算力底座需求爆发，大模型极速迭代推升硬件高景气！借道天弘AI指数基金(A类011839/C类011840)捕捉算力红利

Core Insights - The global competition in the artificial intelligence (AI) sector has shifted from a focus on "model parameter competition" to "monetizing underlying computing power" since 2026, creating significant demand for computing infrastructure [1] - Major tech giants, including Microsoft, Google, Amazon, and Meta, are projected to invest approximately $650 billion in AI by 2026, primarily in data center construction, AI chip procurement, and computing network development [1] - The rapid emergence of multimodal large models and deep reasoning models has led to exponential growth in computing power consumption, necessitating a swift upgrade in network bandwidth [2] Industry Trends - The market for optical modules used in AI clusters is expected to reach $26 billion by 2026, reflecting a nearly 60% year-on-year growth, driven by the demand for high-speed optical modules [2][4] - Goldman Sachs has revised its forecasts for the global optical module market, increasing projections for 2026 and 2027 by 43% and 46%, respectively, with expected shipments of 38 million units for 800G modules and 14 million units for 1.6T modules [2] - The supply-demand balance in the hardware sector is favoring upstream hardware suppliers, who are experiencing faster revenue growth compared to downstream application sectors [2] Investment Opportunities - The Tianhong CSI Artificial Intelligence Theme Index Fund reflects a strong focus on "hard technology infrastructure," with a significant concentration in computing infrastructure stocks, aligning with the current investment logic of returning to "computing infrastructure" from "application bubbles" [5] - The index covers the entire AI industry chain, from underlying computing power to applications, with the top ten holdings heavily weighted in optical modules/communication (19.76%), AI chips/processors (18.77%), and AI servers/computing (4.53%) [5] - The fund's A-class shares are suitable for long-term investors seeking to capitalize on computing industry dividends, while the C-class shares cater to tactical allocation needs during cyclical peaks in computing infrastructure [7]