomic Trap?

June 12, 2026 - Written by Talha Haroon

Prior to the Choose France summit held at the Palace of Versailles last week, SoftBank's Masayoshi Son announced a €75 billion investment package to build AI data centres across Northern France. With the first deal of its kind for SoftBank in Europe, the headlines were predictably rapturous: Son’s backing of the $500 billion Stargate AI campus in the United States, built alongside Sam Altman's OpenAI and Larry Ellison's Oracle, has cemented the Japanese institution as one of the largest financiers of global AI infrastructure spending. However, the deal remains squarely focused on the buildout of data centres, and not AI-related research or start-ups. This raises a crucial question for the global AI race: with every nation seemingly devoted to adopting and specialising in this revolutionary technology, is anyone truly able to pull ahead? 

A new study from researchers at the Sorbonne, University Dauphine-PSL, Télécom Paris, and Agoranov offers the first data-driven answer to that question. Applying economic complexity theory – the model originally developed to measure an economy's development by the diversity and rarity of products it exports – to venture capital data across 17 countries and 18 emerging technology domains, they rank both countries and technologies by the degree to which specialization in them actually confers geoeconomic leverage. The results unsettle a lot of assumptions, particularly for middle powers like France caught between the technological poles of the US and China in the global AI race.

Where Does the Money Actually Flow? 

For many nations in the industrialised world, pouring billions into AI strategies has taken on striking uniformity. The United States invested approximately $109.1 billion in 2024 alone on AI-related sectors and projects. Gulf economies such as the UAE and Saudi Arabia have invested $148 billion since the start of 2024 and earmarked more than $40 billion for AI-related investments. And analysts predict that China’s AI industry and related sectors could grow into a market valued at $1.4 trillion by 2030. The race to develop and dominate this technology has become a financial juggernaut. 

But geoeconomic leverage is inherently a question of relative superiority. For all of these nations, the AI buildout is more an attempt to secure technological sovereignty rather than benefit from the efficiency gains of the technology. As the researchers highlight, the AI strategy is “a necessary condition for strategic autonomy, even if its interpretation and operationalisation vary across geopolitical, economic, and institutional contexts”. That argument certainly holds weight: after all, AI has proven to have a multiplier effect for the value of every other frontier technology, from defense systems and drug discovery to cybersecurity infrastructure and beyond. If it supercharges every other domain, surely dominating AI is tantamount to pulling ahead in the other 17 of the technology basket. 

Not, as the study argues, when we assume that a technology's strategic value is determined equally by a country’s specialisation in a frontier technology and the relative mediocrity of others. A domain in which dozens of nations are simultaneously building data centres may be economically significant, but it confers little leverage. True geoeconomic power flows from specialisation in domains that are rare, concentrated among a small number of leading nations, and structurally difficult to replicate.

A New Measure of Technological Power

The standard measure of technological sovereignty is rooted in the degree to which investments in lucrative sectors result in the greatest control in global supply chains; the countries that spend the most tend to win out in the end. However, the researchers adopt an alternative gauge by focusing on geoeconomic complexity, which measures relative strategic exclusivity and not absolute economic importance. This is done by constructing two interlocking indices.

The first, a Geoeconomic Complexity Index (GCI), ranks countries not by the volume of their venture investment portfolios but by the strategic quality of their composition – whether their start-up ecosystems are concentrated in domains that only a handful of leading nations have explored. The second, an Emerging Technology Geoeconomic Complexity Index (ETGCI), ranks technology domains such as quantum computing, fusion energy, space technologies, AI and others, by whether specialisation in them is concentrated among the highest-ranked countries. 

The two indices are deliberately recursive: a country's standing rises if it specialises in strategically valuable technologies, and a technology's value rises if strategically powerful countries specialise in it. Neither can be gamed out in isolation, meaning that a nation cannot manufacture strategic leverage simply by spending more on a single technology. What emerges when analysing the two indices together is a striking pattern: the countries at the top – what the researchers term “high-diversity, low-ubiquity” nations – have start-up ecosystems which are present across multiple emerging domains, but the domains they dominate are ones few others have entered. It is not enough, in other words, to be everywhere in the technology mix: what matters is investing in sectors which other countries are not. 

For states such as the United States or China, this combination of the GCI and ETGCI showcases the strength of “high levels of diversity” and low “average ubiquities”, indicating not only “broad specialisation” but “strategic concentration of venture investments in less frequent technological domains”. In other words, AI investment is not the sole funnel for capital deployment in the technological domain for these nations. Consider nuclear fusion in America: American companies account for 53 percent of all global private fusion investment, with MIT spinout Commonwealth Fusion Systems alone having raised nearly $3 billion, representing roughly a third of all private fusion capital worldwide. This is precisely what the study identifies as geoeconomic leverage: concentrated positions in domains with limited external competition. 

Moreover, when focusing more closely on ETGCI mapped across the 18 technology domains studied, AI ranks only sixth in strategic importance. Cloud computing, cybersecurity and medtech occupied the top tier of the index, highlighting greater geoeconomic complexity. This is crucial for middle powers: take Sweden, for example, which intends to be “top-five in the world by 2035” in AI. The study finds that it exhibits low diversity and high ubiquity in the cross-indices analysis, highlighting its specialisation in technological domains which are being globally pursued. The rationale for these results resembles a prisoner’s dilemma: since countries seek to specialize in the AI race, including many middle powers, the exclusivity of control over the technology decreases, resulting in lower opportunities for geoeconomic advantage and thus diminished technological sovereignty. 

The Path Ahead for Middle Powers

For each country, the researchers identify a “Simplest Single Sovereignty Enhancing Technology" (SSSET) metric, defined as the technological specialisation which would deliver the greatest improvement in relative geoeconomic standing. For middle powers like the Netherlands, the metric identifies that investment in cybersecurity systems would lift the country four places in overall geoeconomic leverage; an AI-focused strategy, on the contrary, offers no comparable gain. Similarly, for a nation like Sweden, focusing on cybersecurity tools or medtech would deliver a seven-place gain in geoeconomic leverage. 

For middle powers, putting these findings into practice involve three key actions. First, identifying a country's SSSET will be fundamental to understanding not only the technological edge a nation possesses, but where capital should be directed for the greatest return. Second, implementing public-private partnerships on the demand side: by procuring leverage-enhancing technologies, governments can anchor demand for start-ups in rare specialisations, transforming strategic intent into signals for investors and global markets. And last but not least, on the supply side, government policy should focus on developing incubation support and extending R&D subsidies in domains with direct pay-offs for industrial strategy. This means targeted support that rewards innovation in otherwise underfunded technological domains. Crucially, funnelling investments into areas that are underdeveloped globally is a powerful takeaway for middle powers.

Circling back to France, the study highlights the country’s geoeconomic leverage via sustained specialisations in medtech. That advantage is visible across the nation, whether that’s  Toulouse, where decades of precision aerospace engineering have been redirected into robotic surgery systems and advanced imaging technologies, or Montpellier's Med Robotics Place – a dedicated hub for medical robotics start-ups. These are domains where investment in a thriving and growing start-up ecosystem have led to the greatest dividend. Yet, the €75 billion now being pumped into data centres risks reinforcing the very dynamic identified by the study: trading specialisation in rare technologies for compute capacity whose leverage erodes with every AI strategy implemented elsewhere. Middle powers betting on the data centre buildout may win the AI race but lose the broader geoeconomic one.

Written by Talha Haroon

Analyst for Geopol Report