Huatai Securities: AI Agents Spur Demand for Computing Power and Storage, Accelerating Domestic Control in AI Supply Chain
Huatai Securities research indicates a significant shift in the AI industry by 2026, moving from large model pre-training to the commercial implementation of AI Agents. This transition is expected to drive accelerated growth in demand for inference computing power. The firm identifies three key investment themes within this evolving landscape. Firstly, the AI supply chain is seeing a faster formation of a domestic computing power ecosystem. This is supported by advancements in ultra-node interconnection and storage upgrades, which are crucial for meeting the clear inflection point in inference demand. The emergence of new products like foldable phones and AI glasses on the terminal side also presents structural innovation opportunities. Secondly, the increased power consumption driven by AI applications is boosting both volume and prices for MLCCs, inductors, capacitors, and power semiconductors. This price increase cycle is further amplified by the trend of domestic substitution. Lastly, the drive for autonomous and controllable systems is accelerating the localization of upstream manufacturing, equipment, and components, while simultaneously increasing the value of advanced packaging solutions.
AI Agent commercialization marks a critical inflection point, shifting industry focus from foundational model training to practical application deployment. This transition necessitates substantial upgrades in inference computing power and storage, creating complex supply chain dynamics. The emphasis on domestic control and localization within China's AI sector reflects a strategic imperative to build resilient and independent technological capabilities. This approach, while potentially fostering domestic innovation and market growth, also raises questions about global collaboration, interoperability standards, and the long-term efficiency of duplicated R&D efforts. The increasing power demands of AI hardware also present significant challenges related to energy consumption and thermal management, which will become increasingly important as AI integration deepens across consumer electronics and industrial applications over the next decade.
AI-generated to prompt reflection — not editorial opinion, not advice, not a statement of fact. How this works.