China's AI Chip Self-Sufficiency and the Shifting Computing Landscape

Due to US export regulations, the landscape of the AI chip market, once centered on the US, is rapidly changing. China is increasing its self-sufficiency through the development of proprietary AI chips such as Huawei Ascend, Cambricon, and Baidu Kunlun, and recently these domestic chips have begun to be used for inference and training of open-weight models as well.

US regulations have instead become a catalyst that accelerates China's technological innovation. Amid uncertainty in chip supply, Chinese companies are advancing not only in hardware but also toward increasing algorithmic efficiency. A prime example is DeepSeek, which introduced MLA (Multi-head Latent Attention) and GRPO (Group Relative Policy Optimization) to maximize computing efficiency.

In addition, new software platforms competing with NVIDIA's CUDA are emerging, forming a new ecosystem through synergy between hardware and software. This shift is expected to have a significant impact on the training and deployment of future AI models, as well as on the overall structure of the global AI economy.