Europe’s AI ambitions are colliding with physical limits that strategy papers rarely acknowledge. As AI moves from experimentation to infrastructure-scale deployment, the continent faces a harder question about whether its networks, energy systems and digital foundations are fit for what comes next.
Europe is entering the AI era with ambition, talent, and capital, but without the physical foundations required to sustain it. That is the central warning of AI is too big for the European internet, a new report from Nokia based on a survey of more than 1,000 senior European executives across industry and telecommunications. The conclusion is blunt rather than alarmist: AI adoption is accelerating faster than Europe’s networks, energy systems, and digital infrastructure can support, and the gap between aspiration and capability is widening rather than closing
This is not a story about missed innovation or insufficient enthusiasm. Two thirds of European enterprises are already using AI in some form, with a further cohort actively piloting deployment. Business leaders overwhelmingly view AI as transformational rather than speculative. The problem is that Europe is attempting to scale AI on infrastructure designed for a different internet, one built for content delivery and transactional cloud services, not for power-dense, latency-sensitive, data-hungry AI workloads operating at industrial scale.
The report’s argument matters because it reframes the AI debate away from models, funding rounds, and national strategies, and toward something far more prosaic but decisive: whether Europe’s physical and digital backbone can carry the AI supercycle it claims to want.
The productivity imperative Europe cannot defer
Europe’s interest in AI is not optional. Productivity growth across the continent has stagnated for more than two decades, running at roughly half the rate of the United States. This structural weakness now collides with a geopolitical environment in which technological capability increasingly determines economic and political autonomy.
The report situates AI as Europe’s best remaining lever to reverse this trend. Across automotive manufacturing, pharmaceuticals, energy management, and logistics, early adopters are already using AI to reduce energy consumption, compress time to market, and automate complex decision-making. These gains are not theoretical. They are being realised inside operational systems, not experimental labs.
Yet the report exposes a dangerous asymmetry. While executives expect AI to transform everything from cybersecurity to supply chain optimisation, they are far less confident that Europe’s infrastructure can support those ambitions. More than half of surveyed enterprises already report network performance problems linked to latency, throughput, or reliability. For many, these issues are appearing before AI has even reached full scale.
This is the first structural contradiction Europe must confront. AI’s value compounds with scale, but Europe’s infrastructure degrades as demand rises.
Infrastructure, not algorithms, is the binding constraint
The report is explicit in rejecting the idea that Europe’s AI challenge is primarily about talent, regulation, or capital. Those issues exist, but they are secondary to the physical constraints now shaping deployment decisions.
Energy emerges as the most immediate pressure point. Nearly nine in ten executives express concern that Europe’s energy systems cannot keep pace with AI demand, with more than half describing the situation as already strained or at serious risk. These constraints are no longer abstract. Over one fifth of firms report that energy availability is directly delaying AI projects, while others are being forced to rethink site selection, deployment timing, or system architecture altogether.
Connectivity is the deeper, longer-term problem. AI workloads generate traffic patterns that look nothing like traditional cloud usage. Training, inference, model orchestration, and real-time decision systems place extreme demands on latency, resilience, and east-west traffic inside and between data centres. According to the report, global data traffic is projected to increase five to nine times by 2033, with AI alone accounting for roughly one third of that growth.
Europe’s networks were not built for this. Fragmented markets, under-scaled operators, and uneven fibre and 5G deployment leave the continent structurally disadvantaged. The average European telecoms operator serves a fraction of the customer base of its US or Asian counterparts, limiting investment capacity and slowing the rollout of next-generation infrastructure.
The consequence is already visible. AI deployments are hitting performance ceilings, not because models are insufficient, but because the network cannot keep up.
The quiet erosion of digital sovereignty
Infrastructure weakness has a political dimension the report does not shy away from. Nearly a third of executives warn that infrastructure limitations may force them to move AI workloads outside Europe. In practice, this means exporting data, computation, and innovation to regions with cheaper power, larger data centres, and more capable networks.
This is not hypothetical. Many firms are already considering relocation or have begun moving compute abroad. Others are responding by building private AI infrastructure to bypass public network limitations. While rational in isolation, these responses fragment Europe’s digital landscape further and undermine the very economies of scale required to compete globally.
