UBS: AI Drives Surge in Data Center Electricity Demand
Summaries are public. PDF access requires an active membership; all members have the same access. Sign in
According to the Energy Institute's 2026 Statistical Review of World Energy, global electricity demand from data centers surged by 20% in 2025, a growth rate roughly six times faster than total global electricity consumption. UBS highlights that this acceleration is fundamentally altering macroeconomic resource constraints, transitioning AI from a purely technological investment theme into a critical energy infrastructure challenge.
Geographically, the United States leads this surge with a 26% increase, followed by China at 20% and Europe at 10%. Data centers now account for 2.4% of global electricity generation, with the US bearing a disproportionate 6.5% share. UBS correlates the post-2022 demand spike directly with the rapid scaling of large-scale AI models. Traditional server infrastructure is rapidly being replaced by highly energy-intensive GPU-based architectures necessary for AI training and inference.
While energy efficiency improvements are emerging, they remain insufficient to offset the sheer volume growth in computational queries. Advanced generative AI queries can consume up to ten times the electricity of traditional searches. Although newer text-based query efficiencies are improving, complex reasoning and multimodal tasks continue to demand massive power loads.
The macro implications are profound. Future AI capital expenditure returns will increasingly depend on regional power grid capacities, transmission bottlenecks, and localized energy prices. As data center clusters strain local grids, particularly in concentrated US hubs, competition among tech giants will likely manifest as a competition for energy system carrying capacity. UBS suggests investors must evaluate the AI infrastructure value chain holistically, factoring in power generation, grid connection limitations, and cooling infrastructure constraints.
Key exhibits
Exhibit 1
