Ernest Mwebaze built Sunflower LLM, a model handling 31 Ugandan languages, on Alibaba’s Qwen 3. Across Africa, developers are choosing Chinese platforms like DeepSeek, Qwen, and Kimi over American alternatives. The reason is simple: cost and flexibility. “If models are only available in certain western languages, you’re excluding a lot of people from this technology revolution,” Mwebaze told Foreign Policy.
Shikoh Gitau, CEO of Qhala and a leading African AI researcher, said Chinese models are faster and cheaper to train. They are also open-source. That combination is hard to beat. “Africa is going to win AI on minimum viable intelligence,” she said.
Small models serving specific needs, and the best platforms to train them on are currently Chinese. The cost gap is stark. Kimi, from Beijing-based Moonshot AI, costs around $3.40 per million output tokens.
Anthropic’s Opus 4.7 and OpenAI’s GPT-5.5 cost $25 and $30 respectively. For African developers, the disparity can be even worse. Gitau said Qhala’s research found training a model on an African language can cost three to 30 times more than English.
She called it “tokenization bias.” “Our languages are hard, they are not documented, they are not digitized, they are not in a format that anybody else is willing to understand them,” she told Foreign Policy. That bias hits home. UNESCO estimates between 1,500 and 3,000 languages are spoken across the continent.
Some, like Hausa and Swahili, have tens of millions of speakers. Others, like Kakwa, have a few hundred thousand. Uganda alone has 41 spoken languages.
Training a large language model requires vast data. For English and French, that exists. For most African languages, it does not.
Many were not written down before colonization and lack the digital footprint needed for standard LLMs. The solution, Gitau said, is small and specialized language models. These SLMs and SSLMs can be built on minimal data sets and focus on specific applications like agriculture, health, or education.
Chinese platforms currently offer the best tools for that. Mwebaze’s Sunflower LLM is one example. Another is Skillbridge, launched in 2024 by A2SV, an African AI incubator.
Working in Amharic and Afaan Oromo, it helps Ethiopian students study for university entrance exams. It recently expanded to Rwanda. China is not leaving its position to chance.
In April, the Chinese government launched an AI competition for young African developers. Winners will go to China on study visits. When these top developers return home, they will have spent six months learning how to use Chinese models. “My biggest worry is that we are locked in an ecosystem that doesn’t have policies that you can extract yourself from,” Gitau said.
That dependency echoes past patterns. Gitau pointed to smartphones. Transsion, a Chinese manufacturer, now produces 44 percent of the continent’s smartphones.
It did so by making affordable phones marketed solely to Africa. “Everybody now has a Chinese phone, and I feel like that’s what’s happening with AI,” she said. “By the time [Western companies] realize there’s a market, there will be no market.”
AI is often described as a leapfrog opportunity for Africa. Mobile money bypassed traditional banks. Mobile phones meant no need for landlines.
AI could do the same for health care, education, and agriculture. Rwanda is the only country so far to sign a deal to integrate AI into its government, but wider adoption is coming. Just as countries found themselves reliant on Chinese infrastructure through the Belt and Road Initiative, they may find their AI developments similarly tied.
African developers reject the idea of a U.S.-China AI war. “There’s a big war between the U.S. and China on AI. We are not going to be part of that war,” Mwebaze said. For them, it is a choice between technologies.
Up until now, Qwen, DeepSeek, and Kimi have worked best. “Pitching China against Africa is not going to work,” Gitau said. The race, as Western companies understand it, has barely begun. Africa remains behind on AI adoption.
South Africa has the highest rate, with 23 percent of people saying they have used generative AI. The Democratic Republic of the Congo is at 9 percent. Rwanda, despite its government deal with Anthropic, is only at 7 percent.
A Microsoft report found Asian countries have seen rapid growth in AI usage due to increased support for local languages. Since June 2025, South Korea, Thailand, Japan, Mongolia, and Laos have all seen more than 30 percent growth. Mwebaze’s latest work shows the landscape can shift.
He turned to Google’s Gemma for a new project. Google’s smaller models require less compute and can be embedded in phones. They can also be trained for speech and text, crucial in areas with lower literacy rates.
The cost of training a Gemma model is comparable to Qwen. Mwebaze is, however, among the few working on Gemma. Why It Matters: The choice of AI infrastructure today will shape Africa’s digital sovereignty for decades.
If Chinese models become the default, future policies, data governance, and technological upgrades may be dictated by Beijing’s priorities, not Africa’s. For the 2,000 languages spoken across the continent, the window to build inclusive AI on African terms is narrowing. Key takeaways: - Chinese open-source AI models like Qwen, DeepSeek, and Kimi are cheaper and better suited to African languages, driving rapid adoption by developers. - Training AI on African languages can cost three to 30 times more than English, a barrier Western companies have not addressed. - China is actively cultivating African AI talent through competitions and study visits, mirroring its infrastructure playbook. - Western companies risk losing the African AI market entirely if they do not offer competitive, open-source alternatives soon.
What comes next is a test of Western engagement. Google’s Gemma offers a potential alternative, but adoption remains minimal. The Chinese government’s AI competition will bring a new cohort of developers into its ecosystem by year’s end.
As more African governments sign AI integration deals, the platforms they build on will lock in long-term dependencies. The question is whether Western companies will invest in Africa’s linguistic diversity before the market is fully captured.
Key Takeaways
— - Chinese open-source AI models like Qwen, DeepSeek, and Kimi are cheaper and better suited to African languages, driving rapid adoption by developers.
— - Training AI on African languages can cost three to 30 times more than English, a barrier Western companies have not addressed.
— - China is actively cultivating African AI talent through competitions and study visits, mirroring its infrastructure playbook.
— - Western companies risk losing the African AI market entirely if they do not offer competitive, open-source alternatives soon.
Source: Foreign Policy
