Five Factors For AI Leadership
What Does It Take to Be an AI Leader? Five Factors in Which Countries Compete for Global AI Leadership
Pia Hüsch | 2025.08.04
Countries across the globe are seeking to harness benefits from AI technologies whole managing the associated risks.
Introduction
Countries across the globe are seeking to secure the benefits from artificial intelligence (AI) technologies to their national advantage. Whether in form of greater efficiency for public services, compensating for skills shortages, enhancing profits for the private sector or ensuring technological advantage for their defence sector, the race for being at the forefront of this technology and the benefits it promises is on. China and the United States, in particular, are deeply engaged in a competition to establish global leadership in AI technologies. But what defines AI leadership, and how can a country achieve it? This paper explores what variables make an AI leader by proposing a five-factor model that sets out which areas of performance are key to secure AI leadership: 1) economic power and private sector strength; 2) innovation and thought leadership; 3) adoption; 4) regulation and governance; 5) tech diplomacy.
The factors at times overlap and are not all of equal importance. The sheer economic power of the United States or China, for example, will be more decisive in establishing AI leadership than Denmark’s tech diplomacy efforts, no matter how skilfully these are pursuit. But it also illustrates that economic might is not the only decisive factor and that other countries, especially those with smaller economies and without a private sector that can compete with the scale of that in the United States or China, can still carve out meaningful areas of strength that are critical in the race for AI leadership.
Factor 1: Economic power and private sector strength
A countries economic power and the strengths of the private sector are critical to establish a leading role for AI technologies. The United States is often considered world-leading in AI technologies, largely due to its leading technology companies like OpenAI, Nvidia, Google DeepMind or Palantir, but also other tech giants like IBM and Microsoft. Similarly, China hosts a number of large tech companies like Alibaba that have firmly established themselves in the AI market. An economic market of sufficient scale to provide growth opportunities for tech companies is key. As is sufficient venture capital to scale up strong AI start ups – something that small tech-powerhouses such as Estonia or even larger economies such as the United Kingdom, Germany or Japan struggle with. Even where their existing strengths in academic and early-stage research puts them in a strong position, the ability to scale start ups up often heavily relies on US or other foreign funding. This poses the risk that companies move out of these smaller countries which thereby lose the opportunity to retain sovereign capabilities. DeepMind, a UK start up acquired by the US tech giant Google, is a prime example of this dilemma.
Yet the EU demonstrates that a large economic market is not the only decisive factor for AI leadership. It must be paired with a strong private sector that has the ability to attract and retain topic talent, fund cutting edge research and development and, perhaps decisively, finance the necessary infrastructure required to develop AI technologies. AI development requires both large amounts of quality data and computing power. Building data centres to store data and the compute necessary to train AI models requires significant resources – including energy and financial power but also access to the right underlying technologies. Compute is highly dependent on the availability to cutting edge semiconductors, traded through a complex global supply chain featuring multiple bottlenecks – most famously the chip manufacturer TSCM in Taiwan and the Dutch company ASML, which holds the monopoly in the production of extreme ultraviolet lithography machines used to produce the most advanced chips.
The 500 billion US dollar budget of the US-led Stargate project, led by the US company OpenAI and meant to make the United States “a global AI leader”, illustrates the dimensions of investment at stake when it comes to building cutting edge AI infrastructure. Similarly, US companies like AWS, Microsoft and IBM are investing heavily in data centres, including in the Middle East. Despite global collaborations such as under Stargate, US companies remain at the forefront of these initiatives.
China has also been investing in large infrastructure projects tailored for the development of AI. It has been successful in leveraging its government structures to turn economic might to strategic projects, providing leadership and significant resources to strategically position its technology sector. Some have described Chinese efforts to build infrastructure and data centres as “a chaotic building boom”, pointing to inexperienced local governments and companies jumping on the AI hype but struggling to keep data centres afloat. Still, the availability of large data sets and cheap energy have been seen as China’s advantages, while the lack of access to cutting edge semiconductors has often been named as the reason the United States retains an advantage over Chinese AI developments. Chinese private sector with companies such as Alibaba and Tencent have also massively invested in AI infrastructure.
