格隆汇1月13日｜知名投资者、"大空头"Michael Burry警告，大型科技公司靠相对少的投资赚取巨额利 润的时代正在结束，人工智能(AI)是主因。他认为投资人应关注投入资本回报率(ROIC)，而非营收增长 或市场规模。他指出，AI正推动微软、谷歌和Meta等公司，从过去轻资产的软件模式，转向由数据中 心、芯片和能源主导的资本密集型的硬件公司。即使AI帮助科技巨头扩大了市场，ROIC下降仍可能在 未来数年对股价造成压力。 ...

Core Viewpoint - The dialogue between Andrew Ross Sorkin and Dario Amodei highlights contrasting perspectives on AI's future, with Sorkin expressing skepticism about a potential AI bubble, while Amodei emphasizes the tangible value and growth of AI in the industry [6][32]. Group 1: Andrew Ross Sorkin's Perspective - Sorkin views the current AI landscape as reminiscent of historical financial bubbles, suggesting that the rapid growth in AI investment and reliance on AI for GDP growth could lead to a similar collapse as seen in 1929 [33][39]. - He raises concerns about the sustainability of AI investments, questioning whether the returns justify the massive expenditures being made by companies like OpenAI [38][39]. - Sorkin's macro perspective indicates a cautious approach, focusing on the potential risks and uncertainties surrounding AI's economic impact [33][39]. Group 2: Dario Amodei's Perspective - Amodei presents a more optimistic view, citing significant revenue growth in the AI sector, with projections of annual revenues increasing from approximately $1 billion in 2023 to $80-100 billion by 2025 [34][35]. - He argues that the willingness of companies to invest substantial amounts in AI services is a direct indicator of its value, contrasting the skepticism of outsiders with the confidence of industry insiders [35][38]. - Amodei emphasizes the importance of safety and regulation in AI development, advocating for a balanced approach that ensures AI's growth does not outpace its governance [30][31]. Group 3: Industry Risks and Opportunities - Amodei warns that OpenAI could face significant financial challenges due to its aggressive investment strategy, highlighting the inherent risks in the AI industry where companies may either be overly conservative or excessively aggressive [39][42]. - The dialogue suggests that while AI may create opportunities, it will also lead to job displacement, with a focus on the need for individuals to adapt and learn to leverage AI effectively [51][53]. - The conversation underscores the importance of recognizing market fluctuations as opportunities rather than threats, encouraging a proactive approach to investment in the AI sector [53][54].

周二盘前，埃克森美孚(XOM.US)宣布将2024-2030年盈利增长预期上调至250亿美元、现金流增长预期 上调至350亿美元，两项指标较此前计划各提升50亿美元;同时维持至2030年每年280亿至330亿美元的资 本支出预期不变。 埃克森预计，至2030年期间盈利年均增长率将达13%，现金流将实现两位数增长;而通过持续股票回 购，每股收益增速将实现更高水平。 该公司表示，在布伦特原油实际价格维持每桶65美元的前提下，预计未来五年累计盈余现金流将达约 1450亿美元;至2030年，投入资本回报率(ROIC)将超过17%。 埃克森预计，到2030年上游产量将达到550万桶油当量/日，高于此前预测的540万桶油当量/日;其中二叠 纪盆地和圭亚那资产的产量到2030年将达到370万桶油当量/日。 该公司表示，其直至2030年的企业计划中，盈利和现金流前景的改善，反映了优质资产贡献的增强、更 有利可图的业务组合以及运营成本的降低。 截至发稿，埃克森美孚盘前股价上涨0.5%，报116.58美元。 ...

Core Insights - The report focuses on the evolution and optimization of profitability evaluation indicators, particularly the Return on Equity (ROE) within the PB-ROE framework, utilizing DuPont analysis to dissect the structure and potential issues of the ROE metric [1][4] - The report introduces a comprehensive profitability factor by optimizing the indicator system based on the analysis of ROE, aiming to enhance the dual optimization of the PB-ROE strategy framework in both valuation and profitability aspects [1][4] Section Summaries 1. In-depth Exploration of ROE and Profitability Styles - The PB-ROE strategy combines valuation levels with shareholder return rates, reflecting a company's ability to generate profits from shareholder capital, which directly influences net asset growth [10][14] - The relationship between ROE and GDP indicates that listed companies, as a significant part of the economy, have shown substantial growth in revenue, with the total revenue of A-share companies exceeding 72 trillion yuan in 2024, compared to 3.37 trillion yuan in 2004 [14][15] - ROE is categorized under quality style in investment factors, reflecting a company's financial health, profitability, reliability, and long-term growth potential [22][26] 2. ROE and DuPont Analysis - ROE is tested using both quarterly and TTM (Trailing Twelve Months) metrics, with the quarterly ROE factor showing a higher average Rank IC of 4.06% compared to 2.78% for TTM [28][29] - Historical high ROE stocks tend to underperform in future price performance, while portfolios constructed based on future ROE show significant excess returns, indicating the importance of ROE stability [33][34] - DuPont analysis breaks down ROE into three components: net profit margin, total asset turnover, and equity multiplier, providing a comprehensive assessment of a company's profitability, operational efficiency, and leverage [42][46] 3. ROE De-leveraging Analysis - The report discusses the linear separation of leverage factors from ROE, highlighting the economic relationship between ROA and ROE, and the limitations of ROA as a profitability measure [3][15] - The introduction of RONOA (Return on Net Operating Assets) and FCFFIC (Free Cash Flow Return on Invested Capital) aims to provide more accurate profitability assessments by excluding non-core operating activities and mitigating earnings management risks [4][6] 4. Comprehensive Profitability Factor - The integration of stable ROE, stable ROIC, stable RONOA, and FCFFIC forms a comprehensive profitability factor, enhancing the performance of the PB-ROE strategy [4][6] - The active quantitative stock selection strategy based on the PB-ROE framework has achieved an annualized return of 20.42% since 2010, significantly outperforming benchmarks like the CSI 800 [4][6]

