垂直领域 AI SaaS 是目前 VC 投资增长最快的品类——无论是 AI Agent 初创公司、AI 驱动的收购整合（roll-ups），还是原生 AI 全流程服务公司。 先暂时忘掉“AI 泡沫”这种宏观叙事。 垂直领域 AI 正在做一件传统 SaaS 从未做到的事： 它们触达的是 11 万亿美元的美国劳动力市场，而不是那 4500 亿美元的企业软件预算残羹剩饭。 Harvey 在最新公告中达到了 1.5 亿美元 ARR，从 2024 年底的 5000 万美元 ARR 三倍增长，今年已完成三轮融资，估值 80 亿美元。 Abridge 的 ARR 超过 1.17 亿美元，已部署在 150+ 家医疗系统中，包括 Mayo Clinic（梅奥）和 Duke（杜克）。 OpenEvidence 宣称 40% 的美国医生在使用他们的平台，并刚以 60 亿美元估值融资。 Glean 昨天宣布其 ARR 达到 2 亿美元。 这些公司的客户收入留存率普遍很稳健（这是投资界关注的关键点）。 传统 SaaS 往往只能捕获一个员工价值的 1–5%。 而垂直 AI 则能捕获 25–50% 的价值，因为它真正自动化了大量实际工作（数据来自 Scale Venture Partners）。 当软件能起草法律文件、生成医疗记录，而不是仅仅让人工流程更“顺滑”时，收费模型就会发生根本性改变。 于是，那些过去被认为“太小”的市场，如今都变得具备风投可投资性。 法律软件过去是一整个大类； 如今甚至有专门为“车祸律师”提供服务的软件（如 EvenUp）。 TAM 的扩张意味着即使是某个行业的一小块细分市场，也可能诞生十亿美元级公司。 全球垂直 AI 市场预计将从 2024 年的 129 亿美元增长至 2034 年的 1154 亿美元，复合增速 24.5%。 我原以为基础模型公司会在此时带来更强大的模型，但事实并没有发生，而且看起来距离实现仍然很远。 与那些每次 OpenAI 发布新功能就“被灭队”的工具型初创公司不同，垂直 AI 企业通过领域专用数据、内生的合规体系、深度嵌入的工作流构建起真正的护城河。 我也相信 OpenAI、Anthropic、Google 等都将逐步直接进攻每一个大的垂直领域—— 它们更想直接拥有客户，而不是只卖便宜的 token。 原因很简单：直接服务客户的毛利率更高，也更有价值。 OpenAI 巨大的算力成本需要大量资金支撑——单靠卖低价 token 是不够的。 他们必须与最大客户直接合作才能收回投资。 例如：Abridge 拥有超过 150 万次医疗问诊数据用于训练模型——这个数据壁垒根本无法复制。 垂直领域 SaaS 的估值倍数已接近横向 SaaS 的 2 倍： 它们的 EV/Gross Profit 中位数为 11x，而横向 SaaS 为 5x。 投资人已经意识到，这类公司更容易取得行业主导地位、构建粘性产品，并拥有更高效的商业化路径。 当然，大多数垂直 AI 初创公司依然会像普通创业项目一样失败。 但那些成功的公司，其价值将远超上一周期里横向 SaaS 的赢家—— 因为它们瞄准的是更大的预算、更高粘性的产品，并且落地在那些真正需要 AI 帮助的行业。 vertical AI SaaS is the biggest growing category of vc funding across - ai agent startups, roll ups (with ai improvements), or ai-native full service companies forget the broad AI bubble narrative for a second. vertical AI is doing something traditional SaaS never could: accessing the $11 trillion US labor market instead of competing for scraps of the $450 billion enterprise software spend Harvey hit $150M ARR based on recent announcement, up from $50M at end of 2024. now valued at $8B after three funding rounds this year alone. Abridge crossed $117M ARR deployed in 150+ health systems including Mayo Clinic and Duke. OpenEvidence claims 40% of US doctors use their platform, just raised at $6B valuation glean announced $200M ARR yesterday revenue retention numbers seem solid (key point of debate generally) traditional SaaS captured 1-5% of an employee's value. vertical AI captures 25-50% by automating substantial portions of their actual work (Scale Venture Partners). when software can draft a legal document or generate clinical notes, not just streamline how humans do it, the pricing model fundamentally changes this is why markets once considered "too small" are now venture-viable. legal software used to be one category. now theres software just for accident lawyers (EvenUp). the TAM expansion means subfractions of markets can support billion dollar outcomes the global vertical AI market is projected to grow from $12.9B in 2024 to $115.4B by 2034, a 24.5% CAGR i, inititally, believed that foundation model companies will bring way better models by now, but that hasn't happened yet, and they still seem far out unlike wrapper startups getting killed every time OpenAI ships a new feature, vertical AI companies build defensibility through domain-specific data, regulatory compliance baked into the product, and deep workflow integration. i also believe that openai, anthropic, google, and others will target every big vertical directly - they'd rather own the customer, than just provide cheap tokens the reason is that direct customer integrations have way higher margins and are more valuable than selling tokens openai may needs lot of money for compute cost - and selling cheap tokens is a losing proposition they have to work directly with the largest clients to make that money back eg Abridge has 1.5M+ medical encounters training their models. thats not replicable vertical SaaS already trades at nearly 2x the EV/Gross Profit multiples of horizontal peers, 11x versus 5x median. investors recognize these businesses achieve market dominance, build stickier products, and operate with more efficient go-to-market most vertical AI startups will still fail like most startups fail. but the ones that work will be worth more than the horizontal SaaS winners from the last cycle. theyre going after bigger budgets with stickier products in industries that actually need the help this is why im bullish