Huatai Securities: DeepSeek-V4 Long Context Cost Reduction Drives Demand Expansion

Golden Finance reports that on April 27th, Huatai Securities research note stated that on April 24th, DeepSeek released the V4 series models and papers. The core change is a systematic upgrade of model architecture, training process, and infrastructure focused on low-cost practicalization around a 1 million token context. Huatai Securities believes that the market is likely to interpret V4 as “cost reduction and lower computing power and storage requirements,” but a more important marginal change is that, after the decrease in long-context costs, scenarios such as complex agents, multi-document analysis, long-cycle tasks, and online learning can see improved usability, with inference call volume and storage access frequency expected to expand. Potential catalysts include domestic computing power Day 0 adaptation, subsequent deployment of Ascend 950 super nodes, progress in DeepSeek multimodal and Engram routes, and industry development brought by the interoperability of domestic open-source model algorithms. (Jin10)