For much of the artificial intelligence era, discussions about competitiveness focused on algorithms, research talent, and software innovation. However, by 2026, the global AI race has increasingly become a competition for compute capacity. Access to advanced accelerators, hyperscale data centers, high-bandwidth networking infrastructure, and reliable power supplies now determines which organizations can train and deploy frontier AI models. This shift has exposed a growing imbalance between the world’s major economic blocs. The United States controls much of the world’s leading AI infrastructure through NVIDIA and hyperscale cloud providers. China has accelerated efforts to build a sovereign AI ecosystem anchored by domestic hardware development and state-backed investment. Europe, despite possessing critical semiconductor expertise and world-class research institutions, remains heavily dependent on foreign infrastructure. According to economists Alicia García-Herrero and Bertin Martens of Bruegel, that dependence risks creating a structural loss of economic autonomy if left unaddressed. Their proposal is not merely another call for increased investment. Instead, it advocates an Airbus-style industrial strategy that would unite European countries behind a shared AI hardware and compute infrastructure agenda. The proposal raises fundamental questions about digital sovereignty, industrial competitiveness, and the future role of Europe in the global AI economy.
The Global AI Race Has Shifted From Models to Compute
The narrative surrounding artificial intelligence has changed dramatically over the past five years. Early discussions focused largely on breakthroughs in machine learning architectures, advances in large language models, and the capabilities of new AI applications. Today, however, the primary constraint facing many organizations is not access to algorithms but access to compute infrastructure. Training state-of-the-art AI models requires vast clusters of accelerators operating simultaneously across highly specialized data centers. These facilities demand substantial investments in power infrastructure, networking equipment, cooling technologies, and software orchestration systems. As model complexity increases, the importance of compute capacity grows even further because organizations require larger clusters to maintain competitive performance. Consequently, AI development has become increasingly concentrated among a relatively small number of companies capable of funding these investments. The shift has transformed compute from a technical resource into a strategic asset. Governments now view AI infrastructure through the same lens previously applied to telecommunications networks, transportation systems, and energy grids. As a result, national competitiveness increasingly depends on who controls the hardware layer of the AI ecosystem rather than solely who develops the most advanced software.
The emergence of compute as a strategic resource has created new economic and geopolitical dynamics. Access to advanced AI infrastructure determines not only technological leadership but also the ability to support scientific research, healthcare innovation, defense applications, and industrial modernization. Countries that control significant compute resources gain advantages across multiple sectors because AI increasingly influences productivity, automation, and decision-making processes. This reality has prompted governments to develop industrial policies aimed at securing long-term access to compute capacity. In the United States, major cloud providers continue investing hundreds of billions of dollars into AI infrastructure expansion. In China, policymakers support domestic chip development and sovereign AI platforms designed to reduce dependence on foreign technology. Europe, meanwhile, finds itself navigating a more complex landscape. Although the continent remains an important contributor to the global technology ecosystem, it lacks the integrated infrastructure strategy increasingly visible in competing regions. This divergence forms the foundation of the compute gap identified by Bruegel and other policy analysts.
Europe’s Compute Deficit Is Becoming a Structural Problem
Europe’s challenge extends beyond a temporary shortage of hardware. According to the Bruegel analysis, the region faces a structural compute deficit that affects multiple layers of the AI value chain. European organizations often depend on foreign cloud providers for access to advanced accelerators, storage systems, and large-scale training infrastructure. While European startups continue producing innovative AI models, many rely on infrastructure controlled by companies headquartered outside the European Union. This dependence introduces economic, strategic, and operational vulnerabilities that become more significant as AI adoption expands. Furthermore, the gap is not limited to cloud infrastructure. Europe also lacks a dominant domestic producer of advanced AI accelerators capable of competing with NVIDIA’s ecosystem. Consequently, European companies frequently rely on imported hardware to train and deploy advanced AI models. As AI workloads become central to economic growth, these dependencies could constrain Europe’s ability to shape its own technological future.
