African Solarpunk

Holding 60% of the entire planet’s best solar resources, and receiving the most usable sunlight on Earth, you would think Africa harvests the most solar energy globally. You would be wrong. When the International Energy Agency last counted, Africa hosted 1% of the world’s installed solar panels.¹

The age old curse of African Potential strikes again.

Or perhaps an old story, in a way. Drive 50 kilometres northwest of Johannesburg and you will find the Cradle of Humankind, a stretch of limestone caves that has yielded roughly 40% of all known fossils of human ancestors.² In one of those caves, Swartkrans, researchers found burnt bones pointing to some of the earliest controlled use of fire, more than a million years ago.³ Energy was among humanity’s first great technologies, and some of the oldest evidence for it comes from African soil.

Today the world is lighting a new kind of fire. Artificial intelligence runs on electricity, metals, water, data and human language. Africa has an immense supply of every one of them. And yet, surprising to nobody, the continent sits at the very far end of what we call The Current.

Simplifying, quickly. Cobalt and copper leave the Central African Copperbelt as ore. They are refined, built into batteries, chips and cables, and installed in data centres on other continents. Models are trained there, on data that includes African voices and, as we will see, African labour. The finished product then travels back down the current as an app or a subscription, priced in dollars. Africa supplies the headwaters and buys at the mouth. The middle, where most of the value is added, happens elsewhere.

Of all the world’s regions, Africa stands to gain the most from AI, because it holds the inputs the technology needs most and the people who will use it the longest. Capturing that gain means bringing the middle of the current home: generating the power, running the compute, keeping the data and speaking the languages on the continent itself. In business terms, it is vertical integration at the scale of the second largest continent on Earth.

That is what we mean by African Solarpunk. Solarpunk imagines a future where clean energy, advanced technology and living ecosystems work together, and where the benefits are shared widely and fairly. Africa is, to no Africans’ surprise, well placed to build it.

The Illusion of Progress

Naively, you conclude that Africa is already riding the AI wave. Startups in Lagos and Nairobi build on the newest models. Students in Kigali and Cairo study with Claude. Accountants, software engineers, and attorneys in Sandton and Capetown prepare their documents using AI meanwhile pretending not to. All of that is real, and worth celebrating.

Now anti-naïveté…

In June 2026, about 18.8% of the world’s working-age population used a generative AI tool. In the wealthier Global North the figure was 28.8%. In the Global South it was 16.2%, and the gap has widened every quarter this year.⁴ In the first quarter of 2026, South Africa led the continent at 23.1%. Egypt stood at 14.8%, Nigeria and Ghana at about 10%, Kenya at 8.7%, and much of sub-Saharan Africa below that.⁵ The world leader, the United Arab Emirates, has passed 73%.

Let’s be frank: in most African countries, around nine in ten working-age adults did not use a generative AI tool at all in early 2026.

The reason is obvious. To get AI you need basic things, and in much of Africa you do not have these basic things.

• Electricity. Around 600 million Africans, nearly two in five, live without it.⁶
• Connectivity. About 36% of people in the ITU’s Africa region use the internet, against roughly 74% worldwide. Women are further behind than men.⁷
• Compute. Africa is home to nearly a fifth of humanity but about 0.6% of global data centre capacity: roughly 360 megawatts of active capacity, against some 55 gigawatts worldwide.⁸

That last number deserves some coloring in. In March 2026 the Mozal aluminium smelter outside Maputo, which drew a constant 950 megawatts, went into care and maintenance because its owners could not secure affordable electricity.⁹ That one smelter’s taste for power was more than two and a half times the active capacity of every data centre on the continent combined. Africa has industry and ambition. What it lacks is cheap, reliable power where the work happens.

The current shows up elsewhere too. The Democratic Republic of Congo mined roughly 72% of the world’s cobalt in 2025,¹⁰ a key battery metal for the phones, laptops and backup systems of the digital economy. Little of the refining, manufacturing or modelling that follows happens nearby. When OpenAI needed people to label disturbing text so ChatGPT could learn to filter it, the work went to Nairobi through an outsourcing firm. TIME reported that the labellers took home between $1.32 and $2 an hour, while the contract paid the firm $12.50 an hour.¹¹ The work was essential, dehumanizing, and the share of its value that stayed in Kenya was embarrassingly small.