Investment Rating - The investment rating for the industry is "Outperform the Market" (maintained rating) [1] Core Insights - A review of the capital expenditures and performance of major cloud vendors in Q1 2025 shows that Microsoft, Meta, and Amazon experienced a slowdown in year-over-year capital expenditure growth, while Microsoft saw stable revenue and profit growth driven by AI [2] - Overall cloud revenue growth is slowing, but all vendors continue to see increases in cloud revenue. The return on invested capital (ROIC) for Meta and Amazon has decreased quarter-over-quarter due to the impact of overall revenue and profit trends [2] Summary by Sections 1. Capital Expenditures and Performance Review of Cloud Vendors - In Q1 2025, Microsoft's capital expenditure decreased by 5.3% quarter-over-quarter to $21.4 billion, primarily due to normal fluctuations related to data center leasing delivery times. Future capital expenditures are expected to increase [8][14] - Meta's capital expenditure in Q1 2025 was $13.7 billion, a year-over-year increase of 104%, but slightly below expectations. The company raised its full-year capital expenditure guidance to between $64 billion and $72 billion, reflecting additional investments in data centers to support AI initiatives [20][27] - Google's capital expenditure in Q1 2025 was $17.2 billion, a year-over-year increase of 43.2% and a quarter-over-quarter increase of 20.5%, slightly exceeding market expectations [35][41] - Amazon's capital expenditure in Q1 2025 was $24.3 billion, a year-over-year increase of 63.09%, but a quarter-over-quarter decrease of 7.6%, primarily directed towards AWS to meet AI service demands [48][55] 2. Cloud Revenue and ROIC Situation - Microsoft reported Q1 2025 cloud revenue of $26.7 billion, with Azure and other cloud services growing by 33%, contributing 16% to Azure's revenue growth from AI [70] - Amazon's AWS revenue in Q1 2025 was $29.3 billion, a year-over-year increase of 16.9%, with AI business achieving annualized revenue in the billions and continuing to grow at triple-digit percentages [55][70] - Google's cloud revenue in Q1 2025 was $12.26 billion, a year-over-year increase of 28.06%, driven by GCP growth, including AI infrastructure and generative AI solutions [70] - The ROIC for Microsoft and Google increased quarter-over-quarter, while Meta and Amazon saw declines, reflecting the impact of AI demand on capital expenditures and overall performance [71]

Core Viewpoint - Concerns over oversupply in the U.S. data center market have pressured the U.S. stock market, particularly impacting the computing power industry, with significant declines in major tech stocks like Nvidia [2][3] Group 1: Market Impact - As of March 26, the Nasdaq Composite fell by 2.04%, and the Philadelphia Semiconductor Index dropped approximately 3.3%, with Nvidia's stock declining over 5.7%, resulting in a market value loss of $169.1 billion (approximately 1229.1 billion RMB) [2] - The investment bank TD Cowen reported that while data center demand is still showing year-on-year growth, Microsoft has canceled or postponed more data center leases than initially expected [2][3] Group 2: Data Center Supply and Demand - Microsoft has canceled or postponed approximately 2 gigawatts of data center leases in the U.S. and Europe over the past six months, indicating an oversupply relative to current demand forecasts [3] - Despite the cancellations, major tech companies continue to invest heavily in AI infrastructure, with Microsoft planning to invest $80 billion (approximately 585.6 billion RMB) in AI data centers for the fiscal year ending June 2025 [3] Group 3: Capital Expenditure Trends - In 2024, the capital expenditures of major U.S. tech companies are projected to reach record highs, with Microsoft, Amazon, Google, and Meta expected to spend $75.6 billion, $77.7 billion, $52.5 billion, and $37.3 billion respectively, reflecting year-on-year growth rates of 83%, 62%, 63%, and 35% [5] - The total capital expenditure for these four companies in 2024 is estimated at $243.1 billion (approximately 1.8 trillion RMB), representing a 63% increase year-on-year [5] Group 4: Strategic Adjustments - Microsoft is reassessing its data center leasing and construction plans after a period of rapid expansion to support AI training needs, indicating a potential oversupply or idle capacity in certain regions [6] - The company's strategic shift may also be influenced by its partnership with OpenAI, as OpenAI may increasingly rely on self-built or third-party collaborations for its computing needs, leading Microsoft to adjust its data center expansion plans [6]