As data centre construction starts to gather pace in Africa, supporters say investment in AI infrastructure could help expand electricity access for the roughly 600 million Africans who lack it. But critics fear growing energy demand from the sector could strain already weak power systems and divert resources from communities and industries struggling to secure reliable electricity supplies.
On Monday at the UN climate summit in Antalya, three African countries – Somalia, Uganda and Sierra Leone – endorsed a new voluntary pledge proposed by the Turkish presidency which aims “to make AI part of the climate solution – sustainably powered, efficiently designed, transparently measured and available to all”.
The pledge’s signatories – so far just nine countries as well as over 40 companies and some civil society groups – commit to power and build AI infrastructure and hardware sustainably, for example by running data centres with zero- or low-emission energy, supporting their integration into electricity grids, and designing efficient cooling, energy and water systems.
Already have an account?
Log in here →
This story is exclusively available to paid subscribers. A free account gets you a selection of our reporting and our Climate Weekly newsletter. But a subscription gets you every story in full, our Bulletin emails from negotiations and conferences, two subscriber-only newsletters, and our monthly Editor’s column.
Subscribe for £40/quarter →
Or £130/year — best value.
As data centre construction starts to gather pace in Africa, supporters say investment in AI infrastructure could help expand electricity access for the roughly 600 million Africans who lack it. But critics fear growing energy demand from the sector could strain already weak power systems and divert resources from communities and industries struggling to secure reliable electricity supplies.
On Monday at the UN climate summit in Antalya, three African countries – Somalia, Uganda and Sierra Leone – endorsed a new voluntary pledge proposed by the Turkish presidency which aims “to make AI part of the climate solution – sustainably powered, efficiently designed, transparently measured and available to all”.
The pledge’s signatories – so far just nine countries as well as over 40 companies and some civil society groups – commit to power and build AI infrastructure and hardware sustainably, for example by running data centres with zero- or low-emission energy, supporting their integration into electricity grids, and designing efficient cooling, energy and water systems.
The pledge also calls for stronger financial and technical support – including access to computing capacity – to help developing countries use AI to address climate challenges, including improving energy efficiency, optimising electricity systems, reducing emissions and meeting renewable energy needs. It does not, however, explicitly propose that they should be helped to build data centres for that purpose.
UN climate chief Simon Stiell told a high-level event on AI at COP31 that AI’s benefits “must outweigh its growing costs”. “And those benefits must reach every country,” he added.
Bashir Jama, Somalia’s minister of environment and climate change, said his country supports the pledge’s call to reduce AI’s environmental footprint “through cleaner energy, efficient use of electricity and water, transparent measurement and responsible management of electronic waste”.
However, he added that vulnerable countries need to be supported with “accessible finance, affordable computing, reliable data and technical support”, to translate AI’s potential into results.
Project tests data centres running on mini-grids
In Africa, a new project aims to find out whether data-crunching can help rural areas get the power they desperately need. In September, the Rockefeller Foundation launched an “Emergent Grid” initiative that will set up small-scale AI data centres running on solar mini-grids in African villages.
Funded through the Mission300 Accelerator, a World Bank-led programme to bring electricity to 300 million Africans by 2030, pilot projects are planned for Kenya, Democratic Republic of Congo (DRC), Sierra Leone and Djibouti. The idea is to test whether such a set-up could improve revenues for renewable energy developers and expand electricity access in underserved communities.
Mission 300 Accelerator CEO Andrew Herscowitz said global demand for more computing power doesn’t have to push electricity costs higher for people. “We think it can actually do the opposite,” he said in a statement announcing the initiative. “If we build it right, that same demand becomes the reason someone in a remote village finally gets reliable, affordable power,” he added.
Not everyone is convinced, however. Speaking at a Bloomberg Green panel during Climate Week NYC in September, World Bank Group President Ajay Banga warned that the demand for resources needed by data centres, including water and energy, might overburden developing countries.
Explainer: Will AI data centres make or break the energy transition?
Banga said the notion of sharing energy resources with AI would be out of place in those locations and argued there is little prospect of data centres proliferating there.
“You need computing power. Lots of it. You need electricity. Lots of it. More than what the developed world also has in some cases, let alone these developing countries,” he added.
Ketan Joshi, a climate analyst and senior research associate at The Australia Institute, told Climate Home News that data centres should not be seen as a solution for providing electricity access.
