Shenwan Hongyuan Securities·2025-11-22 12:48Core Insights - The report highlights significant advancements in AI technology, particularly with Google's release of Gemini 3 Pro and Nano Banana Pro, which enhance multi-modal understanding and production capabilities [4][5][6] - The Chinese AI industry is expected to see accelerated innovation across computing power, models, and applications in 2026, transitioning from a competitive landscape to a more integrated ecosystem [4][5][6] - Huawei's Flex:ai container technology represents a key breakthrough in AI infrastructure, enabling efficient management of heterogeneous computing resources [4][5][6] Computing Power - The report identifies 2026 as the year of industrialization for domestic computing power, with significant advancements in domestic AI chips and supernodes, showcasing strong engineering capabilities [4][18][22] - Domestic supernodes are categorized into "multi-cabinet" and "single high-density" paths, with Huawei's CM384 and other solutions demonstrating competitive advantages [22][23][31] - Innovations in server architecture and cooling technologies are highlighted, with the potential to enhance overall computing performance and efficiency [27][30][31] Models - The gap between Chinese and American large models is narrowing, with domestic models like DeepSeek and Qwen3 showing competitive performance and cost-effectiveness [33][36] - The report predicts that the monetization of large models will accelerate in 2025, focusing on AI programming and multi-modal applications [33][34] - The introduction of mid-training as a distinct phase in model development is expected to enhance the performance and efficiency of large models [52][55] Applications - The report emphasizes the importance of industry know-how as a competitive advantage in the software sector, suggesting that large models cannot fully replace customized software solutions [58][60] - AI applications are in the early stages of penetration within the software industry, with significant growth potential anticipated as companies begin to disclose AI-related revenues [61][64] - The report draws parallels between the current AI application landscape and the early days of cloud computing, indicating a favorable investment window for software companies [64] AI Infrastructure - Huawei's Flex:ai technology is positioned as a critical component of AI infrastructure, enabling the unified management of various computing resources [65][67] - The report notes that traditional container technologies are insufficient for AI workloads, highlighting the need for specialized AI containers to meet evolving demands [67][68]