The geopolitics of AI: is Africa ready to choose its future?
From Spectator to Player: How Africa Can Win in the AI Era
Artificial intelligence has become the world’s newest arena of strategic competition. Power in this field concentrates across four layers that stack like an industrial pyramid:
Infrastructure at the base, then chips, then models, and finally the applications that reach people and firms. The United States and China currently dominate the first three layers. Europe and the Gulf states are scrambling up the sides. Africa, home to more than a billion people, is mostly watching from the stands.
That view can change, but only if we see the game clearly. Infrastructure is the foundation, chips are the heart, models are the minds, applications are the hands.
Africa is least present in the first three, yet it can gain ground quickly at the top of the stack where solutions are built for real problems. The practical question is not whether artificial intelligence will arrive. It already has. The real question is how African countries can unlock the application opportunity while navigating a global power struggle that is shaping who owns the tools, the rules and the rewards.
Layer one: infrastructure, the bedrock Africa must pour
You cannot run modern AI on a brittle grid or a congested pipe. Electricity, robust networks and local data centers are the minimum entry ticket. This is where Africa faces its steepest gradient. Sub-Saharan Africa’s mobile internet usage sits near 27 per cent, against a global rate of about 57 per cent. Many countries still live with rolling blackouts and limited broadband reach. Without power and fiber, the rest is theory.
There are signs of movement. In 2025 Nvidia announced a 700 million dollar partnership with Cassava Technologies to deploy GPU data centers across the continent. The first phase shipped 3,000 high-end chips to a new South African facility, with a further 12,000 planned for Egypt, Nigeria, Kenya and Morocco. For African engineers who have long worked through high latency links to foreign clouds, that matters. Only about 5 per cent of African AI practitioners currently have the compute needed for advanced work, and many of those rely on thin slices of time on overseas servers. If local cloud capacity grows, developers can train and run models closer to home, reduce costs, and keep sensitive data inside national borders.
Geopolitics threads through these server halls. China has spent the past decade exporting networks, data hubs and surveillance suites as part of its Digital Silk Road. The United States is countering with a push for private investment and what officials call commercial diplomacy. Gulf investors have joined the race, building vast data campuses with tens of thousands of high-end processors and marketing themselves as neutral compute hubs. Infrastructure is now a strategic currency. Countries that host the compute will shape standards and exercise soft power. Countries that do not will rent their digital future by the hour.
For Africa, the immediate path is a pragmatic blend. Expand grids, accelerate last-mile broadband, invite multiple cloud providers to compete, use renewables where they are cheapest, and insist on local jobs and skills transfer in every deal. The Nvidia–Cassava project shows the tone to set. The kit must be first class, not hand-me-down. The contracts should keep data sovereignty and security onshore. The goal is not autarky, it is resilience.
Layer two: chips, the battleground that decides speed and scale
If data centers are the body, semiconductors are the heart and muscle. Advanced AI models devour parallel compute, which today means GPUs and custom accelerators. Control of the chip stack is the high table of tech geopolitics. The United States leads in design through firms such as Nvidia, AMD and Intel. Taiwan’s TSMC fabricates more than 90 per cent of the most advanced chips, which makes the island a linchpin of the global economy. The U.S., Japan and the Netherlands have tightened export controls to slow China’s access to state-of-the-art devices and tools. Beijing has responded with a full-spectrum drive for self sufficiency, from foundries to domestic AI accelerators.
Africa is almost absent here, although the continent supplies the cobalt, lithium, tantalum and platinum that feed global supply chains. Kenya has announced a government partnership with the U.S. to begin small-scale semiconductor manufacturing, and South Africa and Rwanda have floated assembly and packaging ambitions. These are useful seeds, but they will take years to flower and they will not close the performance gap soon. For the next decade the realistic objective is guaranteed access to high performance chips and know-how through smart alliances.
This is where African leverage exists. Minerals and markets are bargaining chips. Regional blocs can negotiate bulk supply of compute credits or long-term GPU allocations in exchange for stable export agreements and investment protections. The Middle East has already demonstrated that capital plus strategic intent can unlock large allocations from U.S. vendors. Africa can do versions of these deals on its own terms, with strict conditions on technology transfer, local integration and workforce training. The principle is simple. In a world defined by silicon, access becomes policy.
