Global South Adopts Hybrid AI Stacks, Splitting US Chips and Chinese Models

Saudi Arabia launched the Humain-m3 Arabic-language AI model earlier this month using technology from Chinese startup MiniMax. [1] The state-backed company that commissioned the model, Humain, has simultaneously formed computing capacity deals with American technology companies including Nvidia. [1] Countries across Latin America, Southeast Asia, and Central Asia are also pairing American chips with Chinese open-source models. [1]

In August, Brazil announced an artificial intelligence investment of roughly $444 million that is effectively divided between American and Chinese technologies. [1] Approximately $255 million of this funding is directed toward supercomputing infrastructure in Rio de Janeiro developed alongside China's Huawei. [1] According to Kyle Chan at the Brookings Institution, many Global South countries want to avoid becoming dependent on a single technology source and do not attach the same security stigma to Chinese technology as Washington. [1]

This bifurcated approach to AI development presents a growing complication for Washington’s technological competition with Beijing. By layering Chinese models like those from MiniMax on top of American hardware from Nvidia, nations in the Global South are establishing a new paradigm of hybrid AI infrastructure. This suggests that U.S. dominance in semiconductor sales does not guarantee a monopoly over the software layer of global AI ecosystems. What the sources don't address: How U.S. export controls might eventually evolve to address hardware being used to run Chinese AI models in third-party nations.