According to a research report by Huatai Securities, domestic AI chips are now on the cusp of a price increase. This is primarily due to a broad-based increase in upstream costs and a mismatch between the supply and demand for local computing power. Huatai Securities notes that, amidst a growing demand for independent control and a stronger bargaining position, domestic AI chip companies are still well-positioned to pass on these cost increases. Moreover, their profit margins are expected to become more flexible as they leverage their scale advantages. Specifically, the driving forces behind this trend include:

Firstly, cost pass-through. There has been a comprehensive rise in costs across various upstream segments, including HBM (High Bandwidth Memory), substrates, packaging, and foundry services. Notably, the procurement cost of HBM is projected to surge by over 50% in the latter half of the year, exerting a significant upward pressure on overall costs.

Secondly, the supply-demand imbalance. With the rapid advancement of multimodal large models and AI Agents, the token call volume of domestic models has skyrocketed. This has, in turn, fueled a rapid growth in demand for computing power on the inference side. Consequently, a substantial supply-demand gap for domestic AI chips is anticipated to emerge from 2025 to 2027.

In addition, potential catalysts for the domestic computing power sector encompass the launch of new-generation products, the outcomes of test adaptations, and the realization of price increases.