Despite differences in government and market systems, both Chinese and American AI leadership ambitions are rooted in their strong economic positions with which other countries cannot compete. Currently hosting about three quarters of AI supercomputer performance, the United States’ strong infrastructure comfortably secures it the world-leading position under this factor. With China in second place, traditional computing powers like Japan, Germany and France are under pressure, arguably now playing “marginal roles in the AI supercomputing landscape”.
Factor 2: Innovation and thought leadership
The launch in early 2025 of DeepSeek, the Chinese answer to OpenAI’s ChatGPT, caused a wave of panic among Western AI enthusiasts. Was this a high-performance AI model, trained for a fraction of the estimated cost to train ChatGPT-4o and on less computational power, proof that China was overtaking the United States in AI competition? Moreover, as the model was built without access to the latest semiconductor chips, does that mean US export controls are failing?
There has been disagreement as to how truly “innovative” DeepSeek is. It became evident that the chips it was built on were sourced before the introduction of export controls and that the model is trained on OpenAI’s models. Some described DeepSeek’s models as “impressive” as “they do the same for less”, while not providing a “step change in capability”. Yet the US market’s plunge after its publication caused considerable concern. While some commentators found that US export controls “fuelled innovation” within China due to limited access to chips and that DeepSeek’s cheap and quick development is a “win for innovation”, others see it as a reminder that access to chips remains a severe restriction to genuine AI innovation in China.
Whether or not DeepSeek was truly “innovative” certainly depends on one’s definition of innovation. At the very least, it was a strong signal that beyond these large models that require enormous infrastructure in the form of data and compute that are really only accessible in the United States, innovation can also take place through smaller models. DeepSeek, if anything, was a cautionary tale to not exclusively focus on the outputs of large-scale infrastructure projects as innovation can materialise in different forms and applications. Depending on adoption and societal impact, it is certainly feasible that other countries may advance innovation and thought leadership in AI technologies, for example for niche applications or by lowering production costs. As it stands, however, the United States remains world leading in terms of AI innovation.
Factor 3: Adoption
Much of the discussion on AI leadership circulates around the development of cutting-edge capabilities, either for national security and defence purposes or those of wider public appeal, such as AI generated images. What this discussion often overlooks is that these capabilities often only represent the tip of the iceberg. And while it is highly relevant to states to maintain the technological edge in a national security context, skilled adoption of AI technologies and wider integration is critical to harness benefits on a societal level, including for economic purposes. Rolling out AI technologies in the energy or transport sector promises greater efficiency and better public services. But the technology behind these applications does not necessarily have to be cutting edge. Instead, it is the AI-human teaming, a skilled workforce that is empowered to make the most out of the technologies in a timely fashion, successful risk management, and the ability to identify existing processes that benefit from new technology enabled opportunities that make for an impactful adoption strategy. Sometimes, there is an overemphasis on AI technologies where a digitalisation process is not even concluded.
Roll out and adoption is by no means restricted to large economies like the United States or China. The United States, for example, seems more concerned about developing frontier capabilities than facilitating wider roll out. China, instead, focuses on application areas such as smart cities or robotics for its manufacturing sector. Smaller countries may prove to be more agile or better placed to technology adoption, for example because of their societal attitude to technology pick up or because of an already digital society that provides quality data. Estonia, a highly digitalised nation, could become leading in AI adoption for public services given its public sector services are already highly digitalised. Finally, AI assurance through regulation, testing and rigorous safety requirements may provide the public with greater confidence to adopt these technologies. The United Kingdom is an example of a middle-power leading on AI safety work. Adoption alone, however, means a country remains a consumer of AI technologies, not a developer or innovator, and as such cannot manifest leadership on its own.