The scale of the challenge becomes clearer when viewed against global investment trends. American technology companies continue allocating unprecedented levels of capital toward AI infrastructure development. Large hyperscale operators are constructing new campuses, expanding GPU clusters, and securing long-term power agreements to support future workloads. China has pursued a different approach but with similar determination, combining industrial policy with state-backed financing to strengthen domestic compute capabilities. Europe, by contrast, remains fragmented across national markets, regulatory environments, and investment priorities. Individual member states often pursue separate initiatives rather than coordinated infrastructure strategies. Although several European countries have announced AI-related investments, the cumulative scale remains modest compared with developments in the United States and China. This fragmentation reduces purchasing power, limits economies of scale, and weakens Europe’s ability to influence global technology supply chains. As a result, the compute deficit reflects not only infrastructure shortages but also broader coordination challenges within the European innovation ecosystem.
Why Compute Has Become a Sovereignty Issue
The concept of digital sovereignty has evolved significantly during the past decade. Initial discussions focused primarily on data protection, privacy regulation, and cloud governance. However, the rapid rise of AI has expanded the conversation to include compute infrastructure itself. Policymakers increasingly recognize that access to advanced compute capacity may influence economic resilience, national security, and technological competitiveness. If critical AI services depend entirely on infrastructure controlled by foreign entities, governments may face limitations when responding to future geopolitical or economic disruptions. This concern is not purely theoretical. Recent years have demonstrated how export controls, trade restrictions, and supply-chain disruptions can reshape technology markets almost overnight. Consequently, compute infrastructure is beginning to occupy a similar strategic position as energy systems and telecommunications networks. The ability to train, deploy, and operate advanced AI models increasingly depends on resources that governments now view as critical national capabilities.
The sovereignty debate also reflects broader concerns about economic value creation. AI models generate significant downstream benefits across industries ranging from manufacturing and healthcare to logistics and financial services. If the underlying infrastructure remains concentrated outside Europe, a substantial share of future economic value could flow toward foreign technology providers. This outcome would not necessarily prevent European innovation, but it could reduce Europe’s influence over key segments of the digital economy. Furthermore, reliance on external infrastructure may affect Europe’s ability to develop independent policy frameworks for AI deployment and governance. Infrastructure ownership often shapes technological standards, platform ecosystems, and market dynamics. Therefore, the compute gap carries implications that extend far beyond hardware procurement. It influences how Europe participates in the next phase of digital transformation and whether the region retains meaningful control over strategic technologies that increasingly underpin economic activity.
The United States Built an AI Infrastructure Advantage
Understanding Europe’s position requires examining how the United States established its current leadership in AI infrastructure. The American advantage did not emerge solely from breakthroughs in machine learning research. Instead, it resulted from decades of investment across multiple layers of the technology ecosystem. Venture capital markets provided substantial funding for startups and infrastructure projects. Public research institutions contributed foundational innovations in computing and networking. Technology companies developed globally dominant cloud platforms capable of supporting large-scale workloads. Semiconductor firms advanced hardware architectures optimized for increasingly complex computational tasks. Together, these factors created an environment where AI infrastructure could scale rapidly. When generative AI adoption accelerated, American firms already possessed many of the assets required to capitalize on the opportunity.
NVIDIA’s rise illustrates the importance of ecosystem development in achieving technological leadership. The company’s success stems not only from hardware performance but also from software frameworks, developer tools, and ecosystem integration built over many years. As AI demand increased, organizations adopted NVIDIA platforms because they offered mature solutions supported by extensive technical resources. Simultaneously, hyperscale cloud providers invested heavily in data centers, networking systems, and accelerator deployments. These investments created a feedback loop in which infrastructure availability attracted AI developers, whose success generated additional demand for infrastructure. The result is an ecosystem that combines capital, talent, hardware, software, and operational expertise at a scale difficult for competitors to replicate quickly. Europe’s challenge therefore involves competing not with individual companies but with an integrated infrastructure environment that has evolved over decades.
China’s Alternative Path: Building a Sovereign AI Stack
China’s approach differs substantially from the American model, yet it reflects a similar recognition of compute’s strategic importance. Faced with export restrictions and growing geopolitical tensions, Chinese policymakers accelerated efforts to strengthen domestic technology capabilities. These initiatives encompass semiconductor design, accelerator development, cloud infrastructure, and AI software ecosystems. Rather than relying entirely on global supply chains, China seeks greater self-sufficiency across critical components of the AI stack. Companies such as Huawei have expanded investments in alternative accelerator architectures designed to support domestic AI workloads. Government agencies, research institutions, and state-owned enterprises contribute demand that helps sustain these initiatives. The strategy aims not only to reduce dependence on foreign technology but also to ensure continued access to compute resources under evolving geopolitical conditions.