“It is demand like any other form of demand – like a steel plant or a household refrigerator – so if you add a data centre, you always need something to generate the energy for it,” he said, warning that without a clear plan to meet that demand, new facilities could divert electricity and investment from other priorities.
Boosting mini-grid economics
Odunayo Onafuye, head of demand partnerships at the Africa Minigrid Developers Association (AMDA), said initiatives like “Emergent Grid” could improve the economics of mini-grids by bringing in large, reliable customers for their power.

Mini-grid developers often struggle to generate enough revenue in underserved communities, where most demand comes from households using electricity mainly for lighting and other basic needs. Their low consumption and limited ability to pay makes it hard for developers to recover the costs of operating and maintaining their plants.
Data centres could help address this problem by providing a steady source of demand and revenue, he said. Developers could sign power purchase agreements with data centre operators, guaranteeing that they will buy a certain amount of electricity.
“It’s improving your consumption and also improving the revenue that will come as a result of that consumption. That improves the financing,” Onafuye said.
Comment: Using energy-hungry AI to detect climate tipping points is a paradox
AI build-out in Africa
While the advantages and disadvantages are still up for debate, Africa is emerging as a potential new frontier for AI and data centre investment, with several technology companies announcing plans to expand their infrastructure across the continent.
Amazon Web Services plans to invest $1.5 billion by 2029 to expand its cloud and AI capabilities across Africa, while California-based Equinix has announced a $438-million investment to tap into the continent’s growing demand for digital infrastructure, including data centres in Nigeria and South Africa.
In South Africa, Microsoft has committed $300 million by 2027 to expand its digital and AI infrastructure. Meanwhile, in Nigeria, energy company Tetracore Energy Group has announced plans to develop a $400-million data centre in Ogun State, in partnership with Chinese firm Huawei.
Globally, the rapid expansion of data centres has sparked growing scrutiny over their environmental and social costs, particularly their demands for electricity, water and land. In the United States, residents have organised protests against proposed and existing facilities in states including Virginia and California.
In June, New York introduced a proposal for a statewide pause of up to a year on new hyperscale data centre developments. Elsewhere, farmers in India and Indigenous communities in Brazil have protested over the impacts of data centre development.
Local concern is rising in Africa too. Citizens and campaigners in Cape Town, South Africa, have pushed back against the expansion of data centres in their communities. With support from the UK-based nonprofit Foxglove, communities filed a lawsuit against two planned data centres in the capital city.
Daniel Hartford, South Africa lead at Foxglove, said the campaigners are not opposed to data centre construction per se, but that “the conversation is just about how transparent these processes are, and really what’s in it for people and the planet”.
So far, communities are not being effectively consulted and there is a lack of information about the energy and water requirements of data centres, and their likely impacts on people’s utility bills, he said.


The climate risks
There are also concerns about the climate impact of AI infrastructure. The Australia Institute’s Joshi said some data centre developers are turning to on-site gas power generation while waiting for grid connections, as they race to bring facilities online. This could lock in fossil fuel infrastructure and increase emissions, he warned.
The African Energy Chamber, which lobbies for fossil fuels, has proposed AI as a way to create new demand for gas in Africa, arguing that “data centres require 24/7 baseload power with near-zero tolerance for interruption, making intermittent renewables alone insufficient without firm backup generation”.
The vote that stopped a data centre: US communities query resource-hungry AI
Many climate advocates – and increasingly politicians – say that additional energy demand for AI data centres should be met from new renewable power capacity – something Rockefeller Foundation’s Emergent Grid initiative hopes to trial in Africa.
The foundation says five sites in Kenya will test whether flexible computing can serve as an anchor customer for new solar-plus-storage systems serving communities of around 50,000 people. In DRC, computing will be added to an existing under-used solar plant that currently serves more than 30,000 people, as well as more than 350 small and medium-sized businesses and 30–40 social institutions.
Other demonstrations will test the model on existing mini-grids in Sierra Leone and through an integrated energy, compute and connectivity site in Djibouti.
But Joshi argued that rather than redirecting surplus mini-grid power in Africa to AI data centres, renewable energy should be used for economic activities more essential for local development such as agriculture and manufacturing, or stored in batteries for use when electricity demand is higher.
“There are always other alternatives. You have to think about all the other potential uses,” he said.


