Karabo Mokgonyana, a Campaigns and Energy Advisor at Power Shift Africa
Artificial intelligence (AI) is often heralded as a transformative force for Africa’s energy future. From optimising electricity grids to forecasting renewable generation, improving demand management, and expanding access through smart mini-grids, AI offers undeniable promise.
In countries facing chronic inefficiencies in power systems, AI-driven tools can reduce losses, enhance planning, and enable more responsive energy distribution.
Today, China and the United States are leading in the deployment of AI within energy systems. The International Energy Agency (IEA) reports a sixfold increase in AI-related patents for electricity grid applications, underscoring the rapid pace of innovation in this space.
In disaster-prone regions, AI-powered early warning systems are already demonstrating life-saving potential, offering cost-effective alternatives to traditional infrastructure and significantly improving resilience in countries like South Africa and Nepal.
Yet, beneath this optimism lies a far more complex and deeply concerning reality. AI is not just a tool for energy efficiency; it is an energy-intensive industry itself. And for a continent already grappling with profound energy access deficits, the rapid expansion of AI infrastructure risks exacerbating existing inequalities in structural and enduring ways.
At the centre of this tension is the rise of AI data centres - specialised, high-performance computing facilities designed to train and deploy machine learning models. Unlike conventional data centres, these facilities are extraordinarily energy intensive.
AI servers consume significantly more electricity than traditional computing systems, with individual racks requiring upwards of 60 kilowatts compared to just 5–10 kilowatts in standard facilities. This steep energy demand is driven by the computational intensity of AI workloads, particularly large-scale model training and real-time inference.
The implications of this high consumption rate for the global energy situation are staggering. Data centres are projected to consume nearly 945 terawatt-hours (TWh) of electricity by 2030, roughly equivalent to Japan’s total annual electricity consumption.
Already, forecasts suggest that data centres could account for up to 3 percent of global electricity demand by 2026, a figure that is expected to rise sharply as AI adoption accelerates. These numbers translate into immense pressure on power systems, increased emissions, and intensified competition for scarce energy resources.
For Africa, the stakes are particularly high. At 0.617 MWh, the continent has the lowest per capita electricity consumption globally and suffers from chronic capacity constraints, with many countries unable to meet existing power demand.
In this context, the rapid deployment of AI infrastructure, often driven by multinational technology companies, raises critical questions about energy allocation and prioritisation. Who benefits from this energy? And at what cost?
Recent developments in South Africa provide a glimpse into this emerging dynamic. The country is positioning itself as a hub for digital infrastructure, with expanding hyperscale data centres and increasing investment in AI capabilities.
Major global firms, including Microsoft and Amazon Web Services (AWS) have established cloud regions in the country, and new AI-focused data infrastructure is being developed to support growing computational demand.
South Africa currently hosts approximately 50 of Africa’s 150 data centres, making it the most concentrated digital infrastructure market on the continent. This installed base is now scaling fast. Vantage Data Centers has committed about $1 billion to develop a new campus in Midrand, signalling long-term confidence in South Africa as a regional compute hub.
At the same time, Teraco has significantly expanded its footprint in the country, reaching a combined capacity of 189MW across key sites in Isando (70MW), Bredell (64MW), Cape Town (53MW), and Durban (2MW). This scale of deployment places South Africa firmly at the centre of Africa’s digital backbone.
Global hyperscale's are reinforcing this trajectory. Microsoft has committed an additional $340 million to expand AI-ready infrastructure, in addition to its initial $1.1 billion investment in Azure, in Johannesburg.
Meanwhile, Google has launched its first African cloud region in Johannesburg, while Amazon Web Services continues to scale its Cape Town hyperscale region, with plans to invest $6 billion across Africa over five years. These facilities are not merely passive consumers of electricity: they are “AI factories” that require constant, high-density power supply to generate and process vast amounts of data.
