By: Mohamed Bah, Salone Energy Forum

Artificial intelligence is rapidly becoming a major driver of electricity demand, transforming data centres from passive digital infrastructure into major energy consumers with direct implications for power systems.

Unlike traditional data centres, which operate with relatively stable and predictable demand profiles, AI workloads introduce new operational challenges. AI training facilities create highly concentrated electricity demand due to the simultaneous operation of large GPU clusters, while AI inference workloads, although less volatile, create sustained and rapidly growing electricity demand over time.

As inference deployment scales, it is expected to become the dominant source of AI-related electricity consumption, making AI increasingly an energy system issue rather than simply a digital infrastructure issue (IEA, 2025).
The scale of this demand growth is substantial. Global electricity generation to supply data centres is projected to rise from around 460 terawatt-hours in 2024 to over 1,000 terawatt-hours by 2030, with AI accounting for a growing share of this increase (IEA, 2025).

According to the International Energy Agency, electricity demand from AI-optimised data centres alone is expected to more than quadruple by 2030, highlighting the extent to which AI deployment is reshaping electricity demand trajectories (IEA, 2025). This increase is already beginning to challenge infrastructure planning assumptions in both developed and emerging markets.

In Europe, this surge in electricity demand is creating increasing tension between digital growth ambitions and climate objectives. European data centre electricity demand is expected to nearly triple to over 150 terawatt-hours by 2030, requiring significant grid upgrades to accommodate expanding digital infrastructure (Reuters, 2024).

This demand growth comes as European electricity systems are simultaneously electrifying transport, heating, and industry, intensifying pressure on generation and network capacity. In Ireland, data centres already account for around 21 per cent of national electricity consumption, with this share expected to continue rising if current expansion trends persist (The Guardian, 2024). In markets such as Ireland and the Netherlands, where electricity networks are already constrained, additional demand from AI infrastructure risks worsening congestion and delaying new industrial and renewable connections.

The highly concentrated nature of AI-related electricity demand also complicates system planning. AI-focused data centres can consume electricity on a scale comparable to large industrial facilities, but because these loads are geographically concentrated they can create localised network stress, requiring additional reserve margins, transmission reinforcement, and flexible generation capacity (IEA, 2025).

In the short term, this may increase balancing costs and delay grid connections. In the longer term, it may force utilities to retain dispatchable fossil fuel generation for reliability purposes, potentially slowing progress toward decarbonisation targets (Chen et al., 2026). The result is a growing structural tension between AI expansion and clean energy transition objectives.

In Africa, the implications of AI-related electricity demand vary significantly depending on the maturity and reliability of national electricity systems. While AI infrastructure deployment remains limited, the expansion of data centres and digital infrastructure is beginning to expose structural differences in grid readiness across the continent. These differences shape both the risks associated with electricity demand growth and the opportunities for digital infrastructure expansion.

At the most constrained end of the spectrum are low-capacity electricity systems, where the principal barrier to digital infrastructure expansion is insufficient reliable supply. In Sierra Leone, Orange Sierra Leone commissioned its first data centre, a €23 million data centre in 2025 in Bo to strengthen domestic digital resilience and improve service redundancy (MoCTI, 2025). While this represents important progress in national digital infrastructure, the country’s reliable electricity capacity remains limited to roughly 100–150 megawatts, constraining the ability to support larger-scale digital infrastructure growth (U.S.

Department of Commerce, 2026). In such systems, the limiting factor is not digital demand but the absence of sufficient electricity capacity to support advanced infrastructure. Even relatively modest data centre expansion requires parallel investment in generation, transmission, and system reliability.
A second category includes medium-capacity systems with chronic reliability challenges, where digital infrastructure can expand but often depends on expensive and carbon-intensive backup generation.

Nigeria fits this profile. Although the country has a larger electricity system and a growing digital economy, unreliable grid supply has forced many data centre operators to rely heavily on diesel generators to maintain uptime (ResearchGate, 2024). This raises operating costs and increases emissions, reducing the economic and environmental benefits of digital expansion. In such contexts, AI-related demand risks reinforcing fossil fuel dependence unless accompanied by improvements in grid reliability and reserve margins.

At the more advanced end of the spectrum are larger but capacity-constrained systems, where digital infrastructure expansion is possible but increasingly limited by network congestion and balancing constraints. South Africa despite having one of the continent’s most developed electricity systems, has chronic load shedding has forced data centres to secure backup power and private electricity arrangements to guarantee reliability (BusinessTech, 2024). At the same time, renewable wheeling arrangements are creating pathways for large users to contract dedicated renewable generation, allowing digital infrastructure growth to support clean energy investment (Creamer Media, 2024). This creates the possibility for AI-related demand to catalyse investment in renewables and network upgrades, but only where regulation, market structures, and infrastructure planning are sufficiently developed.

These contrasting cases suggest that the challenge posed by AI infrastructure expansion in Africa is not uniform. In lower-capacity systems such as Sierra Leone, the priority is basic electricity system expansion. In reliability-constrained systems such as Nigeria, the challenge is reducing dependence on backup fossil generation. In larger systems such as South Africa, the focus shifts to managing grid congestion while aligning digital demand with renewable investment. This means that the relationship between AI growth and electricity demand in Africa depends fundamentally on national infrastructure readiness.

At the same time, these differences reveal a shared strategic opportunity. Across all three contexts, data centres can provide predictable anchor demand capable of supporting investment in generation, storage, and network upgrades. In lower-capacity systems this could justify basic infrastructure expansion; in reliability-constrained systems it could improve grid resilience; and in more advanced systems it could accelerate renewable integration. If supported by effective regulation and coordinated planning, AI-related electricity demand could become a catalyst for broader electricity sector modernisation. Without such coordination, however, digital infrastructure growth may instead deepen existing reliability challenges, raise system costs, and increase dependence on fossil fuels.

A key determinant of future electricity demand is the rapid expansion of AI inference. While model training requires large amounts of electricity upfront, inference occurs continuously throughout the operational life of a model. As AI adoption expands across sectors, inference is expected to account for the majority of lifecycle electricity use, because demand scales directly with the volume and complexity of user interactions (IEA, 2025a). This means that the long-term electricity impact of AI will depend less on how many models are trained and more on how widely AI services are deployed and consumed. As adoption increases, electricity infrastructure readiness will become increasingly important.

The central strategic challenge for both Europe and Africa is aligning AI infrastructure expansion with electricity system planning. In Europe, the priority is ensuring that AI-related electricity demand does not undermine grid reliability or delay decarbonisation. In Africa, the priority is ensuring that digital infrastructure growth supports investment in reliable and low-carbon electricity systems rather than increasing dependence on expensive and polluting backup generation.

More broadly, electricity infrastructure readiness is likely to become an increasingly important determinant of AI competitiveness. Regions that successfully align AI growth with grid modernisation and clean energy investment will be better positioned to capture the economic benefits of AI while maintaining energy security and climate progress. Those that fail to do so risk higher energy costs, greater system instability, and increased emissions.