Layer three: models, the minds that encode values
Above hardware sits the model layer. These are the large language and vision models that interpret, reason and generate. Training them costs tens of millions of dollars, oceans of data and teams of elite researchers. The leaders remain primarily American and Chinese. OpenAI’s GPT family, Anthropic’s Claude and Google’s Gemini sit on one side. Baidu’s Ernie, Alibaba’s Tongyi and Huawei’s PanGu sit on the other. Meta stands out for open releasing its LLaMA line, and France’s Mistral has become a European standard bearer.
No African lab has trained a model at this frontier scale. That is not a moral failing. It is a realistic accounting of capital and compute. The deeper issue is what rides inside these systems beyond weights and tokens. Models carry defaults about speech, identity, politics and risk. They are shaped by the languages that flood the internet and the cultures that annotate the data. Only 0.02 per cent of online content is in African languages. English alone outweighs all African languages by a factor measured in thousands. The result is that general purpose models often stumble on African names and idioms, misread local context, and sometimes reflect stereotypes.
There are two clear strategies. The first is to work with the best proprietary models, push vendors to support African languages and compliance regimes, and negotiate for regional data retention and safety commitments. The second is to embrace open models that can be inspected and fine tuned on local data, even if they trail the frontier by a year. Both paths are valid. Open models reduce barriers to entry and put control in local hands. Closed models offer peak performance and the strongest safety engineering. Either way, Africa should seek partnerships that build research capability on the continent. More fellowships, more joint labs, more compute grants that tie training runs to local datasets and use cases.
There is a geopolitical split here too. Washington has set an explicit aim to expand global adoption of U.S. models and cloud infrastructure. Beijing has leaned into open releasing efficient models that run on cheaper hardware and bundling them with Digital Silk Road offerings. For African governments, the choice is not a simple fork. The better path is competitive alignment. Work with U.S. model providers for quality, governance and democratic norms. Experiment with open models to localise and reduce costs. Keep a healthy distance from any stack that bakes in opaque controls or networked surveillance.
Layer four: applications, the layer Africa can win now
The top layer is where value touches lives. This is Africa’s best near term bet. You do not need a chip fab to build a loan underwriter that sees thin-file customers more clearly. You do not need a frontier model to build a Swahili voice assistant for community health workers. You need access to APIs, a modest budget for training runs, a team that knows the problem and a channel to reach users.
The fundamentals support a surge. The continent has more than 1.3 billion people and will double by mid century. It counts around 44 million micro, small and mid-size businesses that together employ roughly 80 per cent of the workforce. Early adopters are reporting concrete gains. In South Africa, surveys show widespread use of generative tools at work. In Kenya, recommendation engines and AI logistics are lifting sales and cutting fuel. Nigerian fintech’s are using machine learning to detect fraud and extend credit. Estimates suggest AI could add between 1.2 and 1.5 trillion dollars to African GDP by 2030, which translates to a two to three percentage point lift in annual growth if captured at scale.
Startups are doing a surprising share of the work. About 41 per cent of the roughly 2,400 African organizations working with AI are new ventures rather than big corporates or government programs. Hubs in Kampala, Lagos, Nairobi, Cape Town and Kigali show what local ingenuity looks like when constraints are turned into design briefs. Drones and vision models that spot poachers over vast reserves. Mobile clinics triaging cases with AI assistants. Crop disease detectors that run offline on cheap phones. These ideas can travel to other emerging markets. Some will bounce back to rich countries as cheaper, more robust solutions.
Three enablers determine whether this application wave crests or fades. Affordable access to tools, especially compute and APIs. Skills at every level, from prompt savvy front line workers to data scientists who can ship production systems. Local content, which means language support, domain specific datasets and cultural fluency. The good news is that the U.S. ecosystem is already investing in each. Microsoft and Google have pledged large training programs. American and African partners are building data centres. Meta and Mistral’s open models give developers something they can bend and ship today. The task for policymakers is to turn one-off pilots into a pipeline.
Caught in the crossfire, learning to bargain
It is tempting to frame all of this as a clean choice between Washington and Beijing. Reality is messier. China has built much of Africa’s telecom backbone and offers affordable devices and turnkey cloud services, often financed through state banks. Those packages have frequently included surveillance suites branded as safe city. Critics warn that such systems can drift into tools of control if democratic checks are weak. The United States counters with a language of rights, transparency and safety, but often at a higher sticker price and with conditions attached.