The report draws a direct parallel with Europe’s historic dependence on external energy suppliers. Just as energy dependency constrained political autonomy, reliance on foreign AI infrastructure risks eroding digital sovereignty at precisely the moment Europe claims to prioritise it.
Sovereignty in this context is not a slogan. It is a function of control over networks, data flows, and the systems that secure them. Without trusted connectivity, sovereignty becomes performative rather than operational.
Security as both risk and justification
Cybersecurity occupies a central position in the report’s findings, not only as an AI use case, but as a structural justification for infrastructure reform. Almost all surveyed executives believe AI introduces new or evolving security risks, with concern centred on the increasing sophistication and automation of attacks.
At the same time, cybersecurity is the single most common AI application area cited by European firms. This reflects a paradox. AI is simultaneously amplifying threat vectors and becoming essential to managing them. That dual role places even greater pressure on infrastructure, because security systems demand reliability, low latency, and trusted data paths.
The report warns that sovereignty without security is meaningless. Significant portions of Europe’s telecoms infrastructure remain dependent on high-risk vendors, exposing critical systems to potential compromise. AI intensifies this exposure, as automated systems magnify both capability and consequence. In this context, investment in trusted, AI-ready networks is not merely about performance. It is about risk containment at systemic scale.
Fragmentation as Europe’s self-inflicted wound
Perhaps the report’s most uncomfortable argument concerns market structure. Europe’s telecoms fragmentation is not an accident, but a policy outcome. Competition rules designed for consumer pricing have produced an ecosystem unable to support industrial-scale AI infrastructure.
The contrast with other regions is stark. China, India, and the United States have all achieved far higher levels of 5G standalone deployment, with corresponding gains in speed, reliability, and capacity. Europe lags not because of technological inferiority, but because scale has been structurally discouraged.
The report argues that consolidation is no longer optional. Without larger operators capable of sustained capital investment, Europe cannot build the networks AI requires. This is a politically sensitive claim, but one grounded in physical reality rather than ideology. AI does not scale politely. It demands concentration of power, compute, and connectivity. Fragmentation throttles all three.
Policy ambition must translate into concrete build-out
Europe is not standing still. Initiatives such as InvestAI and proposed AI gigafactories signal a shift from rhetoric to capital deployment. Venture funding for AI startups is increasing, and political leaders increasingly frame AI as a strategic priority.
The report acknowledges these developments but stresses that they will fail without corresponding investment in the connective tissue that links data centres, enterprises, and edge systems. Training large models inside sovereign facilities means little if inference traffic bottlenecks at national borders or congested metro networks.
Similarly, building private AI clouds may satisfy short-term security concerns while exacerbating long-term inefficiency. Without shared infrastructure, Europe risks recreating its telecoms problem inside AI itself, a patchwork of isolated systems incapable of global competition.
The report’s recommendations are pragmatic rather than radical. Pool capital, encourage cross-border coordination, align competition policy with infrastructure realities, and treat connectivity as critical national infrastructure rather than a commodity service.
The cost of delay is structural, not cyclical
The most sobering insight in the report is temporal. Infrastructure gaps compound over time. Every year of underinvestment widens the performance gap, increases dependence on external providers, and limits strategic autonomy. AI does not wait for regulatory alignment or multi-year planning cycles. It rewards speed, scale, and reliability. Europe’s risk is not falling behind temporarily but locking itself into a permanently subordinate position as an AI consumer rather than a creator.
Executives understand this. More than eighty per cent believe Europe must either lead or at least keep pace in AI-ready infrastructure. The ambition is there. What remains uncertain is whether policy, industry, and telecoms providers can align quickly enough to deliver the physical systems that ambition requires.
The report’s final message is not pessimistic, but conditional. Europe can still succeed in the AI era, but only if it recognises infrastructure as strategy, not support. Networks, power systems, and secure connectivity are not enablers of AI. They are its precondition.
Without them, Europe’s AI narrative collapses into contradiction: world-class models constrained by third-rate networks, sovereign ambitions dependent on foreign infrastructure, and productivity gains throttled by physical bottlenecks.
AI is not too big for Europe. But it is already too big for the internet Europe has built. Whether that changes will determine whether the continent shapes the AI supercycle or merely adapts to it.
