The rapid expansion of data-intensive technologies such as hyperscale data centres, particularly artificial intelligence (AI) data centres, has significant implications for energy system reliability and resilience, warn experts at UNECE. This is in addition to important implications for infrastructure investment, water and land use, economic development, digital sovereignty, and local communities.
Global electricity consumption by data centres is projected to increase from approximately 485 TWh in 2025 to 950 TWh by 2030, equivalent to 3% of global demand. Additional demand comes from cryptocurrency mining (100–130 TWh annually) and industrial electrification. This growth is rapid and highly concentrated geographically. Capital expenditure on data centres worldwide is forecast to rise from roughly US$800 billion per year in 2026 to $1.8 trillion per year in 2050
As outlined in a new paper prepared to inform debates at UNECE’s intergovernmental Committee on Sustainable Energy later this month, in many regions, demand growth from large-load facilities is outpacing the expansion of electricity infrastructure: while the former can be deployed within 2–5 years, transmission networks often require more than a decade due to planning and permitting constraints.
This mismatch is already visible in several jurisdictions: Ireland has introduced connection restrictions in Dublin; the Netherlands has tightened planning decisions due to infrastructure and land constraints; in the United States, large loads, particularly data centres, are projected to drive significant increases in equipment demand and supply chain pressures by 2030.
Highlighting limited visibility of large-load behaviour as a key operational risk, the paper calls for enhanced data transparency, standardized reporting, and real-time information exchange to support system planning and operation. It also calls for stronger coordination among developers, system operators, regulators, local authorities, and agencies responsible for water, land use, and infrastructure.
Developed by an international Task Force on Digitalization in Energy, the paper recommends that over the long term, data centres, especially AI-oriented facilities, should be assessed not only through grid adequacy and emissions, but also through macroeconomic, distributional, sovereignty, and environmental perspectives. It also warns that regions with low-cost electricity may attract large-scale computing, but cheap energy alone does not ensure durable economic benefits.
Without deliberate policy design, countries risk using scarce electricity to generate externally captured AI value while absorbing infrastructure costs and labour-market impacts domestically.
Aligning data centre growth with public interest therefore requires integrated governance, treating energy planning, economic development, digital policy, environmental management, spatial planning, and national security as a unified strategic framework.
In systems with high shares of renewables-based generation, AI data centres can have disproportionate effects on voltage and frequency stability due to their rapid demand variability. In extreme cases, this can lead to voltage oscillations, unintended disconnections, and cascading failures.
The paper highlights that regulatory responses to large electricity loads are evolving but remain incomplete. Existing frameworks were designed for predictable demand growth, while emerging loads such as data centres and AI introduce rapid, concentrated, and largely uncertain demand, creating a mismatch between facility expansion and grid readiness.
Environmental and sustainability regulation is also evolving beyond electricity consumption but remains fragmented. Large loads have significant impacts on water use, emissions, and local resources, while limited data transparency constrains effective regulation.
Environmental frameworks and reporting should therefore cover lifecycle impacts (energy, carbon, water, land, materials, and e‑waste footprints). They should also consider the cumulative impacts where multiple facilities are concentrated in the same areas.
The paper further notes that a transition toward conditional connection frameworks is emerging, reflecting the growing system relevance of large loads. For example, Ireland links data centre connections to new generation capacity. However, inconsistent implementation and lack of standardization continue to create regulatory uncertainty.
Cost allocation for grid expansion remains unresolved as well. Large-load growth requires significant infrastructure investment, raising questions over cost sharing and creating risks of cross-subsidization, inefficient investment, or delayed development. Clear frameworks are, therefore, needed to allocate costs and risks fairly among developers, infrastructure providers, and consumers.
Locational and siting policies are expanding but remain only partially effective. Large loads continue to cluster in areas with favourable connectivity and regulation, leading to local grid stress and misalignment between private investment decisions and system needs. Siting decisions should also account for water availability, land constraints, environmental impacts, digital connectivity, and opportunities for local economic development.
Although large loads offer flexibility potential, it remains underutilized. Data centres, for example, could provide demand response and load shifting, but prevailing economic incentives favour continuous operation, limiting system benefits. They could also contribute through energy storage, on-site generation, waste heat recovery, and that way integrate with other energy and industrial assets.
The paper concludes that authorities should respond to the growth of data centres and other large loads through a phased roadmap combining immediate risk management with longer-term planning reform. The objective is to move from reactive power connection decisions to strategic integration, ensuring that large electric loads support grid resilience, decarbonization, economic development, responsible resource use, and public value.