Chinese AI dominance in Africa, pressure on Silicon Valley:Kenya developers embrace Chinese open-source AI over costly US models

Last year, while building such an AI system, American and Chinese tools were tested. China’s model proved better because it handled dozens of Ugandan languages more effectively. It was also cheaper and could be customized with its own data. Mwebaze’s ‘Sunflower’ system is now being used across Uganda. Farmers are receiving weather information and agricultural advice in local dialects. In Kenya, entrepreneurs are facilitating legal and business services. In Nigeria, educational tools have been created to teach students. In Ghana, developers are creating local chatbots. Chinese AI is rapidly advancing in developing countries that seek excellent systems at low cost.
Models from American companies like OpenAI and Anthropic are closed and fee-based. In contrast, Chinese systems can be downloaded and modified without any fees or permissions. Approximately half of the total AI usage on Openrouter is from Chinese open-source models. A year ago, this was less than 25%. This has been revealed by the data analysis of eighty million customers. Chinese models are also leading among the most downloaded systems on Hugging Face.
Developers in interviews in Kenya and Uganda have confirmed this. For them, where the model was built does not matter. Cost, computing resources, and data control are most important. American models are like rented houses, while Chinese models are like owning your own home. Training a model from scratch costs millions. In contrast, a new product can be built on top of a downloaded Chinese system. This only requires paying for server costs.
American models are based on subscriptions and fees. Moses Kemibaro, who runs a digital agency, said that if a Toyota hatchback can do the job, why drive a Ferrari? When computing costs are factored in, Chinese models are up to 90% cheaper. American companies are also making efforts Africa’s experience suggests that the global AI competition is now a two-horse race. Kamal Budhabhatti, Chief Executive of Craft Silicon, feels pulled in both directions. He re-engineered his firm’s systems using Anthropic’s programming tool, Claude Code. The work of a hundred engineers was completed by a team of 10 people in half the time.
The bill was heavy, but the speed was worth it. Huawei also offered many major incentives, such as a year of free computing, to adopt Chinese systems.