Huatai Securities: AI Agents Accelerate Demand for Computing Power and Memory, Speeding Up Supply Chain Localization

According to a research report from Huatai Securities obtained by 36Kr, the AI industry is transitioning from large model pre-training to AI Agent commercialization in 2026, leading to accelerating growth in inference computing demand. The report identifies three key investment themes: first, the computing power chain, where domestic computing power closed-loop is accelerating, super-node interconnection and storage upgrades are underway, and the inflection point for inference demand is clear; on the edge side, new product cycles for foldable phones and AI glasses are approaching, presenting structural innovation opportunities. Second, power and passive components, where AI power consumption drives volume and price increases for MLCCs, inductors, capacitors, and power semiconductors, with price hike cycles coinciding with domestic substitution. Third, self-sufficiency, where localization progress in upstream manufacturing, equipment, and components is accelerating, and the value of advanced packaging is systematically increasing. Separately, South Korea's Ministry of Economy and Finance stated in a press release that the government will closely communicate with the U.S. Treasury to enhance mutual understanding and trust in the foreign exchange market, and will continue to cooperate to stabilize the market. This follows the U.S. Treasury's decision to keep South Korea on its currency monitoring list.