Europe is at risk of becoming an also-ran in the global artificial intelligence race. Despite its economic power and talent base, Europe today significantly underperforms in AI capabilities compared to the United States and China – and even finds itself learning hard lessons from AI military innovations in Ukraine. The central issue is a strategic shortfall in AI infrastructure and investment. Europe controls only a minuscule share of global computing power for AI, lacks indigenous cloud and semiconductor capacity, and remains dependent on foreign tech giants for critical AI resources. This infrastructure gap, coupled with slower investment and fragmented efforts, leaves Europe lagging behind its peers in both civilian and military AI. The following report examines how the AI models work, Europe’s AI shortcomings and the strategic consequences. It also reviews Europe’s initiatives to catch up, assessing whether they are enough to bridge the capability gap before it becomes a permanent strategic liability.

Having a flavour of how AI models are developed can be useful for understanding why this matters. AI model development occurs in two phases: training — which requires massive computing resources and high-cost experimentation — and inference, where trained models are deployed in real-world applications. The training of frontier models like GPT-4 or Gemini Ultra demands vast clusters of advanced GPUs, high energy consumption, and large interdisciplinary R&D teams. Costs have grown exponentially, and only a few actors worldwide, such as the US and China, can compete at this scale. Moreover, these capabilities are no longer confined to commercial or academic use; they are increasingly deployed in national security domains. Ukraine has become a live AI battlefield test ground, where Western technologies like Palantir and Clearview AI are being used in intelligence, targeting, and strategic decision-making.

In this context, Europe’s late start in building AI infrastructure is a strategic vulnerability. As of 2024, European companies control less than 5% of global cloud market share and only 4–5% of global computing power dedicated to AI. Energy costs, regulatory fragmentation, and talent shortage further complicates the issue. To deal with these risks, the EU launched the AI Continent Action Plan, a wide-reaching effort to build a network of AI Factories and Gigafactories, mobilize €200 billion in public-private investment, and foster sovereign capabilities in high-performance computing and defense applications.

This report is aiming to explain how frontier AI models work, details Europe’s AI shortcomings, and analyzes their strategic implications. It then evaluates Europe’s initiatives to close the gap and assesses whether these are sufficient to prevent a long-term dependency on external powers for critical AI systems—especially in defense. 

The question is no longer whether Europe must catch up, but whether it can do so fast enough to avoid a permanent loss of technological sovereignty. Ultimately, we argue that the EU should position itself as a military AI innovation hub for the Western multitech industry, complementing U.S. advances in generative AI. Europe’s geographic proximity to, and deeper involvement in, the Russo-Ukrainian war gives it a unique opportunity to lead in the development and deployment of AI for defense applications.