Factor 4: Regulation and AI governance
Governance of AI technologies remains a highly relevant issue. While the EU tries to clarify its comprehensive regulatory framework, the EU AI Act, the United States exercises pressure on the EU to abandon the voluntary code of practice preceding the full act. The United States has made clear that the Trump Administration holds on to a light touch approach to regulating its tech sector – even if at odds with allies’ concerns over AI safety and security. Meanwhile, many other countries such as the United Kingdom are still figuring out how to strike the right balance between innovation friendliness and risk management. Initially, the British approach advanced five voluntary principles for regulation, but the government led by Prime Minister Keir Starmer announced that it was pursuing binding regulation for frontier AI models. What this entails exactly is still to be seen. Similarly, Japan is currently committed to voluntary principles with emphasis on sectoral regulation.
Although many countries are still facing the same questions around getting this balance of regulation right and the EU is often seen as a leader in AI regulation, there remains a gap in international best practice for technology regulation that also allows for innovation. Getting this balance right is a powerful way in which other countries can mark their spot in the AI leadership race, for example by providing regulatory stability and attracting AI innovators to their jurisdiction. In the absence of US leadership on global AI governance, China has recommended a global AI governance body, illustrating that AI governance remains a factor where countries want to be seen as leaders.
Factor 5: Tech diplomacy
A final factor in AI leadership is international tech diplomacy, an effective tool to forge international alliances and advance soft power. The “America first” approach championed by the second Trump Administration, however, does not emphasise international tech diplomacy as a key priority. Instead, the Trump Administration’s AI Action Plan arguably “seeks customers, not partners”. China, by contrast, has long been a successful tech diplomat, particularly with respect to telecommunications infrastructure in the global south and in tech standard setting bodies such as the International Telecommunication Union, and is continuing its diplomatic engagements backed by its manufacturing capabilities providing access to cheap infrastructure.
Both the United Kingdom and the EU have also positioned themselves as AI and broader tech diplomats. But while the United Kingdom-led AI safety summit in November 2023 was a success in putting London and AI safety on the international agenda, the legacy of the AI safety – and later security – summits are unclear. The 2025 Paris summit, for example, was not a beacon of hope for tech diplomats but instead, a sobering event demonstrating how difficult it is to find consensus among key AI powers. The fact that the United Kingdom and the United States did not sign the Paris summit declaration was seen as a strong concession from the British government to keep the Trump Administration happy at the early days of the new administration and leaves open questions as to the future of the AI Summits. Similarly, the voluntary nature of the principles suggested under Japanese leadership face significant challenges. Without binding commitments, they are fragile to bow down to (perceived) economic and national pressures.
This illustrates how fragile tech diplomacy is as a pillar of AI leadership: it ultimately remains subject to external pressures. As such, tech diplomacy can enhance tech relations but will not be able to make a country an AI leader unless it is backed by elements of other pillars, for example Chinese economic power.
Conclusion
AI leadership has different facets and by no means is purely restricted to the development side. The suggested five factor model illustrates that while the economic power and a strong private sector remain the most important factor for AI leadership, especially as they support additional factors such as innovation and thought leadership, other factors offer opportunities for smaller and middle power countries to establish themselves among a group of AI leaders. For example, niche application areas, innovative, smaller models, the ability to adopt AI technologies at speed and on a societal level or striking the right balance in regulation between risk management and innovation friendliness can make middle ground countries develop strong positions in the race for AI leadership. While the global race for AI leadership remains one between the United States and China, smaller countries should focus on their ability to make themselves indispensable to the global supply chain and double down on existing strengths to leverage these in order to be able to offer something that the scale of China and the United States cannot provide.
Pia Hüsch is a Research Fellow in cyber, technology and national security. Her research focusses on the impact, societal risks and lawfulness of cyber operations and the geopolitical and national security implications of disruptive technologies, such as AI.