Is the AI Stock Pullback a Buying Opportunity? Morgan Stanley Names TSMC and 30 Key Players — 'Non-Tech Stocks' to Be the Next Frontier

On the 10th, Taiwan's stock market opened higher and fluctuated, with TSMC's share price gains narrowing by the close. The weighted index rose 702.85 points to finish at 44,928.76, reclaiming the quarterly line at 44,331. Morgan Stanley noted that the recent correction in AI infrastructure-related stocks primarily reflects short-term fund adjustments, not a deterioration in fundamentals. According to the latest 'AI Infrastructure Value Chain Heatmap,' Taiwan's foundry, semiconductor packaging and testing, and power supply supply chain firms—including TSMC, Hon Hai, and Quanta—are prominently featured. TSMC, a heavyweight stock, rose as much as NT$40 to NT$2,410 during the day but pulled back, closing up NT$10 at NT$2,380—the lowest point of the day. MediaTek gained NT$60 (1.54%) to close at NT$3,960, while Delta Electronics surged to its daily limit at NT$1,815, up NT$165, marking a two-week high. Morgan Stanley's list of 30 Taiwanese companies in the 'AI Infrastructure Value Chain Heatmap' includes: Semiconductor Manufacturing: ASE Holding, King Yuan Electronics, Tong Hsing Electronic, TSMC, UMC, Vanguard International Semiconductor, Win Semiconductors, Powerchip Technology, Grace Semiconductor, Andes Technology, Silicon Motion Technology. Server Components: Delta Electronics, Lite-On Technology, Yageo, Chicony Power, Sunon, Twinhead, Unimicron, Zhending-KY, Nanya PCB, Gold Circuit Electronics. Server Manufacturing: Chintech, Wistron, Wiwynn, Pegatron, Hon Hai, Quanta, Gigabyte,仁宝. Networking: Accton. Stephen Byrd, Morgan Stanley's Global Head of Thematic and Sustainable Development Research, stated that AI infrastructure delivers significant economic benefits globally, and recent stock corrections do not indicate weakening fundamentals. Concerns exist that companies may limit employee use of AI tokens to control costs, potentially affecting large language model (LLM) revenues. However, current average token usage per employee remains low, indicating substantial room for future growth. Morgan Stanley points out that AI development bottlenecks are shifting from chip supply to infrastructure, especially power supply. The report estimates that U.S. data centers will face a 38GW power deficit by 2028, expanding to 122GW by 2030. Investors are advised to monitor fuel cells, natural gas turbines, energy storage, and data center REITs with power assets. The next phase of the AI race will hinge on power supply stability.