展望2026年，如何把握AI算力投资机会？

Group 1: AI and Chip Industry Outlook - The introduction of the 1.6T optical module is anticipated next year, corresponding to NVIDIA's next-generation Rubin architecture, which is expected to drive upgrades in the related supply chain and achieve volume and price increases [1] - The AI application sector is projected to experience significant growth next year, with ASIC fields likely accelerating, as companies like Google, OpenAI, and AWS are advancing ASIC-related solutions, which the domestic supply chain can meet [1] - The North American computing segment shows high certainty in demand due to the transition from NVIDIA's Blackwell architecture to Rubin architecture and the ramp-up of Google's TPU, positively impacting domestic optical modules, PCBs, and server demand [1] Group 2: Domestic Chip Market Dynamics - The core logic for domestic chips revolves around domestic substitution and self-control, with AI chips expected to continue gaining market share from traditional chips [2] - The semiconductor equipment sector has high certainty, as the production and capacity expansion of GPU chips rely on upstream equipment and materials, with demand growth in the storage sector likely reflected in etching and thin-film deposition equipment [2] - The chip sector, including chip ETFs and semiconductor equipment ETFs, is expected to perform well next year due to the ongoing demand expansion cycle across various segments of the domestic chip industry [2] Group 3: AI Applications and Market Sentiment - Current AI applications include lightweight solutions like AI writing and video generation, with larger fields such as robotics and smart driving, although the robotics sector has shown volatility due to Tesla's production guidance adjustments [3] - The AI application sector has not yet seen large-scale deployment, suggesting a cautious investment approach until clear demand signals emerge, although low valuations may present early investment opportunities [3] - The U.S. faces potential electricity supply issues, which could impact computing power support for future AI applications, highlighting a need for ongoing monitoring [3]