《内部价值播客》讨论了Constellation Software (CSI)这家公司，它常被称为“软件界的伯克希尔·哈撒韦”。尽管其创始人兼前首席执行官马克·伦纳德（Mark Leonard，因健康原因于2023年9月卸任）早在2025年就曾表示该股被高估，投资者在当时的价格水平上只能获得8%的回报，但该股最近经历了有史以来最大的跌幅，在人工智能叙事背景下暴跌50%。他表示，需要“大幅下跌”才能实现25%的年化回报。 马克·伦纳德的背景独特，在进入风险投资领域并于1995年创立CSI之前，他曾从事各种工作。受巴菲特启发，他以长期主义的理念建立了CSI，专注于收购和持有垂直市场软件（VMS）公司。VMS指的是小众的、行业特定的软件，例如墓地管理系统，这些软件通常在少数几家公司占据主导地位的小型市场中运营，从而确保高利润率和资本回报。 CSI的业务模式高度去中心化，总部规模很小（不到20人），监管着几个运营部门（如Topicus、Harris、Vela）。这些部门又将资本配置决策进一步下放给其收购的公司。这种旨在最大程度减少官僚作风、培养小型团队自主性的结构，使CSI得以收购1,000多家公司。伦纳德的激励方法侧重于高投资资本回报率（ROIC）（历史上在30%左右，最近在20%中期），鼓励运营部门将利润再投资于新的收购或项目，而不是将现金上缴，以避免扭曲奖金结构。 CSI历史上一个显著特点是其股本数量：自IPO以来，它没有发行过任何新股，所有收购都通过现金流和债务进行。这反映了强大的股东友好文化，尽管伦纳德也曾表示由于信息不对称，他对股票回购感到不安。高管被要求将其税后奖金的75%用于购买CSI股票，这些股票必须持有至少四年。 股价近期下跌的主要原因是“人工智能叙事”。许多人担心AI可能会颠覆CSI的业务，方式一是让公司能够“随心所欲地编写”（vibe code）自己的软件，从而降低软件开发成本；方式二是导致“席位压缩”（seat compression），即AI代理取代人类用户，从而减少对软件许可证的需求。然而，CSI的辩护指出，其75%的收入来自具有粘性、任务关键型的*维护*合同，涉及基本服务（如公共交通或医院IT系统），在这些领域，转换成本高昂，客户规避风险。伦纳德还认为AI可能会增强现有职位而非取代它们，而在内部，AI可以提高效率、节省成本（员工占开支的70%）并加速收购来源的寻找。 伦纳德此前曾暗示会探索VMS之外的投资，这可能表明核心VMS市场存在潜在局限性，或者他希望更广泛地配置资本。一个例子是在热油（thermal oil）领域的“反向”大型交易，但最终未能实现。尽管马克·米勒（Mark Miller）是新任首席执行官，他在CSI拥有悠久的历史，但他能否在VMS之外执行新的战略转变，是投资者需要考虑的一个问题。 在估值方面，鉴于该股目前已从高点下跌50%，其交易市盈率远低于往年。丹尼尔的估值，采用12-13%的营收复合年增长率（CAGR）、小幅利润率扩张和25倍现金流倍数，估算出2,300-2,400美元的公允价值，这意味着12-13%的内部收益率（IRR）。然而，肖恩指出，在没有显著增长的情况下，25倍的市盈率本身并不便宜，并质疑考虑到AI的不确定性和领导层的变动，市场是否真的“过度反应”了。 两位主持人均表达了谨慎乐观的态度。丹尼尔虽然承认CSI是一家以合理价格出售的优秀公司，但更倾向于探索规模较小的VMS“模仿者”，以寻求潜在的更高增长和灵活性。肖恩表示同意，建议在投资Constellation之前，更深入地研究这些剥离公司或类似公司，特别是考虑到现有投资组合中已持有伯克希尔和Transdigm等其他收购型公司的股票。

The Intrinsic Value Podcast discusses Constellation Software (CSI), a company often dubbed the "Berkshire Hathaway of software." The stock recently experienced its largest-ever drawdown, plummeting 50% amidst the AI narrative, despite founder and former CEO Mark Leonard (who stepped down in September 2023 due to health) having stated in early 2025 that the stock was overvalued and investors should expect only 8% returns at those levels. He suggested a "massive drop" was needed for 25% annualized returns. Mark Leonard's background is unique, having worked diverse jobs before entering venture capital and founding CSI in 1995. Inspired by Buffett, he built CSI with a long-term mindset, focusing on acquiring and holding Vertical Market Software (VMS) companies. VMS refers to niche, industry-specific software, such as cemetery management systems, which often operate in smaller markets where a few players dominate, ensuring high margins and returns on capital. CSI's business model is highly decentralized, with a tiny head office (less than 20 people) overseeing several operating units (e.g., Topicus, Harris, Vela). These units, in turn, push capital allocation decisions further down to their acquired companies. This structure, designed to minimize bureaucracy and foster autonomy in smaller teams, has allowed CSI to acquire over 1,000 companies. Leonard's approach to incentives focused on high Returns on Invested Capital (ROIC) (historically in the 30s, recently mid-20s), encouraging operating units to reinvest profits in new acquisitions or initiatives rather than upstreaming cash to avoid distorting bonus structures. A remarkable aspect of CSI's history is its share count: it has issued zero new shares since its IPO, funding all acquisitions through cash flow and debt. This reflects a strong shareholder-friendly culture, though Leonard also expressed discomfort with buybacks due to information asymmetry. Executives are required to use 75% of their after-tax bonuses to buy CSI shares, which must be held for at least four years. The main driver of the recent stock decline is the "AI narrative." Many fear AI could disrupt CSI's businesses, either by allowing companies to "vibe code" their own software, reducing the cost of building software, or by leading to "seat compression" where AI agents replace human users, decreasing the need for software licenses. However, CSI's defense highlights that 75% of its revenue comes from sticky, mission-critical *maintenance* contracts for essential services (like public transport or hospital IT), where switching costs are high and clients are risk-averse. Leonard also suggests AI might enhance existing roles rather than replace them, and internally, AI could boost efficiency, save costs (staff is 70% of expenses), and speed up acquisition sourcing. Leonard has previously hinted at exploring investments beyond VMS, suggesting potential limitations in the core VMS market or a desire to allocate capital more broadly. An example was a large, "contrarian" deal in the thermal oil sector, which ultimately didn't materialize. While Mark Miller is the new CEO, with a long history at CSI, his ability to execute new strategic shifts outside VMS is a point of consideration for investors. In terms of valuation, with the stock now 50% down from its highs, it's trading at significantly lower multiples than in previous years. Daniel's valuation, using a 12-13% top-line growth CAGR, slight margin expansion, and a 25x cash flow multiple, estimates a fair value of $2,300-$2,400 USD, suggesting a 12-13% internal rate of return (IRR). However, Sean notes that a 25x multiple isn't inherently cheap without significant growth, and questions whether the market is truly "dramatically overreacting" given the AI uncertainty and the shift in leadership. Both hosts express cautious optimism. Daniel, while acknowledging CSI as a great company at a fair price, leans towards exploring smaller VMS "copycats" for potentially higher growth and agility. Sean agrees, suggesting a deeper dive into these spin-offs or similar companies before committing to a Constellation investment, especially given existing portfolio exposure to other acquirers like Berkshire and Transdigm.