Even moving money follows the current. In 2024 only about 14% of Africa’s trade was with itself, and sending money to sub-Saharan Africa costs more than sending it anywhere else: about 8.45% in fees on average, or roughly $17 on every $200 sent home.¹²

Nobody shaped the meanders on purpose. Each curve, came one reasonable-looking deal at a time. The result, though, is clear. Using AI and owning the value of AI are two very different places to swim, and today most of the African continent is drowning in the first.

The Continent’s Balance Sheet

If that was the liabilities side, here are the assets. Read them together and a pattern appears: nearly everything AI needs, Africa holds in abundance.

Two rows demand your attention more than the rest.

People. Africa’s median age is about 19. The next-youngest region, Latin America and the Caribbean, sits at 32.¹⁶ By 2050, roughly one in four people on Earth will be African. AI is a general-purpose technology whose rewards compound over decades of use, and the continent with the most decades ahead of its population is positioned to collect the most. That holds only if the foundations are in place.

Sunlight, because it connects every other row. Minerals become more valuable when there is power to process them near the mine. Farmland becomes more productive when there is power to pump water, run cold rooms and dry grain. Data becomes an asset when there is power to store and compute it locally. The sun is the one input every other asset depends on, and it arrives free every morning, unless you’re in Cape Town during the winter.

That is why the latest solar figures are so striking. Ember, the energy think tank, found that Chinese solar panel exports to Africa reached 23 gigawatts in the twelve months to June 2026. In 2023, South Africa took more than half of the continent’s imports. By mid-2026 its share had fallen to a fifth, because everyone else was buying too.¹³ Households and businesses tired of load-shedding and diesel bills are already building the energy layer of a solarpunk continent, one rooftop and parking lot at a time.

What “Solar-Powered Inference” Actually Looks Like

For every complex problem there is an answer that is clear, simple and wrong.

H. L. Mencken

An AI model has two phases in its life. Training is when the model learns. It takes months, thousands of specialised chips (currently, mainly from NVIDIA) and power on the scale of a small city. Inference is when the model is used: every question answered, every You met me at a very Chinese time in my life, every sentence translated. Training happens once per model. Inference happens billions of times a day, for as long as the model is in use.

Africa does not need to win the training race to benefit from AI. Open-weight models, which anyone can download, run and adapt, have become remarkably capable. They come from many places: Chinese labs such as DeepSeek and Alibaba’s Qwen team, and American and European companies such as Meta, Google and Mistral. Microsoft’s data found that DeepSeek, free to use and openly licensed, saw two to four times more use in Africa than in other regions.¹⁹ Africans are already choosing the tools they can afford and adapt. The next step is to run those tools on African soil, on African power.

Being visionary in name helps nobody, so picture this.

A farming cooperative in Limpopo. Beside the packhouse sits a steel shipping container holding racks of GPUs, a battery bank and a liquid-cooling loop. Solar panels cover the roof and the field next door.

By day, the container does the cooperative’s thinking. A camera on a tractor spots early blight on a row of tomatoes. A sensor network decides which fields need water, and when. A farmer asks a question aloud in Sepedi and hears the answer in Sepedi. At solar noon, when power is most plentiful, the container runs its heaviest jobs: retraining the disease model on this season’s photos, forecasting next week’s prices at the fresh-produce market.

The heat goes to work as well. The cooling loop carries it to a crop dryer, so maize and chillies dry evenly and store longer. Some of the power runs a cold room. That matters enormously, because an estimated 30 to 50% of food produced in sub-Saharan Africa is lost after harvest.²⁰ In Nigeria, ColdHubs has shown that a solar cold room can stretch the shelf life of fresh produce from about two days to about twenty-one.²¹

And the data, the photos and soil readings and voice questions, stays on the farm unless the farmers choose to share it.

This is the scale at which inference fits Africa. An IFC executive made the case in March: by focusing on inference, African markets can run real-time services such as payments, local-language tools and digital government without the huge power loads of a training hub.²² Larger facilities have their place too. Cassava Technologies launched an Nvidia-powered “AI factory” in South Africa this year, starting in Cape Town, with a second planned for Johannesburg and an aim of 12,000 to 13,000 GPUs across the continent.²³ The healthiest picture is an ecosystem: large AI factories in the cities and thousands of small pods at the edge, much as the solar boom already mixes utility plants with rooftops.

The voice layer

The most important application in this picture may also be the simplest: speaking.

