Artificial intelligence is rapidly transforming digital infrastructure — and, increasingly, energy infrastructure as well. Training large AI models requires massive computing clusters powered by specialized processors that operate continuously at high power densities. As a result, the expansion of AI computing is now directly influencing electricity systems, grid planning, and industrial infrastructure strategies across advanced economies. Yet much of the current policy discussion still treats the energy implications of AI primarily as a question of electricity supply. A systemic perspective suggests a different conclusion: the central constraint is not simply electricity production but the synchronization between electricity generation, transmission infrastructure, territorial grid capacity, and digital infrastructure deployment.

Recent research from the International Energy Agency indicates that electricity demand from data centers is already substantial and growing rapidly. Data centers consumed roughly 415 terawatt-hours of electricity globally in 2024, equivalent to about 1.5% of total global electricity demand, and the growth rate of electricity consumption in the sector has accelerated due to artificial intelligence workloads (International Energy Agency, 2025). AI computing clusters require far more energy than traditional enterprise servers because they rely on high-performance GPUs operating continuously at high utilization rates. According to the IEA, electricity consumption from data centers could more than double by 2030 as artificial intelligence adoption accelerates (International Energy Agency, 2025).

Industry analyses confirm the scale of the emerging demand. Large hyperscale AI campuses often require 100 to 500 megawatts of continuous electricity, and some proposed facilities approach gigawatt-scale power consumption (McKinsey & Company, 2026). These electricity requirements place AI infrastructure firmly within the domain of heavy industrial energy consumers. As a result, decisions about where to locate data centers increasingly depend on electricity infrastructure rather than purely digital considerations.

This shift is already visible in Europe. According to the European Data Centre Association, the growth of artificial intelligence, cloud computing, and digital services is driving unprecedented expansion in European data center capacity (European Data Centre Association, 2026). Data centers have become critical infrastructure supporting financial systems, digital services, and industrial data platforms across the continent.

A recent investment decision illustrates how rapidly this transformation is unfolding. According to Bloomberg, Digital Realty plans to invest €2 billion over the next five years to build new data center campuses in Italy, with facilities planned in Rome and Milan (Bloomberg, 2026). The investment reflects the growing importance of the Mediterranean region in global data traffic flows.

Italy’s geographic position is increasingly valuable within the global internet infrastructure. Subsea fiber-optic cables connecting Europe with Africa, the Middle East, and Asia increasingly land along the Italian coastline, positioning the country as a gateway for intercontinental data traffic (TeleGeography, 2024). Digital infrastructure providers therefore view Italy not only as a national market but also as a strategic node within a wider European digital network.

Digital Realty’s expansion strategy centers on two major facilities. The first is a 62-megawatt campus in Rome, expected to begin operations around 2027. The second is a larger campus in Milan designed to reach approximately 84 megawatts of capacity, initially starting with an 8-megawatt phase before expanding as demand grows (Bloomberg, 2026). Combined, the two facilities will provide approximately 146 megawatts of data center capacity.

From the perspective of national electricity statistics, a project of this size appears manageable. Electricity systems in advanced economies routinely operate at peak demand levels measured in tens of gigawatts. However, interpreting infrastructure feasibility purely through national electricity supply statistics overlooks a critical structural feature of energy systems: electricity networks are spatially constrained physical infrastructures.

Electricity cannot be delivered everywhere simply because it exists somewhere within the system. Transmission lines, substations, and voltage stability constraints determine where large industrial electricity consumers can connect to the grid (International Energy Agency, 2026). As a result, the location of digital infrastructure increasingly depends on the configuration of electricity networks.

This spatial constraint is becoming visible as AI infrastructure expands. The International Energy Agency notes that electricity demand from data centers could increase sharply during the second half of the decade, placing new pressure on electricity networks in regions where digital infrastructure is concentrated (International Energy Agency, 2026). In several advanced economies, the main challenge is not the availability of electricity generation but the ability of transmission infrastructure to deliver electricity to rapidly growing digital clusters.

Financial institutions and energy analysts increasingly recognize this infrastructure bottleneck. Research from Morgan Stanley argues that the global energy sector is now racing to address an emerging “AI power bottleneck,” as the electricity demands of AI computing begin to reshape power markets and grid investment priorities (Morgan Stanley, 2026).

The concentration of data centers in specific geographic regions further amplifies this problem. Academic research suggests that the clustering of AI data centers can create regional electricity system stress even when national electricity supply appears sufficient (Chen et al., 2026). These clusters often develop near financial centers, telecommunications hubs, and major internet exchange points — locations that may already have high electricity demand.

As digital infrastructure expands, the competition for grid connection capacity intensifies. Developers frequently submit grid connection requests in advance of construction in order to secure access to transmission infrastructure. When multiple projects target the same electricity network nodes, connection queues emerge even if the national electricity system has adequate generation capacity.

Electricity planners are increasingly responding with large-scale grid investment programs. The International Energy Agency emphasizes that the expansion of artificial intelligence infrastructure will require not only new electricity generation but also significant investments in transmission networks, storage systems, and grid flexibility technologies (International Energy Agency, 2025).

This infrastructure challenge reflects a broader transformation of the energy–technology nexus. Artificial intelligence is no longer merely a software innovation; it has become a structural driver of energy demand and electricity infrastructure development.

Understanding this transformation requires shifting the policy debate. The key question is not simply whether electricity supply is sufficient. Instead, policymakers and infrastructure planners must ask whether electricity systems can deliver large quantities of power in the right locations, at the right time, and at the scale required by AI infrastructure.

The expansion of data centers in Italy illustrates this systemic challenge clearly. The country is not facing an immediate electricity shortage. Instead, it confronts a coordination problem: aligning energy infrastructure, grid capacity, and digital infrastructure deployment within an increasingly AI-driven digital economy.

In the coming decade, the success of AI infrastructure expansion may depend less on how much electricity countries produce and more on how effectively they coordinate the development of energy systems and digital infrastructure. Artificial intelligence is therefore transforming electricity networks into strategic infrastructure for the digital economy.

Q&A