China’s aggressive open-source artificial intelligence strategy could pose a credible challenge to OpenAI and other Western AI leaders, according to Tiezhen Wang, a former executive at Hugging Face and current chief of staff at Moonshot AI.
Speaking to Rest of World, Wang outlined how Chinese tech companies are betting heavily on open-source models as a counterweight to the closed, proprietary systems developed by US firms like OpenAI, Anthropic, and Google.
The strategic shift has implications beyond the US-China tech rivalry. For African developers and startups, the rise of open-source AI models from China could mean greater access to cutting-edge technology without the cost barriers associated with commercial APIs from Western providers.
Open-source as competitive strategy
Wang said Chinese companies view open-source AI not just as a technical philosophy but as a deliberate competitive strategy. By releasing powerful models publicly, Chinese firms can accelerate adoption, build developer ecosystems, and reduce dependence on US-controlled infrastructure.
“The open-source approach allows Chinese companies to leapfrog some of the advantages that come from being first,” Wang told Rest of World. “It’s about building a coalition of users and developers globally.”
Major Chinese AI labs including Alibaba, Baidu, and Moonshot have released models rivalling GPT-4 in capability, often with fewer restrictions and lower inference costs. These models are increasingly being adopted by developers in emerging markets.
Why this matters for Africa
African AI developers have long faced barriers accessing frontier models due to cost, API restrictions, and data sovereignty concerns. Open-source models reduce these friction points.
Kenyan and Nigerian developer communities have begun experimenting with Chinese open-source models for applications in local languages, agriculture, and fintech. The models can be fine-tuned on local datasets and run on cheaper infrastructure than proprietary alternatives.
“For African developers, open-source means you’re not locked into paying OpenAI or Google every time you make an API call,” said Segun Adeyemi, a machine learning engineer based in Lagos. “You can host the model yourself or use cheaper providers.”
However, experts caution that open-source models still require significant compute resources to train and fine-tune, which may limit accessibility for smaller African startups without cloud credits or subsidised compute access.
Geopolitical dimensions
Wang’s comments come amid escalating AI competition between Washington and Beijing. The US has imposed sweeping export controls on advanced chips to China, aiming to slow Chinese AI development. In response, Chinese firms have doubled down on efficiency and open-source collaboration.
Wang said the chip restrictions have forced Chinese researchers to focus on optimising models rather than scaling them indefinitely. That efficiency focus could benefit developers in bandwidth-constrained environments, including much of sub-Saharan Africa.
The geopolitical tug-of-war also raises questions about data governance and digital sovereignty. African policymakers are weighing whether to align with US-led AI frameworks or explore alternatives, including models from China and the European Union.
Wang said he expects the gap between open-source and proprietary models to narrow further in 2026, particularly as Chinese labs refine reasoning and multimodal capabilities.
As the AI race intensifies, access to powerful, affordable models could determine which markets can participate meaningfully in the next wave of digital transformation.