Sub-Saharan Africa is home to more than 2,000 languages, and fewer than 5% of them have the digital resources language AI needs.¹⁸ For hundreds of millions of people, a chatbot that only reads and writes English or French is a door into the wrong room. Speech recognition in their own languages opens the right one.

The work is well underway. WAXAL, an open dataset funded by Google and the Gates Foundation and built with Makerere University, the University of Ghana, Digital Umuganda in Rwanda, AIMS Senegal and others, holds about 1,250 hours of transcribed natural speech in 19 African languages spoken by more than 100 million people.²⁴ In 2025 Nigeria released N-ATLAS, an open model for Yoruba, Hausa, Igbo and Nigerian-accented English.¹⁸ And when WAXAL’s Wolof data first shipped with misaligned transcripts, a Senegalese community of AI developers, GalsenAI, found and fixed the problem.²⁵ That is the solarpunk pattern in miniature: open tools, improved locally, owned by the people who use them.

The wild layer

The sensors and models that watch crops can also watch wildlife corridors. EarthRanger, an open-source conservation platform, has been deployed at more than 500 sites in 70 countries, and 72% of those sites are in Africa. In Malawi’s Liwonde National Park, better monitoring and response supported by EarthRanger cut deaths from human–wildlife conflict by more than 91%. Across the parks managed by African Parks, rangers using it have removed more than 50,000 snares.²⁶

Africa’s wild places are part of its wealth and part of the planet’s life support. The Kavango–Zambezi Transfrontier Conservation Area spans five countries and about 520,000 square kilometres, and it is home to roughly 228,000 elephants, about half of Africa’s remaining savanna elephants.²⁷ The Congo Basin, the world’s second-largest tropical rainforest, sits on peatlands holding around 30 billion tonnes of carbon, and recent research warns that the forest’s ability to absorb carbon is weakening.²⁸ Solar-powered edge AI can count herds from the air, pick out the sound of chainsaws in acoustic recordings, and warn a village before elephants reach its fields. In African solarpunk, compute serves the savanna as faithfully as it serves the farm.

How a Young Continent Can Organise Its Future

Power, compute and language are the hardware and software of an African AI economy. Its operating system is institutions: how a continent decides about its money, its data, its laws and its borders.

This is where AI promises the most and where the loudest objections will stem from. We recommend, in public decisions, AI should advise and accountable people should decide. Models can surface patterns faster than any ministry. It cannot be voted out, it cannot be held responsible, and it is incapable of cadre corruption (phenomena prevalent in the United States, Europe, Africa, China, Australia, and South America.)

Money. Central banks set interest rates using data that is often weeks or months old. Machine learning can “nowcast” an economy from faster signals such as mobile-money flows, market food prices, shipping data and satellite images. The South African Reserve Bank has published research on nowcasting GDP with a suite of statistical models,²⁹ and in May 2025 central banks including the BCEAO, the BEAC and those of Kenya, Morocco, Nigeria and South Africa met in Dakar to present AI prototypes for forecasting inflation and supervising banks.³⁰ Add PAPSS, the Pan-African Payment and Settlement System, which lets a trader in Accra pay a supplier in Lagos in their own currencies without a detour through the dollar. By late 2025 it connected 19 countries and more than 150 commercial banks.³¹ Faster data, local settlement and transparent models add up to monetary policy made on the continent, for the continent, with evidence anyone can inspect.

Data. Data is the raw material of AI, and much of Africa’s is stored and processed abroad under other countries’ laws. The African Union’s Malabo Convention on cyber security and data protection took nine years to gather the 15 ratifications it needed to enter into force in 2023, and the AU’s Continental AI Strategy, endorsed in 2024, puts data governance at its centre.³² More than 40 African countries now have national data protection laws.⁸ Sovereignty here should mean something precise: data about Africans is stored on the continent by default, protected by enforceable rights, and shared on terms set by the people it describes. Solar-powered local data centres are what make that default affordable.

Laws. A law people cannot read is a law they cannot shape. Language and speech models can translate every bill into the languages people actually speak, read it aloud over a basic phone line, and help a parliament digest thousands of public comments in days. They can flag where a new regulation contradicts an old one. Elected representatives still debate, amend and vote. AI simply widens the doorway into the process.

