芯片ETF（512760）回调超2.4%，AI算力成本下探，把握布局回调机会

Group 1 - The core viewpoint is that the reduction in AI training and inference costs is driving application prosperity, with significant potential in edge AI, particularly through devices like headphones and glasses [1] - NVIDIA's Rubin platform AI chip mass production is expected to lead to a multiple increase in computing speed, spurring a surge in demand for computing power across various sectors including servers, AI chips, optical chips, storage, and PCBs [1] - Storage prices have bottomed out and are beginning to recover, with the packaging and testing sector's utilization rates gradually increasing, benefiting from advanced packaging demands driven by AI chips [1] Group 2 - TSMC's fourth-quarter revenue exceeded expectations, driven by strong demand for AI applications [1] - Domestic equipment is making continuous breakthroughs in advanced processes, with "advanced process expansion" becoming a key focus for self-sufficiency, while CoWoS and HBM are positioned to capitalize on AI industry trends [1] - The Chip ETF (512760) tracks the China Semiconductor Index (990001), which selects listed companies in semiconductor materials, equipment design, manufacturing, packaging, and testing from the Chinese A-share market to reflect the overall performance of semiconductor-related listed companies [1]