读完一早上的访谈.
主题综述

SaaS 死了吗 · The SaaS Postmortem

主题综述

主题页（活文档）· 最近更新 2026-06-11 · 取材 8 篇访谈（按逐字稿全文核对）

更新日志

主流共识

几乎所有人在这一点上不分歧：应用层正在被 AI agent 重写，customer service / coding / 内部工具是当前最显眼的三个战场。

"The wave that we're riding of large language models and this next generation of AI is greater than any company riding it. And so don't fight AI."
「我们所乘的这波大型语言模型和下一代人工智能的浪潮，比任何公司驾驭它都更强大。所以不要与人工智能作对。」
Bret Taylor · Uncapped #42 Bret Taylor from Sierra
"The thing that is fundamentally different about this product cycle is that the software itself can actually do the work."
「这个产品周期从根本上不同的地方在于，软件本身实际上可以完成工作。」
Why AI Moats Still Matter · Why AI Moats Still Matter

第二点共识：per-seat pricing 在 agentic 场景里说不通，但要被什么取代还在吵——outcome-based、按使用量、按结果分成都有人押。

分歧在哪

阵营 A · "软件成本归零派"——SaaS 真的要塌

Klarna 的 Sebastian Siemiatkowski 给出了最干脆的版本：

"You should think that cost of creating software is going down to zero. And that means that everyone will be able to generate software at any point in time."
「你应该认为创建软件的成本会降到零。这意味着每个人都可以在任何时间点生成软件。」
Sebastian Siemiatkowski · 20VC: SaaS is Dead

而他认为打击 incumbents 最狠的不是软件本身变便宜，是迁移成本崩塌：

"The next thing that's going to hit everyone bad is the switching cost of data."
「接下来会严重影响每个人的事情是数据转换成本。」
Sebastian Siemiatkowski · 20VC: SaaS is Dead

Turing 的 Jonathan Siddharth 把同一逻辑外推到 SaaS 之外的整个知识工作领域：

"The glimpse of the future that I see is that all knowledge work is going to be automated. If a human's job involves looking at a computer, analyzing what's on the screen, using different tools, using a keyboard and a mouse, it's going to be automated. It's only a matter of time."
「我所看到的未来一瞥是，所有知识工作都将被自动化。如果一个人的工作涉及看电脑、分析屏幕上的内容、使用不同的工具、使用键盘和鼠标，它就会被自动化。这只是时间问题。」
Jonathan Siddharth · 20VC: Scale, Surge, Turing, Mercor

阵营 B · "SaaS 没死，只是被错读"——延伸而非替代

a16z 的 Anish Acharya 是最明确的反方声音：

"You have this innovation bazooka with these models. Why would you point it at rebuilding payroll or ERP or CRM? Right, you're going to take it and use it to extend your core advantage as a business or you're going to take it to optimize the other 90% that you're not spending on software."
「你手上这堆模型是一门创新火箭筒。你为什么要把它对准重建工资单、ERP 或 CRM？你应该用它来延伸你作为一家企业的核心优势，或者用它去优化你那 90% 没花在软件上的开支。」
Anish Acharya · Anish Acharya: Is SaaS Dead in a World of AI?

他的实证武器是 incumbents 涨价：

"If you look at public market SaaS companies, 75% have raised prices since ChatGPT was released. 75%. And they've raised prices meaningfully — the mean is 8% to 12%, but there's a large group that have raised it 25% or more. … Price is a measure of product market fit. If you have enormous competitive pressure, you are not raising prices, you're typically cutting prices."
「看上市的 SaaS 公司——自 ChatGPT 发布以来，75% 涨过价。75%。而且涨得很实在：均值 8%–12%，还有一大批涨了 25% 以上。……价格是产品-市场契合的度量。如果你面临巨大的竞争压力，你不会涨价，通常是降价。」
Anish Acharya · Anish Acharya: Is SaaS Dead in a World of AI?
"What you instead see is the native categories that did not exist before the product cycle being owned by startups."
「你反而看到的是，产品周期之前不存在的本土类别被初创公司所拥有。」
Anish Acharya · Anish Acharya: Is SaaS Dead in a World of AI?