Borders. The African Continental Free Trade Area now counts 50 ratifications among its 54 signatories, covering about 1.4 billion people and a combined GDP of around $3.4 trillion.³³ It only works if goods cross borders quickly and fairly. AI can speed up customs clearance, predict truck queues at border posts and catch document fraud without delaying honest traders. Borders shape wildlife too. In KAZA, researchers tracking GPS-collared elephants found that fences between Namibia and Botswana stopped female elephants completely: no collared female moved between the two countries.³⁴ Good data can help design borders that keep people safe while letting trade, and elephants, move.

Here the continent’s greatest asset returns. The engineers who will build these systems, the technicians who will run them, the civil servants who will govern them and the citizens who will hold them to account are overwhelmingly young. For them, organising for this future means study, cooperatives, startups, public service and active participation in the institutions that already exist.

The Counter-Argument: Overcoming the “Impossible”

A careful reader should be sceptical by now. Here are the strongest objections, answered as honestly as we can.

”The grid is too unreliable for data centres.” For a conventional hyperscale campus, that is often true. It is exactly why this model starts with solar panels and batteries on site. Africa’s households and businesses have already voted with their wallets: panel imports keep setting records, driven largely by people escaping load-shedding and diesel bills.¹³ Mozal is the same lesson seen from the other side. Heavy industry leaves where power is expensive and gathers where it is cheap.

”Capital costs too much.” This is the hardest objection. It is also the hardest to refute outright. The IEA estimates that the cost of capital for clean energy projects in Africa is at least two to three times higher than in non-African economies and China,³⁵ and the continent attracts under 3% of global energy spending while holding around a fifth of the world’s people.³⁶ This is incredibly disappointing, but the remedies are known – though none are quick: concessional finance that absorbs early risk, local pension funds investing in local infrastructure, cooperative ownership that spreads small stakes across many people, and a continental market big enough to reward patient investors. Small modular projects help, because a single container can be financed and proven long before a campus can.

”Africa doesn’t make the chips.” Obviously, and leading-edge chip fabrication will stay concentrated in a few places (Taiwan) for years. Africa can still capture a large share of the value, because chips are a purchased input, like tractors on a farm. The value lies in what they produce: services, data, models tuned to local needs, and the energy that runs them. Older chips from the world’s upgrade cycles can find second lives in edge pods, school labs and community networks.

”AI will take the jobs.” Some tasks will change. A solar-powered AI economy, though, is full of hands-on work: installing panels, building grids, cooling and maintaining servers, recording and cleaning language data, repairing field sensors, advising farmers. Africa’s developer community is already growing faster than any other.¹⁷ The real risk is that these jobs are created somewhere else.

”This will become a tool for surveillance.” It could, anywhere in the world. It already is in Europe and The United States. Africa should target open models wherever possible, public records of how government uses AI, independent data protection authorities with real power, and a person who signs every consequential decision. Solarpunk is community-first by definition. A system that watches citizens without their consent fails the culture test. We believe the value of honest disclosure, especially when it comes to user data, is heavily undervalued in 2026. In-so-far-as a technology that gets better with more data, should respect and reward the human beings it relies on for its own value. You get further together.

A Call to the New African Engineer

When we say engineer, we mean it broadly.

The engineer is the technician who keeps a container of GPUs cool through a Sahel summer. The agronomist who turns ten thousand leaf photos into a disease model. The linguist who records her grandmother’s stories in Kinyarwanda so a speech model can learn the language. The central bank economist who builds a nowcast from mobile-money data. The ranger reading an alert on a tablet at midnight. The developer in London or Toronto who decides to host her next project on an African cloud.

Each of them is building the middle of the current.

The community already exists. The Deep Learning Indaba has gathered Africa’s machine learning researchers every year since 2017, and in 2025 its local IndabaX events reached 47 countries.³⁷ Open research communities keep building datasets and models for African languages. Solar installers work in every major city. What is often missing is the connection between them: the installer who knows the ML engineer who knows the farming cooperative.

With the express goal of spreading awareness of the growing global Solarpunk movement, Snow asks the following of our African readers, and any non-African-identifying readers who align closely with African Solarpunk.

To builders: run open models locally. Contribute to open African language datasets. Design for the sun: schedule heavy work for midday, share your heat, share your power.

To leaders: treat solar-powered, locally hosted compute as public infrastructure. Ratify and enforce data protection. Connect to regional payment systems. Buy from local AI providers whenever they can do the job.

To investors and to the wider world: Africa’s success is a shared interest. The Congo Basin helps regulate the planet’s climate. Africa’s farmland will help feed the world. One in four people alive in 2050 will be African. A prosperous, solar-powered, AI-literate Africa is good news for every continent.

