DeepSeek V4 fully adapts to Huawei Ascend chips for AI independence

DeepSeek's V4 model has achieved full adaptation to Huawei's Ascend AI computing series, marking a significant step in China's push for domestic AI infrastructure independence. The integration, completed in April 2026, allows the model to run inference workloads natively on Huawei's Ascend chips without requiring any overseas hardware. This eliminates dependency on suppliers like NVIDIA, which has faced export restrictions to China. The move leverages Huawei's Ascend ecosystem, which includes the Ascend 910B and 910C processors, designed for high-performance AI training and inference.

This adaptation enables Chinese enterprises to deploy DeepSeek V4 at scale for applications such as natural language processing, code generation, and enterprise search, all while maintaining supply chain security. The successful integration also signals that Huawei's Ascend platform has reached enterprise-grade maturity, with software toolchains and runtime optimizations capable of handling complex AI models. As a result, China's AI sector reduces geopolitical risk and accelerates its path toward full-stack self-reliance in artificial intelligence.