快手二季度表现稳健，华尔街热议“可灵AI”：流水表现超过账面收入，全年指引有点低？

Core Viewpoint - The article highlights that despite the seemingly modest revenue figures for Kuaishou's AI video generation application "Keling," its actual cash flow performance may significantly exceed reported financial data, indicating a conservative revenue guidance for the year [2][3]. Group 1: Keling's Revenue Performance - Keling's revenue for Q2 reached 250 million yuan, representing a 67% quarter-on-quarter growth, and the full-year revenue guidance for Keling has been raised to 125 million USD, doubling previous expectations [2][5]. - Multiple investment banks believe that Kuaishou's revenue guidance for Keling is overly conservative, with JPMorgan noting that actual cash receipts for the quarter could exceed 300 million yuan due to delays in revenue recognition [4][5]. - Goldman Sachs projects that Keling's total revenue will continue to grow rapidly over the next two years, reaching 250 million USD and 365 million USD in 2026 and 2027, respectively [3][4]. Group 2: Core Business Performance - Kuaishou's core business showed strong resilience, with online marketing service revenue growing 13% year-on-year to 35 billion yuan, driven by external advertising and e-commerce [6][7]. - The e-commerce segment was a highlight, with GMV increasing by 18% year-on-year, surpassing the previous quarter's growth of 15%, primarily driven by significant growth in shelf-style e-commerce and short video e-commerce [6][7]. Group 3: AI Investment and Profitability Balance - Kuaishou is increasing its capital expenditure for the year to 12 billion yuan, primarily for Keling-related investments, while maintaining that the impact of AI investments on overall profitability will be controlled within 1-2 percentage points [9]. - JPMorgan emphasizes Kuaishou's ability to maintain stable profit margins while investing in AI, reflecting strong execution capabilities [9]. - Despite Keling achieving gross profit profitability, the company expects that incremental investments in computing power will have a manageable impact on the profit and loss statement, with further potential for cost savings in training and inference [9].