African states are not powerless in this contest. They can trade access for capacity, request secondment programs into top AI labs, insist on in-country data processing for sensitive domains, and require that any surveillance technology meet strict, publicly debated safeguards. They can also play coalition politics. A negotiating bloc of several countries can secure better pricing and longer term commitments than a single buyer can.
Europe, India and the Gulf complicate the picture further. The EU does not own frontier models, but it does write rules, and its AI Act will influence what firms need to ship into African markets. India offers frugal engineering, digital public goods and a template for running national scale platforms at low cost. Gulf investors bring capital and a hunger to become an alternative compute hub. A wise African strategy will engage all three without letting any single external power set the terms of domestic governance.
None of this requires flag waving. It does require a clear-eyed view of trade offs. The U.S. stack is more advanced, benefits from the deepest talent pools, and is embedded in a research culture that publishes, peer reviews and debates openly. Its leading firms are under pressure, from regulators and civil society, to align AI with democratic norms. That is not a guarantee of virtue, but it is a structure of accountability that travels. If Africa wants models that can be interrogated on bias, cloud platforms that meet robust security baselines, and partners who will be pushed by their own systems to respect privacy, then a default alignment with U.S. technology, complemented by open models and diversified infrastructure, is a rational choice.
This alignment should never be uncritical. It should be contractual. Access in exchange for training, grants for local research, dedicated African language roadmaps, and regional data centres that meet African standards. It should also be plural. Keep room for open source and European providers. Continue to work on large scale identity, payments and health data systems. Welcome Gulf capital, but ensure it finances African capacity rather than merely renting African markets.
What readiness looks like, a practical program
Readiness is not a slogan, it is a checklist.
Build the base. Expand electricity access, shore up grids, invest in submarine cables and last mile fiber, and invite multiple cloud providers to bring GPUs onshore. Tie licenses to apprenticeships and knowledge transfer. Use public land and power purchase agreements to accelerate green data centers.
Secure access to chips. Negotiate regional allocations of GPUs and accelerators that do not vanish when global shortages hit. Offer minerals, market access and long term offtake agreements in exchange for predictable supply, local integration partners and training for African engineers.
Invest in people. Put AI and data science into curricula at secondary and tertiary levels. Fund scholarships and research labs. Close the gender gap in digital skills so AI does not widen inequality. Create fellowships that keep top researchers tied to African institutions even when they collaborate globally.
Win at applications. Prioritize use cases that move needles in business, agriculture, health, education, finance and public administration. Buy what works. Kill what does not. Publish results so others can learn.
Open the data and create it responsibly. Release non sensitive public datasets, from agricultural extension records to satellite layers, under clear licenses. Pay communities to contribute voice, text and image data in local languages with strong privacy protections.
Write the rules and enforce them. Update data protection laws. Create clear guidance on AI ethics, nondiscrimination and accountability. Stand up small, capable regulators who can learn fast, use sandboxes, and work with industry rather than against it. Join international forums early so African interests are not an afterthought.
Protect democracy. Build capacity to detect deepfakes and coordinate responses during elections. Regulate surveillance technology with bright lines and independent oversight. Keep the internet open, because an open internet is the soil where local AI ecosystems grow.
A decade that will decide a century
Africa missed earlier industrial revolutions. It does not have to miss the AI revolution. The continent’s greatest assets are its people, its problems and its freedom to leap without legacy. A billion young citizens will become the largest pool of new workers and entrepreneurs on earth. Daily challenges in logistics, health and agriculture are not barriers, they are briefs for innovation. The lack of legacy systems can be an advantage when building national scale platforms that are digital first.
The risk is also clear. If Africa remains a consumer of foreign models, if its data trains distant systems without consent or benefit, if its public spheres are shaped by imported defaults, then AI will deepen dependency rather than reduce it. The difference between those futures will be made by choices that begin now. Choose to pour concrete at the base, to bargain hard for chips and compute, to demand partnerships that teach, and to build applications that pay for the rest.
Artificial intelligence is here to stay and will seep into every system we run. The superpowers will keep writing their chapters, but Africa still holds the pen for its own. If the continent commits to the application layer while investing steadily down the stack, if it negotiates shrewdly with those who control chips and clouds, if it trains millions and writes rules that protect rights, then the next decade can be the one where Africa does not just adopt AI, it adapts it and exports it.
The window is open. Readiness is a choice.