With artificial intelligence promising to transform economies around the world, African nations are looking to build infrastructure for what is coming. China, which has been intimately involved in Africa’s telecom build-out and cultivated relationships with governments across the continent, has an advantage.
Without a plan to compete, the United States risks ceding an enormous emerging market — on the continent with the world’s youngest and fastest-growing population — to its leading geopolitical rival. American AI companies should take note.
China has made waves in the global economy by offering open-weight AI models that developers and businesses can download, modify and fine-tune with their own data.
Software developers across the continent are adopting Chinese models to build applications tailored for African users. Ugandan researchers used Alibaba’s open Qwen model to build their own version, Sunflower LLM, which works in 31 local languages.
An early-stage mobile app called Yotu Health, developed using DeepWisdom’s Atoms software platform, is being tested in Kenya to help diabetes patients log their glucose levels, medications, meals and symptoms, and receive personalized health tips and warnings about when to seek care.
Chinese AI applications are also finding their way into African infrastructure. Ethiopia has adopted China’s AI-powered meteorological early warning system. South Africa is using a Huawei system to help secure its sprawling railway system from thieves and vandals.
But China has no monopoly on AI projects in Africa. Penn State helped develop an app that runs on cheap cellphones, using Google’s open-source TensorFlow software, and lets farmers in West Africa take a photo of a dying crop and get an immediate diagnosis — even without an internet connection. Google’s AI-powered Flood Hub now provides flood forecasts across most of the continent, and its warnings have been used in Nigeria to allow a nonprofit to prepare relief so it could arrive quickly after disaster struck.
Nor is China uncontested when it comes to open models. American chip manufacturer Nvidia is building its Nemotron family of open-weight LLMs and assembling an international coalition to further develop them. A Nemotron-powered solution is already being tested in Gambia to help health workers retrieve medical information.
Running AI requires a lot of computing capacity, which in turn requires a lot of electricity. If African leaders are to make good on their ambitions of achieving data sovereignty in the AI age, they have to figure out how to generate a lot more of their own power.
Sub-Saharan Africa has only about 0.4 gigawatts of data-center capacity. That’s less than 1 percent of the world’s total. Much of the existing capacity there supports telecommunications and conventional cloud services.
In compute, the U.S. has far more to offer than China. Nvidia is the leader in developing the advanced chips needed to run AI, and one of the biggest publicly announced AI-compute build-outs in Africa is Nvidia powered: Cassava Technologies’ planned $700 million deployment across five countries. Microsoft and the United Arab Emirates’ G42 holding company have committed to building a massive data center in Kenya, which will likely also run on Nvidia chips.
Notably, the Microsoft project is currently stalled over, among other issues, Kenya’s nervousness about the project’s power draw. At full build-out, the data center would require as much as 1 gigawatt to run — almost a third of Kenya’s existing capacity. That’s the continent’s challenge in miniature: Only 53 percent of sub-Saharan Africa’s population has access to electricity.
Washington once had a serious answer to this challenge. The Obama administration’s Power Africa program, created in 2013, saw the U.S. Agency for International Development partnering with African governments and the private sector to increase electric power across the continent. But that program came to a halt because of President Donald Trump’s dismantling of USAID and his cuts to foreign aid in general.
AI is not just advanced models solving difficult problems. It’s an infrastructure play. If the U.S. engages more aggressively in electricity generation and computational build-out, American companies will get deeper footholds into emerging African markets.
The post How the U.S. can beat China in Africa’s AI race appeared first on Washington Post.




