虚高的ARR，才是AI商业最大“泡沫”

Core Viewpoint - The article discusses the misuse of Annual Recurring Revenue (ARR) as a valuation metric for AI companies, highlighting that the traditional application of ARR, which works well for SaaS businesses, does not translate effectively to the AI sector due to different revenue models and volatility in income [1][4][12]. Group 1: Misapplication of ARR in AI - ARR is a key metric for SaaS companies due to high customer retention rates and predictable revenue streams, with U.S. SaaS companies typically achieving a Net Dollar Retention (NDR) rate of over 100% [3]. - In the AI sector, ARR is often manipulated, with companies using inflated metrics such as highest monthly or daily revenues to calculate ARR, leading to a distorted view of their financial health [2][7]. - New AI startups have reportedly increased their total revenue from $0 to $2 million in just a few months, raising concerns about the authenticity of these figures [5][6]. Group 2: Reasons ARR is Inapplicable to AI Companies - The business model for AI companies is shifting from fixed pricing to performance-based pricing, which increases revenue unpredictability and complicates annual revenue forecasts [12]. - Many early revenues for AI companies come from experimental contracts with large clients, which do not guarantee long-term commitment or stable income [13]. - AI companies face higher operational costs, with resource consumption rates between 50-75%, compared to 20% for traditional SaaS companies, affecting profitability [14][15]. Group 3: Alternative Valuation Approaches - The article suggests moving from static income predictions to evaluations that reflect actual market conditions, considering both the speed of customer revenue growth and the share of customer spending on AI products [15]. - Emphasis should be placed on assessing the quality of revenue by breaking down different income types and applying varied evaluation standards [15].