华福证券发布研报称，谷歌(GOOGL.US)、微软(MSFT.US)、Meta(META.US)、亚马逊(AMZN.US)均已 披露第二季度财报，今年二季度单季资本开支分别为224亿美元/242亿美元/170亿美元/322亿美元，分别 同比+70%/+27%/+101%/+83%，同比仍保持高增速。大模型的竞争与迭代仍在持续，算力资本开支仍将 维持较高水平，二季度财报披露的资本开支数据和对下个财季/下一财年的指引基本验证了此前观点， AI产业趋势进一步得到强化。AI产业趋势自今年以来重新强化，驱动力来自推理需求大幅度增长，AI 资本开支有望持续增加。 今年以来推理算力消耗呈倍数级增长，推理侧算力对训练侧算力需求的承接推动总算力需求进一步增 长。根据华尔街见闻援引火山引擎公布的数据，截至2025年5月底，豆包大模型日均tokens使用量超过 16.4万亿，较去年5月刚发布时增长137倍。 根据谷歌5月I/O大会披露的数据，2025年4月每月处理的tokens总量超过480兆，这一数据在去年4月为 9.7兆，同比增长50倍。本次业绩会上，谷歌披露当前月tokens处理量达到980兆，达到4月的2倍，近一 个月时间 ...

Investment Rating - The industry rating is "Outperform the Market" indicating that the overall return of the industry is expected to exceed the market benchmark index by more than 5% over the next 6 months [14]. Core Insights - The report highlights a significant increase in capital expenditures among major North American cloud providers, with Q2 capital expenditures reported as follows: Google at $22.4 billion (+70% YoY), Microsoft at $24.2 billion (+27% YoY), Meta at $17 billion (+101% YoY), and Amazon at $32.2 billion (+83% YoY) [3]. - The upward revisions in capital expenditure forecasts for 2025 by these companies reflect strong demand for cloud products and services, with Google increasing its forecast from $75 billion to $85 billion, Microsoft expecting over $30 billion in the next quarter, Meta adjusting its range from $64-72 billion to $66-72 billion, and Amazon indicating that Q2 expenditures represent the level for the second half of the year [3]. - The report emphasizes the ongoing competition and iteration in AI models, suggesting that capital expenditures related to computing power will remain high, further reinforced by the strong Q2 earnings reports and future guidance [3]. Summary by Sections Industry Dynamics - The report notes a substantial increase in inference computing power consumption, which has seen a multiplicative growth this year, driven by the demand for training-related computing power. As of May 2025, the daily token usage for the Doubao large model exceeded 16.4 trillion, a 137-fold increase from the previous year [4]. - Google reported that the monthly token processing volume reached 980 trillion, doubling from April's figure, indicating significant growth in user engagement and demand for AI applications [4]. Investment Recommendations - The report suggests that the AI industry trend has been reinforced this year, driven by a substantial increase in inference demand, leading to sustained growth in AI capital expenditures. It highlights the importance of PCB in the AI hardware supply chain, with a recommendation to focus on companies such as Huadian Co., Shenghong Technology, Shennan Circuit, and others due to expected benefits from supply-demand dynamics [5]. - Additionally, it recommends monitoring server manufacturing firms like Industrial Fulian and liquid cooling supply chain companies such as Invec and Sihua New Materials [5].

Group 1 - The core viewpoint of the news highlights the significant growth in the AI sector, particularly driven by Google's Gemini app, which has surpassed 450 million monthly active users and experienced over 50% growth in daily request volume compared to Q1 [1] - The company has increased its capital expenditure for Q2 2025 by 70% year-on-year, with a focus on technology infrastructure, allocating two-thirds to server investments and one-third to data centers and network equipment to meet customer demand [1] - The company anticipates a tight supply-demand relationship in 2026 and has raised its 2025 capital expenditure forecast by 13% to $85 billion, up from the previous $75 billion, with further increases expected in 2026 [1] Group 2 - The ChiNext AI ETF (159388) has seen an opening rise of over 3% and can fluctuate by up to 20% in a single day, tracking the ChiNext AI Index and covering key sectors such as hardware, software, and computing power [1] - The ETF is heavily invested in leading companies in the optical module, communication equipment, and integrated circuit sectors, benefiting from the global demand for AI computing power and hardware upgrades [1] - Investors interested in the AI sector can consider the ChiNext AI ETF (159388) for a streamlined investment approach [1]