华为昇腾、寒武纪、海光信息称适配DeepSeek新模型 AI概念股集体拉升

Group 1 - AI-related concept stocks experienced a collective surge, with companies like TongHuang Technology hitting the daily limit up of 20%, and others such as Kaipu Cloud and TuoerSi rising over 5% [1] - DeepQiuSuo officially released its new generation large language model DeepSeek-V3.2-Exp, which features a sparse Attention architecture that reduces computational resource consumption and enhances model inference efficiency [1] - Cambrian announced its adaptation to DeepSeek-V3.2-Exp, leveraging the Sparse Attention mechanism to significantly lower training and inference costs in long-sequence scenarios [1] Group 2 - Haiguang Information announced seamless adaptation and deep optimization of DeepSeek-V3.2-Exp on its DCU, showcasing excellent performance and validating its high versatility and ecological compatibility [2] - Huaxin Securities reported that the era of domestic AI chips has arrived, with a complete industrial chain established from upstream advanced processes to downstream model acceleration by major companies [2] - Zhongyin Securities indicated that the commercialization of AI applications is approaching a turning point, with accelerating demand for computing power and significant breakthroughs in domestic computing capabilities [2] Group 3 - Guotai Haitong Securities noted that large models and generative AI are rapidly transitioning from the cloud to consumer terminals, with major companies intensifying the construction of integrated AI ecosystems [3] - The ability to create "AI + hardware" killer applications will be a key competitive factor in the next wave of computing platform competition, influencing capital market assessments of technology company growth [3]