Atlassian 的 Mike Cannon-Brookes 同样不接受"软件即将塌"的版本——他押注的是工程师*更多*而不是更少：

"Five years from now, we'll have more engineers working for our company than we do today, more software developers. … There's no doubt in my mind that we will create far more technology."
「五年后，我们公司雇的工程师会比今天多，软件开发者会更多。……我毫不怀疑，我们会造出多得多的技术。」
Mike Cannon-Brookes · 20VC: Atlassian CEO

他给护城河重新指了一个新位置——设计：

"I do think it makes design more valuable. Good design. Scarce resources probably get more valuable, and good design is quite scarce. … In the AI era, where software gets cheaper to create, there will certainly be far more software."
「我确实认为这让设计更值钱。好设计。稀缺资源大概率会更值钱，而好设计相当稀缺。……在 AI 时代，软件造起来更便宜，于是软件一定会多得多。」
Mike Cannon-Brookes · 20VC: Atlassian CEO

阵营 C · "重画护城河派"——SaaS 形态在变，但 defensibility 没消失

Bret Taylor 拒绝两边的二分法，把焦点放在定价：

"Our whole hypothesis is every company needs a website in 1997. Every company needs an agent in 2027."
「我们的整个假设是，每家公司在 1997 年都需要一个网站。每家公司在 2027 年都需要一个代理。」
Bret Taylor · Uncapped #42 Bret Taylor from Sierra
"Let's say you had an AI agent to generate leads for your sales team. You care about the number and quality of the leads. You really don't care how many tokens the model uses. In fact, it's not obvious to me that there's a correlation between used tokens and leads generated. … To the degree agents have a measurable outcome, outcome-based pricing feels like the secular business model for agents."
「假设你有个 AI agent 给销售团队找线索。你在意的是线索的数量和质量。你根本不关心模型用了多少 token。事实上，我都不确定'用掉的 token'和'生成的线索'之间有什么相关性。……只要 agent 有可衡量的结果，outcome-based 定价就感觉是 agent 时代那个'长期成立'的商业模式。」
Bret Taylor · Uncapped #42 Bret Taylor from Sierra

"Why AI Moats Still Matter" 的那期 a16z panel 给了一个比较精细的版本——SaaS-as-business-model 可能在变，但 SaaS-as-defensibility-pattern 没消失，只是位置换了：

"The defensibility of a software product resides, in my opinion, from owning the end-to-end workflow, from the context in which that it's applied, becoming the system of record, having a network effect, deeply embedding yourself within your customer."
「我认为，软件产品的防御性在于拥有端到端的工作流程、在于它被应用的上下文、成为记录系统、拥有网络效应、深度嵌入到客户内部。」
Why AI Moats Still Matter · Why AI Moats Still Matter
"The market opportunity for software today is no longer just IT spend, it's largely labor."
「如今软件的市场机会不再仅仅是 IT 支出，而很大程度上是劳动力支出。」
Why AI Moats Still Matter · Why AI Moats Still Matter

阵营 D · "Startup-shaped holes 派"——实证视角

Garry Tan / Lightcone 那期对 MIT 报告的反驳是这场争论里最贴地的声音——不谈 SaaS 形而上学，只看哪类公司实际在赚到企业的钱：

"The success rate of the ones where the enterprise went with an outside vendor like a Greenlight or a Tactile was much higher than the success rate of when they tried to build stuff themselves."
「企业选择像 Greenlight 或 Tactile 这样的外部供应商的成功率，远高于他们自己尝试构建东西的成功率。」
The Lightcone / Garry Tan · Inside The MIT AI Study
"There's this startup-shaped hole in basically every process or every sort of annoying system that should exist that doesn't exist yet."
「基本上每个流程、每个该存在但还不存在的烦人系统里，都有一个 startup 形状的洞。」
The Lightcone / Garry Tan · Inside The MIT AI Study

OpenAI Codex 的 Alex Embiricos 从更上游给了一个旁证——*所有* agent 终究会落回到 coding agent 形态，这对"SaaS 形态会变成什么"这件事其实暗示很强：

"All agents are actually coding agents because coding is just the best way for an agent to use a computer."
「所有 agent 实际上都是 coding agent，因为编码是 agent 使用电脑最好的方式。」
Alex Embiricos · 20VC: Codex vs Claude Code vs Cursor

都没说透的

我的看法

判断（不是事实）：这场辩论里两边都对了一部分，但他们在争的其实不是同一个问题。"SaaS is dead" 派多数在讲商业模型（per-seat、software-as-license），"SaaS not dead" 派多数在讲企业关系（embedded、systems of record、distribution）——前者大概率真的会被冲掉，后者大概率会幸存甚至加强。把"SaaS"这个词同时绑在两件事上，是这场辩论一直绕圈的根源。

我目前对这个判断的把握：中等偏低。我对"商业模型部分会塌"这一半的信心更高（per-seat 在 agent 场景里几乎不成立是显见的），对"客户关系部分会加强"这一半信心较低（很可能反而是 startup-shaped holes 派的判断更准——不是 incumbents 加强，而是新公司在 incumbents 没建好 AI 战壕之前抢走整片新地）。

还想知道什么

取材

继续读
线索
阅读线索：agent 不到 9 nines 可靠，那应该卖给企业还是消费者
线索
阅读线索：如果 coding 是 commodity，AI startup 的护城河长在哪儿
线索
阅读线索：推理成本一年跌 10×，AI app 该按什么定价
线索
摘报
综述
AI Agent 安全与治理
综述
Agent 代购时代：广告与商业的重写