More than a million years ago, in a cave not far from Johannesburg, one of our ancestors kept a fire burning. It was among the first times anyone turned energy into a tool, and everything that followed grew from it.

The next fire is already burning, 150 million kilometres away. It rises over this continent every morning, more generously than anywhere else on Earth. This time, its warmth can be shared. The question for this generation is what it will build in that light.

Sources and notes

Figures were checked against the sources below in September 2026.

Notes

1. International Energy Agency, Africa Energy Outlook 2022, key findings. https://www.iea.org/reports/africa-energy-outlook-2022/key-findings ↩
2. University of Cape Town News, “Rocking the Cradle” (2023), on the Cradle’s share of known hominin fossils. The Fossil Hominid Sites of South Africa were inscribed as a UNESCO World Heritage Site in 1999. https://www.news.uct.ac.za/article/-2023-05-12-rocking-the-cradle-new-evidence-shows-site-is-much-younger-than-thought ↩
3. C. K. Brain and A. Sillen, “Evidence from the Swartkrans cave for the earliest use of fire,” Nature 336 (1988). The dating and interpretation of early fire evidence continue to be debated. ↩
4. Microsoft On the Issues, “The continued state of global AI diffusion in 2026” (September 2026). Microsoft measures the share of people aged 15 to 64 who used a generative AI product during the period. https://blogs.microsoft.com/on-the-issues/2026/09/21/the-continued-state-of-global-ai-diffusion-in-2026/ ↩
5. Microsoft AI Economy Institute, Global AI Diffusion Q1 2026 Trends and Insights (May 2026). https://www.microsoft.com/en-us/research/wp-content/uploads/2026/05/Microsoft-AI-Diffusion-Report-2026-Q1.pdf ↩
6. IEA, Financing Electricity Access in Africa (2025), executive summary. https://www.iea.org/reports/financing-electricity-access-in-africa/executive-summary ↩
7. ITU, Measuring Digital Development: Facts and Figures 2025. The ITU’s “Africa” region covers mainly sub-Saharan countries; North African states are counted in its Arab States region. https://www.itu.int/itu-d/reports/statistics/facts-figures-2025/ ↩
8. Africa Data Centres Association with Rising Advisory, Data Centres in Africa 2026: The Economic Report. https://africadca.org/en/data-centres-in-africa-2026-the-economic-report ↩
9. South32, Mozal Aluminium. https://www.south32.net/what-we-do/our-locations/southern-africa/mozal-aluminium ↩
10. GlobalData analysis via Mining Technology (January 2026). https://www.mining-technology.com/analyst-comment/drc-indonesia-anchor-global-cobalt-supply/ ↩
11. Billy Perrigo, “OpenAI Used Kenyan Workers on Less Than $2 Per Hour,” TIME (January 2023). https://time.com/6247678/openai-chatgpt-kenya-workers/ ↩
12. TechCabal (September 2026), citing 2024 intra-African trade data and World Bank remittance prices for Q1 2025. The World Bank’s cost measure is for sending $200. https://techcabal.com/2026/09/10/papss-payment-network-africans-use/ ↩
13. Ember, “The take-off in African solar that official statistics can’t yet see” (2026). https://ember-energy.org/latest-insights/the-take-off-in-african-solar-that-official-statistics-cant-yet-see/ ↩
14. African Development Bank, Feed Africa strategy. Estimates of 60 to 65% are widely cited. https://www.afdb.org/fileadmin/uploads/afdb/Documents/Generic-Documents/Brochure_Feed_Africa_-En.pdf ↩
15. Tanger Med Port Authority, Port Activity Report in 2025 (February 2026). https://www.tangermed.ma/wp-content/uploads/press-releases/2026/CP-TMPA-PORT-ACTIVITY-REPORT-IN-2025.pdf ↩
16. Pew Research Center, analysis of UN World Population Prospects 2024 (July 2025), for median ages; UN World Population Prospects 2024 for projections to 2050. https://www.pewresearch.org/short-reads/2025/07/09/5-facts-about-how-the-worlds-population-is-expected-to-change-by-2100/ and https://population.un.org/wpp/ ↩
17. Microsoft AI Economy Institute, Global AI Diffusion Q1 2026, as reported by Engineering News (May 2026). The 21% figure is annual growth from 2019 to 2024. https://www.engineeringnews.co.za/article/global-ai-use-increases-in-first-quarter-microsoft-2026-05-18 ↩
