Huatai Securities: AI Agents Accelerate Demand for Computing Power and Memory, Speeding Up Supply Chain Localization
Huatai Securities analysts predict that by 2026, the AI industry will shift from pre-training large models to commercializing AI agents, leading to accelerated growth in inference computing demand. They highlight three key areas: a closed loop of domestic computing power, increased demand for passive components due to AI energy consumption, and faster import substitution.
According to a report from Huatai Securities, by 2026 the artificial intelligence industry will shift from pre-training large language models to the commercial deployment of AI agents, driving a significant increase in demand for inference computing. This optimism is tied to three main areas: first, the formation of a closed loop for domestic computing power, including advancements in superinterconnects and memory, to meet the growing inference demand; second, the rise in AI energy consumption is boosting demand for passive components such as MLCC capacitors, inductors, capacitors, and power semiconductors, leading to higher prices and volumes; third, the localization of upstream equipment and components is accelerating, and the value of advanced packaging is increasing. The news also includes a paragraph unrelated to AI, which states that South Korea's Ministry of Finance will continue close cooperation with the U.S. Treasury to stabilize the currency market after Korea was added to the currency monitoring list.
Source: 36Kr —
original