美股云计算和互联网巨头24Q4总结：云厂Capex指引仍乐观，AI应用ROI路线清晰或将迎来催化

Investment Rating - The report maintains a positive outlook on major US cloud computing and internet giants, indicating a favorable investment rating for the sector [1]. Core Insights - The cloud computing industry is experiencing a temporary slowdown in growth due to capacity constraints, but the capital expenditure (Capex) for 2025 is expected to remain high, reflecting strong demand for AI computing power [2][9]. - Major players like Google, Microsoft, Amazon, and META are projected to significantly increase their Capex in 2025, with a total expected to reach approximately $340 billion, representing a year-over-year growth of around 35% [10][30]. - The report highlights the importance of AI applications and the return on investment (ROI) associated with them, suggesting that advancements in AI will drive further penetration into various user scenarios [15][23]. Summary by Sections Cloud Computing Performance - Google Cloud's revenue growth for FY24Q4 was 30.1%, which was below Bloomberg's expectations due to capacity constraints [2][3]. - Microsoft Azure's revenue growth guidance for FY25Q3 is projected at 31%-32%, indicating no significant improvement from the current quarter [2][3]. - Amazon AWS's growth remained stable, aligning with expectations, but faced a reduction in operating profit margin due to early server replacements [2][3]. Capital Expenditure (Capex) - The total Capex for Google, Microsoft, Amazon, and META in 2024 is estimated at $250.4 billion, with a year-over-year increase of 63.2% [9][10]. - Google has provided a more optimistic Capex guidance for FY25, expecting it to reach $75 billion, a growth of over 40% compared to FY24 [25][30]. - Amazon's FY24Q4 Capex was $27.8 billion, with expectations for FY25 to exceed $83 billion, reflecting a growth of over 30% [10][30]. AI Applications and ROI - The report emphasizes the declining cost of computing power for AI applications, which is expected to enhance the ROI of AI implementations [9][15]. - The optimization of AI large model algorithms is anticipated to accelerate the adoption of AI applications in internet and SaaS companies [15][23]. - The report notes that the integration of AI into traditional search functions, such as Google's AI Overview, is gradually increasing its penetration rate, although the speed is still limited by computing costs [23][26]. ASIC Chip Development - Major cloud providers are accelerating the development of ASIC chips, with a focus on optimizing for AI inference workloads [21][23]. - Google is set to launch its TPU v6 Trillium chip, designed specifically for AI inference, which is expected to lower the cost of internal AI workloads [21][23]. - Amazon is advancing its Rainier project to build a large-scale Tranium2 chip cluster for AI training and inference, which is expected to enhance AWS's performance [30].