18. TechCabal (February 2026), on WAXAL, N-ATLAS and language resources in sub-Saharan Africa. https://techcabal.com/2026/02/02/google-joins-push-to-localise-ai-for-african-languages-with-speech-database/ ↩
19. Microsoft AI Economy Institute, Global AI Adoption in 2025: A Widening Digital Divide (January 2026). https://www.microsoft.com/en-us/corporate-responsibility/topics/ai-economy-institute/reports/global-ai-adoption-2025/ ↩
20. “From farm to fork: a review of strategies for sustainable reduction of post-harvest losses in Sub-Saharan Africa,” Cogent Food & Agriculture (2025). Estimates vary widely by crop and method. https://www.tandfonline.com/doi/full/10.1080/23311932.2025.2588851 ↩
21. International Institute of Refrigeration, “Reducing post-harvest food losses in sub-Saharan Africa” (2023). https://iifiir.org/en/news/reducing-post-harvest-food-losses-in-sub-saharan-africa ↩
22. The Guardian (Nigeria), “Africa’s digital capacity lags as 1.4b people share 1% compute power” (March 2026). https://guardian.ng/news/africas-digital-capacity-lags-as-1-4b-people-shares-1-compute-power/ ↩
23. ITWeb, “Cassava plans AI factory in Joburg, as Cape Town goes live” (May 2026). https://www.itweb.co.za/article/cassava-plans-ai-factory-in-joburg-as-cape-town-goes-live/6GxRKMYQbWnMb3Wj ↩
24. WAXAL dataset card, Hugging Face (2026). https://huggingface.co/datasets/google/WaxalNLP ↩
25. “Opportunities and Challenges of Natural Language Processing for Low-Resource Senegalese Languages in Social Science Research,” arXiv:2601.09716 (2026). https://arxiv.org/abs/2601.09716 ↩
26. Wall et al., “EarthRanger: An open-source platform for ecosystem monitoring, research and management,” Methods in Ecology and Evolution (2024). https://besjournals.onlinelibrary.wiley.com/doi/10.1111/2041-210X.14399 ↩
27. Naidoo et al., “Landscape connectivity for African elephants in the world’s largest transfrontier conservation area,” Journal of Applied Ecology (2024). https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2664.14746 ↩
28. Global Witness, on the Congo Basin peatlands; National Geographic (March 2026), on signs that the forest’s carbon uptake is declining. https://globalwitness.org/en/campaigns/forests/why-is-the-congo-basin-the-worlds-largest-forest-carbon-sink-at-risk/ and https://www.nationalgeographic.com/environment/article/congo-basin-carbon-research-climate-change ↩
29. Botha, Olds, Reid, Steenkamp and van Jaarsveld, “Nowcasting South African gross domestic product using a suite of statistical models,” South African Reserve Bank Working Paper WP/21/01 (2021). ↩
30. African Diplomats, on the May 2025 BCEAO-hosted meeting of central banks on AI (May 2025). https://africandiplomats.com/artificial-intelligence-african-central-banks-take-the-helm-of-a-new-financial-order/ ↩
31. All Business Africa, “PAPSS and the Rise of Pan-African Payment Corridors” (April 2026). https://allbusiness.africa/insights/papss-cross-border-payments-africa ↩
32. Future of Privacy Forum, “The African Union’s Continental AI Strategy” (November 2024). https://fpf.org/blog/global/the-african-unions-continental-ai-strategy-data-protection-and-governance-laws-set-to-play-a-key-role-in-ai-regulation/ ↩
33. tralac, AfCFTA ratification status (September 2026); market size as stated by the African Union. https://www.tralac.org/resources/by-region/cfta.html ↩
34. “Challenges to Elephant Connectivity From Border Fences in the World’s Largest Transfrontier Conservation Area,” Frontiers in Conservation Science (2022). https://www.frontiersin.org/articles/10.3389/fcosc.2022.788133/full ↩
35. IEA, Financing Clean Energy in Africa (2023). https://www.iea.org/reports/financing-clean-energy-in-africa ↩
36. IEA, Clean Energy Investment for Development in Africa, executive summary. https://www.iea.org/reports/clean-energy-investment-for-development-in-africa/executive-summary ↩
37. Deep Learning Indaba, IndabaX programme. https://deeplearningindaba.com/2026/indabax/ ↩