Serenity: The pullback in the AI hardware sector is mainly due to short-term deleveraging, while its long-term demand logic remains unchanged.

Serenity stated in a recent report that the sharp pullback in the AI hardware sector is more an overcorrection driven by short-term deleveraging rather than a deterioration in fundamentals. It pointed out that the latest financial results of companies such as Bloom Energy and Teradyne show significant revenue and profit growth, performance pre-announcements from Chinese optical module manufacturers are also strong, and upcoming financial reports from firms including Lumentum, SanDisk, and SK Hynix will further confirm accelerating AI demand. Alphabet has raised its 2026 capital expenditure target to $195 billion to $205 billion, which is one of the most important signals of AI demand from large hyperscale cloud providers. Meanwhile, market concerns about further sharp interest rate hikes by the Federal Reserve this year and overcapacity in China’s memory chips have been significantly exaggerated. Long-term agreements (LTAs) between cloud providers such as Meta and Google, as well as collaborations among memory manufacturers including Samsung, SK Hynix, and Micron, also reflect that AI storage demand such as HBM has a structural growth trend. Serenity noted that as long as revenue and EPS of AI industry chain enterprises continue to accelerate, the long-term investment logic remains valid, and it expects the AI theme will gradually recover in the future, though it cannot determine the exact timing.