The global AI race, long dominated by the US and China, may become multipolar and more fragmented if current trends continue. This development would enable a wider range of countries to develop competitive AI ecosystems.
The AI race is frequently presented as a race between the US and China, in which other countries may be able to build specific niches, but will ultimately remain reliant on the solutions and underpinning infrastructure provided by either of the two superpowers. At present, the US and China remain firmly in the lead when it comes to the development of frontier AI – with most research, investment, computing power, infrastructure availability, talent, patents and new product deployment centred around those two markets. Other countries are usually afforded little agency in this framing.
As the global security order deteriorates, countries worldwide are becoming increasingly concerned that, in a more hostile, transactional new world, overreliance on others for pivotal technologies like AI is a source of profound vulnerability.
However, geopolitical tensions and the rapid securitization of the AI sphere may start to challenge the status quo. As the global security order deteriorates, countries worldwide are becoming increasingly concerned that, in a more hostile, transactional new world, overreliance on others for pivotal technologies like AI has become a source of profound vulnerability. This fear has led to a growing number of countries seeking trusted technology alternatives with greater urgency, and encourage efforts to bolster their domestic AI industries. The increasing level of entanglement between private sector AI companies and governments and militaries has begun to prompt similar concerns over external dependencies.
This evolving security environment also provides a catalyst for this change. High defence spending (which has already reached a level not seen since the early Cold War) – with growing shares allocated to AI development – could see smaller markets benefit from spillover effects into their wider AI ecosystems and help their domestic solutions to achieve scale. A rapidly growing global ecosystem of ‘dual-use’ AI solutions may similarly enable smaller countries and developers to gain a larger market share.
This paper argues that the dynamics described above may result in a more securitized, multipolar – and likely also fragmented – global AI industry, in which it will become more difficult for one or two actors to substantially dominate the value chain or the technology’s wider rollout. To understand how this scenario may come about, the paper explores four important trends that accelerated in 2025 and early 2026, and which could start a shift towards a multipolar AI market:
- The dual-use and defence tech boom. In 2025, interest and growth in the market for defence and dual-use AI applications accelerated. Though the majority of leading commercial AI companies are still based in either the US or China, the market for dual-use AI applications has the potential to become more geographically diverse. For example, the European Union (EU), Israel, South Korea, the UK and Ukraine are all developing their own increasingly vibrant ecosystems. High levels of defence spending, mounting concerns over sovereignty and efforts by governments to put cutting-edge innovation at the heart of rearmament efforts, are all helping to fuel this growth.
The growth of innovation ecosystems on the back of military spending would not be without precedent. Silicon Valley itself has its origins in defence contracting, with many of its technologies ranging from semiconductors to GPS – even the internet – having been built as part of Cold War military contracts. In the long term, spillover benefits from dual-use innovation, as well as ecosystem growth fuelled by increased government spending and domestic adoption of AI, could see smaller AI ecosystems become more mature and better able to compete.
- The rise of ‘patriotic tech’, blurring the boundaries between civil and military use. 2025 was characterized by an increasingly more intimate relationship between AI companies and governments, with many commercial companies now willing to explore the defence and national security applications of their AI products. Though this is a global dynamic, the trend is especially visible in the US, driven not just by new defence tech companies entering the market, but also by established Silicon Valley ‘hyperscalers’ like Alphabet and Microsoft and frontier AI labs like OpenAI and Anthropic pursuing contracts with the US Department of Defense and allied defence ministries, and presenting themselves as working in support of US government (and, by extension, NATO) objectives.
Closer collaboration between tech companies and the military will likely draw in larger parts of the economy, which would especially benefit smaller markets that can then allocate finite resources more effectively towards shared strategic ends. This increased entanglement and mutual dependence may, however, give rise to a new powerful type of tech–industrial complex, with tech companies and militaries become increasingly mutually interdependent. Such a development could sow further distrust between countries, and encourage further decoupling efforts. The recent spat between the Pentagon and Anthropic also shows the perils of this kind of intimate relationship for the private sector.
- A global push towards sovereignty and decoupling of interdependencies. Globally, there has been a clear push by governments to become more self-reliant in defence, AI and technology more widely. As the geopolitical climate grows progressively more tense and the geostrategic utility of having independent AI capabilities becomes more evident, many countries have started to reassess the composition of their technology stacks. This national security-driven reassessment has put a renewed focus on the development of sovereign capabilities and resilient supply chains (even if that comes at a higher cost or means a transition towards less mature solutions), as countries seek to disentangle themselves from a perceived over-reliance on foreign AI and AI infrastructure suppliers.
This concern over sovereignty may see more competitive alternative AI ecosystems emerge. Rhetoric around sovereignty is also likely to become a growing source of friction between countries, as Washington and Beijing will seek to preserve the market positions of their respective AI ‘champions’.
- Growing concerns about a potential AI valuation bubble. Throughout 2025, concern became increasingly apparent over the levels of spending by and valuations of companies in the AI and wider tech industry. Large US-based hyperscalers continued to invest significant amounts – over $300 billion in 2025 alone – into new data centres, computing resources and other infrastructure to maintain an edge in the race to develop frontier AI capabilities. Investors, meanwhile, are becoming more sceptical about the likelihood of frontier AI spending generating the promised financial returns.
The securitization of the AI race may allow such high levels of spending to persist for longer, as geostrategic considerations can put pressure on governments to step in and keep the market afloat. The big AI spenders are already pursuing government contracts and support, further amplifying the trend towards ‘patriotic tech’ discussed above.
A potential market correction could, however, also lead to the geostrategic competition being reconfigured, as focus could move away from highly capital-intensive frontier AI development towards cheaper, open-source models.
As access to investment and cutting-edge infrastructure remains among the main barriers to smaller AI ecosystems catching up with the major players, a shift towards ‘good enough’ tech development, and rapid diffusion and adoption could allow more actors to gain a larger market share. If the AI race is considered as three parallel races – the race towards the frontier (the development of the most cutting-edge AI), the race towards diffusion (spreading AI throughout the economy), and the race towards application (finding different use cases for the technology) – emphasis may shift towards the latter two. In these two races, China, but also smaller actors, will find themselves better positioned.
About this paper
The remainder of this paper is made up of two parts. The first part focuses on the four key trends outlined above. These four trends were selected on the basis of an extensive literature review, and were stress-tested by a diverse range of domain and regional experts.
The final part draws conclusions and presents recommendations on how the private sector can best prepare for a more geopolitically charged and multipolar AI environment. While the art of prediction is necessarily inexact, by extrapolating from ongoing trends and drawing on subject expertise, the paper seeks to provide
decision-makers with the tools to better anticipate changes to their operational environment.
The paper focuses primarily on the political economy of AI, especially dual-use AI systems. Critical though other issues may be, the analysis does not extend to the ethical implications, or actual battlefield utility, of AI solutions. Nor does the paper seek to comment on how dual-use technologies should be governed.
Though this paper draws on existing trends and developments in AI, its conclusions and the trajectories it sets out are inevitably speculative. Many of the dynamics discussed remain uncertain, and are subject to a highly volatile geopolitical environment and rapidly evolving AI space, in which assessments of the technology’s current state can become obsolete in months or even weeks.
However, given the potentially transformative impact that securitization of AI may have on the global marketplace, and the reconfiguration of the AI race that may result, decision-makers in business (as well as in government and elsewhere) must be proactive in thinking about what this direction of travel may mean for their own operations.
