DeepSeek plans shift to Huawei processors for training next generation models

DeepSeek is prioritizing domestic chipmaker Huawei to support its future AI model development due to American trade limits on Nvidia hardware. The company plans to use these chips for training models with trillions of parameters by late 2026 or early 2027. While DeepSeek still relies on existing Nvidia stock, the transition aims to secure a stable long term supply chain.