China’s experience offers important lessons for Europe because it demonstrates how industrial policy can influence technology development. While China’s political and economic system differs significantly from Europe’s, both regions face questions about technological dependence and long-term competitiveness. Chinese policymakers recognized early that compute infrastructure would become a strategic asset and responded with coordinated investment programs. Europe has traditionally favored market-driven approaches and regulatory frameworks rather than large-scale industrial coordination. However, the emergence of AI infrastructure as a strategic resource has prompted renewed discussions about whether existing policy tools remain sufficient. The Bruegel proposal emerges within this context. Rather than replicating China’s model, it seeks a distinctly European solution rooted in cross-border cooperation, shared investment, and industrial coordination. The proposal draws inspiration from one of Europe’s most successful examples of collective industrial strategy: Airbus.
The Airbus Model: Europe’s Most Successful Industrial Playbook
The centerpiece of the Bruegel proposal is the idea that Europe should stop treating AI infrastructure as a collection of national projects and instead approach it as a continental industrial challenge. García-Herrero and Martens argue that Europe has already solved a similar problem before through Airbus. During the second half of the twentieth century, European governments recognized that fragmented aerospace industries could not effectively compete against dominant American manufacturers. Individual countries possessed strong engineering capabilities, yet none had sufficient scale to challenge established competitors independently. The solution was to pool expertise, coordinate investment, and distribute production across multiple countries while maintaining a shared strategic objective. Over time, Airbus evolved from a political experiment into one of the world’s largest aerospace companies. Importantly, Airbus succeeded because European governments aligned procurement, industrial policy, research funding, and manufacturing capabilities around a common goal. The Bruegel paper suggests that AI infrastructure now presents a comparable challenge. While Europe possesses significant technical capabilities, those strengths remain dispersed across national boundaries and often operate without sufficient coordination. An Airbus-style framework could transform these fragmented assets into a coherent industrial platform capable of supporting Europe’s long-term AI ambitions.
Applying the Airbus model to AI would not mean creating a direct competitor to NVIDIA overnight. Instead, it would involve building strategic capabilities across several layers of the AI infrastructure stack. Europe could coordinate investments in accelerator design, advanced packaging, photonics, networking technologies, software frameworks, and data center infrastructure. Member states could align procurement strategies across healthcare systems, defense organizations, scientific research institutions, and public administrations. This coordinated demand would create a predictable market for European AI technologies and reduce dependence on foreign suppliers. The objective would not be technological isolation but rather strategic resilience. European organizations would continue participating in global technology ecosystems while developing stronger domestic capabilities in critical areas. Such a strategy would require patience because industrial ecosystems take years to mature. Nevertheless, Airbus demonstrates that Europe can achieve global competitiveness when governments commit to sustained collaboration rather than fragmented competition.
ASML: The Strategic Asset Europe Already Controls
Any discussion of European technology competitiveness inevitably leads to ASML. The Dutch company occupies a unique position within the global semiconductor industry because its lithography systems are essential for manufacturing the world’s most advanced chips. Every major semiconductor producer, including Taiwan Semiconductor Manufacturing Company, Samsung, and Intel, depends on ASML equipment to fabricate cutting-edge processors. As a result, Europe already controls one of the most important bottlenecks in the global semiconductor supply chain. This reality distinguishes Europe from many other regions attempting to strengthen their AI capabilities. Unlike countries that must build entirely new industrial foundations, Europe begins with a globally dominant technology asset that remains difficult to replicate. ASML’s position demonstrates that Europe can produce world-leading industrial champions when the right conditions exist. The question is whether similar success can be extended into other parts of the AI value chain.
The Bruegel analysis argues that ASML should serve as a foundation rather than an endpoint. Europe’s semiconductor ecosystem includes additional strengths through organizations such as IMEC in Belgium, Infineon Technologies in Germany, STMicroelectronics in France and Italy, and numerous research institutions across the continent. Collectively, these organizations contribute expertise in chip design, materials science, manufacturing technologies, and advanced electronics. However, Europe often struggles to convert technical leadership into large-scale commercial ecosystems. Many innovations developed in European laboratories eventually generate economic value elsewhere because downstream infrastructure and investment remain concentrated outside the region. An Airbus-style AI initiative would seek to address this disconnect by linking research excellence with industrial deployment. Rather than viewing ASML as an isolated success story, policymakers could use it as evidence that Europe possesses the technical capabilities necessary to compete when strategic coordination exists.
