The United States attempts to force other countries to follow its lead in the AI competition between China and the US. However, in Africa, many companies are not planning to follow the US’s choices. Chinese large models are cheaper and more open, and they even have an advantage in processing local languages in Africa. As a result, they are being adopted by more and more local companies.
According to a report by the South China Morning Post on August 23, Chinese large models such as DeepSeek, Alibaba Tongyi Qianwen, Kimi of Moon's Dark Side, and Zhipu AI are gaining more users in Africa. For many local companies, choosing a Chinese or American model is not a binary political choice, but rather a matter of cost and performance: they use the model that is best suited for their current tasks.
Nigeria's insurance technology platform Curacel is one of them. This company provides AI claims processing and fraud detection services for African insurance companies and fintech enterprises. It has already integrated the GLM-5.3 model from Zhipu AI into its internal system, and uses it in conjunction with Western models.
Curacel co-founder and CEO Henry Mascot stated that, in tasks requiring large model deployments such as programming, data extraction, classification, and customer service, Chinese model costs are lower while the effect can be comparable to Western products.
"Based on our experience, the performance gap between Chinese and Western language models has narrowed significantly." he said. Curacel doesn't bet on a single vendor, but allocates different tasks to different models based on quality and cost. The company is also using cheaper Chinese models in some client projects.
The insurance technology company Curacel, located in Nigeria, utilizes both Chinese and Western AI models, selecting the appropriate model based on task requirements, costs, and performance. Social Media
The iterative speed of the Chinese model has also left an impression on local practitioners.
Michael Michie, co-founder of AI consulting firm EverseTech based in Nairobi, said that about a year ago, it often took a long time for Chinese models to experience a significant performance improvement. Now, significant progress can be seen within six months or even a shorter period of time.
EverseTech has helped some customers switch to Chinese models. One client in the legal industry appreciated the lower cost per million Tokens; another client extensively used retrieval-enhanced generation technology and needed AI to process internal documents within the company. Tests revealed that the Chinese model could convert these documents into searchable data more quickly and cheaper.
For most customers, the problem has shifted from "which model is strongest" to "which model can provide me with the ability at a commercially viable price." Mickey said.
Besides price and performance, the Chinese model has another practical advantage in Africa: local languages.
Uganda has an AI system called "Sunflower" that can provide weather information and agricultural advice in over 30 local languages. The development team chose Quora 3 as the basis for the system, and the reason is straightforward: tests showed that Quora performed better than models from Meta and Google when dealing with Ugandan local languages.
Open weights is also an important reason why the Chinese model has gained attention. Many Chinese models allow companies to download the model weights and deploy them on their own infrastructure, allowing for fine-tuning without relying entirely on external service providers.
Sanifu AI Co-Founder Bernard Momanyi Nyagaka stated that this is particularly important for banks and financial technology companies, as sensitive information such as bank records and customer identity authentication documents can remain within their own systems instead of being sent to third-party API service providers.
He also pointed out that compared to the closed-ended services of Western companies, China's openweight models are in some cases less likely to tie users to a single supplier. This is because companies can store and deploy models on their own. In his view, regardless of where the technology comes from, it is the closed and hosted API services that are more likely to lead to dependence.
Chinese companies are also bringing models and other infrastructure such as cloud computing to Africa, providing users with Chinese models like DeepSeek, Qwen, and GLM, along with tools for development and deployment. In July, China also initiated the establishment of a global artificial intelligence cooperation organization to promote open-source technologies and international AI collaboration.
In contrast, the pressure from the United States to side with certain regions, such as Africa, is becoming more direct. A draft document from the U.S. State Department revealed earlier that countries joining the World Artificial Intelligence Cooperation Organization could be excluded from the U.S.-led “Pax Silica” alliance. The competition between China and the United States has expanded from large models to areas such as chips, critical minerals, cloud infrastructure, and AI governance rules.
Some African enterprises are unwilling to accept this "two choices" system.
Mascott said outright that he does not agree with simply viewing Africa as a “test ground” for Sino-American competition. “African companies are proactive buyers,” he said. They choose technologies that can truly solve problems and are priced reasonably, “regardless of where they are developed.”
Mickey also stated that in the competition between China and the United States in terms of technology, "the most advantageous position for Africa is not to choose one side," but rather to accumulate sufficient capabilities and resources, to cooperate with both parties, and ultimately develop more technologies that belong to Africa itself.
A Chinese Foreign Ministry spokesman, Lin Jian, previously stated at a regular press conference that countries have the right to choose partners based on their own national conditions and development needs. China firmly opposes any attempts to side with certain positions or engage in factional confrontation regarding artificial intelligence. It hopes that all parties will abandon the zero-sum mentality, work together for common benefit, respect each country's digital sovereignty, and jointly build a fair and reasonable global governance system for artificial intelligence. This would enable artificial intelligence to become an important driving force for promoting common prosperity and maintaining common security.
