The research report from Huatai Securities points out that globally, mainstream large models are primarily concentrated in China and the United States. According to data from Artificial Analysis, leading large model manufacturers in the United States include OpenAI, xAI, Anthropic, and Google; in China, DeepSeek, Alibaba, Zhipu, Kimi, and MiniMax are relatively prominent. Due to limited access to high-performance computing resources in China, under the context of adopting reinforcement learning and post-training paradigms, overseas models tend to enhance performance through scale expansion, while domestic manufacturers excel more in architectural optimization. In terms of investment advice, focus can be placed on the four directions of computing power, storage, electricity, and applications.