Building a European AI Hardware Consortium
Creating a European AI hardware consortium would require a broader vision than simply designing new chips. Modern AI infrastructure consists of interconnected systems that include accelerators, networking equipment, software frameworks, power systems, cooling technologies, and cloud platforms. Success depends on integrating these components into reliable and scalable environments capable of supporting demanding workloads. Consequently, any European initiative would need to encompass the entire compute ecosystem rather than focusing narrowly on semiconductor manufacturing. The consortium proposed by Bruegel would likely involve public institutions, private companies, research organizations, and infrastructure providers operating across multiple countries. Such collaboration would help distribute costs while enabling participants to benefit from shared expertise and economies of scale. Importantly, this approach recognizes that no single European country possesses sufficient resources to independently replicate the infrastructure ecosystems emerging in the United States or China. Collective action therefore becomes a practical necessity rather than a political preference.
The consortium model could also help Europe overcome one of its most persistent challenges: market fragmentation. Despite decades of economic integration, technology markets across Europe remain less unified than those in the United States. Different regulatory frameworks, procurement processes, and investment priorities often limit the scalability of emerging technologies. A coordinated AI infrastructure initiative could provide common standards and shared objectives that reduce these barriers. Furthermore, pooled investment would enable larger projects than most individual countries could support independently. Similar approaches have proven effective in sectors such as aerospace, scientific research, and energy infrastructure. While AI presents unique technical challenges, the underlying principle remains consistent. Scale matters, and Europe must find mechanisms that allow it to achieve scale without sacrificing the diversity and innovation that characterize its technology ecosystem.
Public Procurement Could Become Europe’s Secret Weapon
One of the most intriguing aspects of the Bruegel proposal involves public procurement. Governments represent some of the largest purchasers of technology services in the world. Healthcare systems, public administrations, defense organizations, educational institutions, and scientific research facilities collectively generate substantial demand for compute resources. Yet procurement decisions often occur independently across different agencies and countries. As a result, governments fail to leverage their combined purchasing power to shape technology markets. The Airbus strategy suggests a different approach. By coordinating procurement across the European Union, policymakers could create a predictable demand base for domestic AI infrastructure providers. This demand would help reduce commercial risk for companies investing in new technologies while accelerating the development of local ecosystems. Importantly, procurement would function as an industrial policy tool rather than merely an administrative process.
Historical experience suggests that coordinated procurement can influence technological development significantly. Many transformative technologies initially benefited from government demand before achieving broader commercial adoption. The internet, semiconductor manufacturing, aerospace systems, and satellite technologies all illustrate how public-sector purchasing can support innovation. AI infrastructure may follow a similar trajectory. European governments already spend billions of euros annually on digital services, cloud platforms, and research infrastructure. Redirecting a portion of that spending toward coordinated AI initiatives could generate substantial market signals. Furthermore, public-sector workloads often involve sensitive data and critical services, making them particularly relevant to discussions of digital sovereignty. If governments choose to prioritize European infrastructure providers for certain strategic applications, they could help create the scale necessary for domestic ecosystems to develop. This approach would not eliminate competition but could provide emerging technologies with a stronger foundation for growth.
Data Centers and Energy: The Physical Foundations of Compute
The AI infrastructure debate often focuses on chips and algorithms, yet compute ultimately depends on physical infrastructure. Advanced accelerators require specialized data centers capable of supporting high-density workloads. These facilities consume significant amounts of electricity and increasingly rely on sophisticated cooling technologies to maintain operational efficiency. Consequently, energy policy and data center strategy have become inseparable from discussions about AI competitiveness. Europe faces both opportunities and challenges in this area. On one hand, the continent possesses strong renewable energy capabilities and ambitious sustainability goals. On the other hand, electricity costs in many European markets exceed those in competing regions. These dynamics influence investment decisions because energy expenses represent a major component of AI infrastructure operating costs. Any strategy designed to close Europe’s compute gap must therefore address power availability alongside hardware development.