To support this AI expansion, energy solutions are being rapidly mobilised, including dedicated renewable energy projects such as solar plants built specifically to power data centres. While this may appear to align with sustainability goals, it raises a fundamental concern: renewable energy capacity that could otherwise be directed toward expanding public access is now being diverted to meet the demands of private, energy-intensive digital infrastructure.
In a country where rolling blackouts (load shedding) remain a persistent challenge, this reallocation of energy resources is both politically and ethically fraught.
Moreover, the energy demands of AI are not limited to electricity alone. Data centres require vast quantities of water for cooling, further straining already scarce resources. Globally, AI infrastructure is projected to consume 33 billion gallons of potable water annually by 2028, with usage expected to increase dramatically in the coming years.
For water-stressed regions across Africa, where more than 400 million people lack access to clean, safe drinking water, this introduces an additional layer of environmental pressure that is often overlooked in discussions of digital transformation.
Equally concerning are the environmental implications. While AI is frequently framed as a tool for climate mitigation, its rapid expansion is contributing to increased carbon emissions, particularly where data centres are powered by fossil fuels.
In some cases, AI facilities have resorted to gas-powered backup systems for reliability, effectively locking in new sources of emissions. This paradox, where a technology designed to optimise systems simultaneously drives new environmental externalities, demands urgent scrutiny.
Beyond environmental and resource considerations, the expansion of AI in Africa also raises profound questions about digital sovereignty and structural inequality. Much of the continent’s AI infrastructure is owned and operated by foreign entities, with profits and data flows largely externalised.
Meanwhile, the energy costs are borne locally, often by already strained national grids. These dynamics risk reproducing historical patterns of extraction, where Africa supplies the raw inputs, in this case, energy and data, while value is captured elsewhere.
The concept of “AI factories” is particularly instructive in this regard. These facilities are designed to produce intelligence at scale, but they are also emblematic of a broader shift toward centralised, capital-intensive digital infrastructure.
For Africa, participation in this system is constrained by limited access to high-performance computing, fragmented policy environments, and inadequate energy infrastructure. As a result, the continent risks being positioned as a peripheral node in the global AI economy, consuming energy without fully capturing the benefits of innovation.
Critically, the expansion of AI infrastructure is occurring within a fragmented and underdeveloped policy landscape. There is currently limited regulation governing the energy use of data centres in most African countries, and few mechanisms to ensure that these developments align with national development priorities.
Without deliberate intervention, there is a real risk that AI will deepen existing inequalities by diverting scarce resources toward elite, export-oriented sectors while leaving most of the population behind.
As global capital flows into hyperscale infrastructure, African institutions are significantly scaling investment in AI. The African Development Bank (AfDB) has announced the “AI 10 billion initiative” aimed at accelerating Africa’s capacity to develop, deploy, and govern AI systems by 2035. The initiative seeks to expand digital infrastructure, strengthen data ecosystems, and create new jobs.
Emphasising the development of regional policy frameworks for AI is not an argument against it. Rather, it is a call for a more critical and context-sensitive approach to its deployment.
Africa must assert greater control over how AI infrastructure is developed, financed, and integrated into its energy systems. This includes establishing clear regulatory frameworks, prioritising public interest outcomes, and ensuring that AI investments contribute to, rather than detract from, universal energy access.
A sustainable pathway forward would require several key interventions. Firstly, AI infrastructure development must be explicitly linked to national energy planning, with clear limits on energy allocation and strong incentives for efficiency.
Secondly, there must be a concerted effort to localise value creation, ensuring that African countries benefit not only as consumers of AI technologies but as producers and innovators. Thirdly, transparency and accountability mechanisms must be strengthened to track the energy and environmental impacts of AI systems.
Ultimately, the relationship between AI and energy in Africa is not merely a technical issue: it is a political, socioeconomic, and ethical question about the continent’s future.
Will AI serve as a tool for inclusive development, or will it become another vector of extraction and inequality? The answer will depend on the choices African leaders make today.
Karabo Mokgonyana is a Campaigns and Energy Advisor at Power Shift Africa
