AI vs SaaS：先卖再问，市场只“卖对了一半”？

Core Viewpoint - Barclays highlights a critical technological distinction: AI tools are indeed encroaching on the application layer of SaaS companies, but they cannot shake the foundational "system of record" infrastructure, which is the core moat for companies like Salesforce and SAP [1][2]. Group 1: Impact of AI on SaaS Companies - The recent release of products like Claude Cowork by Anthropic has led to a significant decline in enterprise software stocks, with Salesforce and Workday dropping over 40% in the past 12 months [2]. - Investors are confused about the boundaries of AI capabilities, leading to a panic sell-off as they believe new AI tools will completely replace traditional SaaS software, resulting in a zero valuation for legacy companies [2][3]. - Barclays' report argues that a simplistic "one-size-fits-all" logic does not apply to most enterprise software companies [3]. Group 2: AI Capabilities and Limitations - Generative AI excels in pattern recognition and "draft generation," but its probabilistic nature poses fundamental limitations, particularly in scenarios requiring absolute accuracy [5]. - Traditional software operates on deterministic rules, ensuring consistent outputs, while AI software is probabilistic and cannot guarantee the same level of consistency [5][6]. - This indicates that AI operates at a higher level of abstraction and is not a direct replacement for traditional software [6]. Group 3: Mispriced Software Companies - Barclays identifies three categories of enterprise software companies that have been mispriced during the sell-off, starting with system of record companies like Salesforce, which provide critical data requiring certainty [9]. - SAP's position is even more secure, as it manages essential business data and workflows that cannot be handled by advanced generative AI models [9][10]. - The report suggests that AI will not replace these systems but will increase their importance, as AI agents will create more data touchpoints, raising the complexity that system records need to manage [10]. Group 4: Additional Misjudged Investment Opportunities - Besides system of record companies, Barclays points out two other categories that have been misjudged: beneficiaries of AI agents and AI computing providers [11]. - Companies like JFrog, Snowflake, and MongoDB may see increased usage due to the demand for more code and data driven by AI expansion [11]. - There is a logical contradiction in the market's reaction; if AI is powerful enough to disrupt the software industry, the demand for computing power should surge, yet companies like Oracle and CoreWeave have also faced significant sell-offs [11]. Group 5: Reevaluation of Software Sector Valuations - The market correction is deemed necessary for the application layer of enterprise software, which has long enjoyed inflated valuations due to controlling both infrastructure and interface [15]. - If AI technologies can overlay on system records, they may begin to erode the pricing power of SaaS companies [15]. - Barclays concludes that the era of easy high profits for bloated application layers may be over, but this does not signify the end of the entire industry [15][16]. Group 6: Market Sentiment and Future Outlook - The indiscriminate nature of the current sell-off indicates that investors with limited understanding of the software industry are making decisions based on extreme viewpoints [16]. - As understanding of AI capabilities and SaaS business models deepens, the market may reprice companies incorrectly categorized as "AI victims" [16].