The scale of future demand underscores the importance of this issue. AI workloads continue increasing in complexity, driving demand for larger GPU clusters and higher utilization rates. Data center operators across Europe are already expanding capacity to support cloud services, enterprise applications, and AI deployments. However, infrastructure growth requires long-term planning because power generation, transmission networks, and facility construction involve multi-year development cycles. Policymakers therefore face a dual challenge. They must accelerate AI infrastructure deployment while ensuring that growth aligns with energy security and sustainability objectives. This balancing act will influence where new data centers are built, how they are powered, and which regions emerge as future AI hubs. As a result, the compute gap cannot be understood solely as a technology issue. It also reflects broader questions about industrial infrastructure and resource allocation.
The Obstacles Are Significant
Although the Airbus strategy presents an attractive vision, implementation would encounter substantial challenges. The first obstacle involves funding. Building competitive AI infrastructure requires enormous capital investment across hardware, facilities, software, and research. American technology companies possess market capitalizations and cash reserves that enable rapid expansion. China benefits from state-backed financing mechanisms capable of supporting strategic initiatives over long time horizons. Europe must identify financing models that can achieve comparable scale without relying exclusively on either private capital or public subsidies. This challenge becomes more complex when considering the continent’s diverse economic priorities and fiscal constraints. Securing political agreement on large-scale investments may prove difficult even when policymakers broadly support the underlying objectives. Consequently, funding remains one of the most critical questions facing any European AI strategy.
Political coordination represents an equally significant challenge. The European Union consists of multiple member states with different industrial structures, economic interests, and policy priorities. Achieving consensus on technology initiatives often requires extensive negotiation and compromise. Furthermore, national governments may prefer supporting domestic champions rather than participating in multinational projects. These dynamics have historically complicated industrial policy efforts across Europe. The Airbus example demonstrates that coordination is possible, but it also illustrates the time and political commitment required to achieve success. AI infrastructure development occurs within a rapidly evolving technological landscape, meaning delays can carry substantial opportunity costs. Therefore, policymakers must balance the need for consensus with the urgency of technological competition. Whether Europe can move quickly enough remains one of the central uncertainties surrounding the Bruegel proposal.
What Success Could Look Like by 2035
If Europe successfully implements elements of the Airbus strategy, the benefits may extend far beyond AI infrastructure itself. A stronger compute ecosystem could support scientific research, advanced manufacturing, healthcare innovation, and defense modernization. European startups would gain greater access to local infrastructure, potentially reducing dependence on foreign cloud providers. Universities and research institutions could train advanced models using compute resources aligned with regional priorities. Public-sector organizations might deploy AI systems with greater confidence regarding governance, transparency, and security. Importantly, success would not require Europe to surpass the United States or China in every category. The goal would be establishing sufficient capacity to preserve strategic autonomy while remaining deeply integrated within global technology markets. This distinction matters because the objective is resilience rather than isolation.
By 2035, a successful strategy could result in a diversified European AI ecosystem supported by domestic infrastructure providers, coordinated procurement programs, and expanded semiconductor capabilities. Data center development would likely play a central role, particularly in regions with abundant renewable energy resources. Cross-border research initiatives could accelerate innovation while reducing duplication of effort. Furthermore, stronger infrastructure foundations might encourage additional private investment by reducing uncertainty surrounding long-term market demand. None of these outcomes are guaranteed. However, the alternative pathcontinued dependence on external infrastructure—also carries risks. The Bruegel paper argues that Europe must decide whether it intends to remain primarily a consumer of foreign AI infrastructure or become a more active participant in shaping the next phase of technological development.
Conclusion: Europe’s Next Industrial Test
The debate surrounding Europe’s AI compute gap ultimately concerns more than technology. It reflects broader questions about economic sovereignty, industrial strategy, and long-term competitiveness in an increasingly digital world. The rise of AI has transformed compute infrastructure into a strategic resource that influences innovation, productivity, and national resilience. While Europe continues producing world-class research and maintaining critical positions within global semiconductor supply chains, it remains dependent on external providers for many of the infrastructure capabilities that underpin modern AI systems. García-Herrero and Martens argue that this dependence represents a structural vulnerability that could become more significant as AI adoption expands. Their proposed Airbus-style strategy offers a practical framework for addressing that challenge through coordinated investment, shared procurement, and cross-border collaboration. Whether policymakers embrace that vision remains uncertain. However, the broader issue is unlikely to disappear. As the global AI race increasingly becomes a race for compute, Europe faces a strategic choice that may shape its technological position for decades to come.
