中国银河证券:AI“泡沫化”程度仍然有限 维持对硬件端的推荐
智通财经网·2026-01-09 01:24

Core Viewpoint - The current stage of AI development differs from the internet bubble period, with hardware supply still in a state of shortage and a safety margin remaining in the market [1] Group 1: AI Development and Market Dynamics - Since 2022, AI technology has led to unprecedented infrastructure investments in the AI industry, creating a boom in the semiconductor hardware sector [1] - Concerns about an "AI bubble" are valid, as capital expenditures of major global cloud providers are expected to exceed their combined operating cash flow by 2026 [1] - Key differences between the current AI phase and the internet bubble include different driving enterprises, incomplete commercialization of AI, limited transmission of investment bubbles to the stock market, and the ongoing loose global monetary environment [1] Group 2: Supply Chain Constraints - The demand for high-performance computing is rigid due to the rapid development of multimodal large models and AI Agent functionalities, leading to a tight supply-demand relationship [2] - Global data center vacancy rates are at historical lows, and the consumption of electricity by data centers is impacting the elasticity of power supply [2] - Domestic computing power is still in a catch-up phase, with supply constraints persisting [2] Group 3: PCB Market Insights - The current PCB upcycle is primarily driven by growth in HDI and high-layer board shipments, with a projected compound annual growth rate of 20.3% for high-end HDI PCBs and 11.6% for boards with 14 layers or more from 2024 to 2029 [3] - The technology barriers for high-end HDI and high-layer boards are significant, with major companies leading the expansion [3] - Capital expenditures in the industry are expected to increase by 8.4% in the first three quarters of 2025 compared to the entire year of 2024, indicating that capacity expansion will still be limited relative to demand growth [3]

CGS-中国银河证券:AI“泡沫化”程度仍然有限 维持对硬件端的推荐 - Reportify