AI and global structural demand
Electricity is entering a new phase of structural demand growth, driven by electrification, digitalisation, and artificial intelligence, according to Thomas Spencer, senior analyst at the International Energy Agency (IEA).
Thomas Spencer, a senior analyst at the International Energy Agency, asserts that electricity is becoming the defining aspect of global energy infrastructure. He says that modern economies are increasingly “electricity intensive”, particularly in high-value sectors such as financial services, advanced manufacturing, and digital industries.
He explains that this shift is structural and tied to both economic development and decarbonisation. “If we are to decarbonise our energy systems, electrification will be a key route. Electricity demand is not only growing because economies are expanding, but because they are actively switching away from fossil fuels.”
Spencer highlights that this transition is already visible in the data. Global electricity demand grew by 4.5 per cent in 2024 and a further 3 per cent in 2025. While these figures may appear incremental, he emphasises their significance at system level. “The 2025 increase alone represents around 1,000 TWh of additional consumption, roughly equivalent to adding the entire electricity demand of Japan in a single year.”
Between now and 2035, global electricity consumption is expected to increase by around 10,000 TWh. Spencer says this shift is comparable to the combined electricity consumption of all advanced economies today.
Advanced economies
One of the most notable changes, Spencer argues, is the return of electricity demand growth in advanced economies.
“For roughly two decades, consumption in these regions was largely flat, driven by efficiency gains and structural changes in industry, but that pattern is now breaking.”
He explains that electrification is a key driver, particularly through electric vehicles and heating systems. However, he also identifies artificial intelligence and data infrastructure as increasingly important factors.
Spencer states that data centres and AI are expected to account for around 10 per cent of global electricity demand growth to 2030. While significant, he stresses that most demand growth is still coming from broader forces, particularly industrialisation in emerging markets.
In regions such as India and southeast Asia, he explains that electricity demand is being driven by “cooling, appliance growth, and industrialisation”, and in these contexts, AI plays “a relatively minor role”.
Data centres and AI
Spencer points to rapid expansion in global data centre electricity use, which reached around 415 TWh in 2024, comparable to the electricity consumption of France. Despite this scale, he emphasises that it still represents only around 1.5 per cent of global electricity demand.
However, data centre demand has been growing at around 12 per cent annually over the past five years, roughly four times faster than overall electricity consumption. He links this directly to AI, as new AI-optimised servers are significantly more energy-intensive than conventional computing infrastructure.
Regional divergence
Spencer states that the contribution to electricity growth by emerging economies is relatively small. In contrast, in advanced economies they are becoming a key driver of renewed demand growth.
The most pronounced case is the United States, where data centres are projected to account for around half of electricity demand growth out to 2035. He attributes this to the US position as the dominant global data centre hub, hosting around 45 per cent of installed capacity.
The People’s Republic of China follows with approximately 25 per cent, while Europe accounts for around 15 per cent. Spencer outlines that Europe’s share has declined significantly over the past decade, largely due to slower deployment and regulatory and infrastructure constraints. This has prompted policy concern at EU level, with efforts now underway to accelerate data centre development and strengthen digital sovereignty.
System challenges
Spencer identifies data centres as a new category of energy infrastructure that does not fit neatly into existing planning frameworks.
“First, they can be built extremely quickly, often within 18 to 24 months, compared to several years for transmission infrastructure. Second, they are becoming extremely large, reaching gigawatt-scale consumption. Third, they tend to cluster geographically, creating concentrated pressure on local grids.”
He points to rapid growth in project pipelines, outlining a 70 per cent increase in proposed data centre capacity in the first three quarters of 2025 alone. While not all of these projects will proceed, he warns that speculative applications already create significant strain on grid connection processes.
Spencer describes this as a coordination challenge between infrastructure planning and private sector development timelines. In his view, existing systems are struggling to manage the speed and volume of applications.
Flexibility
The response, he argues, will require greater flexibility in how data centres interact with electricity systems.
He suggests that even small levels of operational flexibility could have significant system benefits. For example, allowing data centres to shift consumption for just 1 per cent of hours in a year could, in theory, enable three times more capacity to be integrated in the EU than currently projected.
However, he acknowledges that this is not straightforward as data centres rely on highly expensive IT infrastructure, meaning that reducing utilisation carries real economic costs.
As a result, he outlines a hierarchy of potential solutions. The most challenging to deliver is direct flexibility in computing workloads. More feasible options include flexibility in supporting systems such as cooling, including thermal storage technologies.
He concludes by highlighting the potential for more flexible power arrangements, including hybrid and off-grid solutions, and requirements to contract dispatchable generation.
