科股早知道：AI应用普及拐点已至，海内外算力需求持续景气

Group 1 - DeepSeek online model has been upgraded to version DeepSeek-V3.1-Terminus, indicating a significant advancement in AI applications driving the demand for computing power [2] - First Shanghai Securities expresses optimism about the sustained high growth of computing power demand driven by AI applications, suggesting that the domestic computing power capacity bottleneck has been broken and is expected to see a release by 2026 [2] - The overseas computing power demand is anticipated to remain robust as AI applications continue to expand [2] Group 2 - High Bandwidth Memory (HBM) has emerged as the main growth driver in the DRAM market, effectively addressing bandwidth bottlenecks, high power consumption, and capacity limitations [3] - Yole forecasts that the global HBM market size will grow from $17 billion in 2024 to $98 billion by 2030, with a compound annual growth rate (CAGR) of 33% [2] - The introduction of HBM significantly enhances AI performance and quality, surpassing traditional memory performance [3] Group 3 - The HBM industry chain includes upstream suppliers of semiconductor materials and equipment, midstream HBM production, and downstream applications in AI, data centers, and high-performance computing [3] - The TSV process in HBM manufacturing is identified as a major challenge, involving complex processes such as photolithography, coating, and etching, representing the highest value segment [3] - The domestic HBM production is deemed essential, with current developments indicating it is in the early stages, presenting expansion opportunities for upstream equipment and materials [3]