20VC：SpaceX完成对xAI的收购 | 2026年SaaS大屠杀：公开市场崩盘 | 微软损失3600亿美元市值 | 英伟达与OpenAI的1000亿美元投资纠纷 | Waymo以1100亿美元估值融资160亿美元

我认为只有两种类型的公司。

I think there are only two types of companies.

对于私营公司，它们要么以惊人的速度增长，要么无法获得融资。

For private companies, they're growing at insane rates or they're unfundable.

而对于上市公司，它们的增长在加速，而不是放缓。

And for public companies, they're accelerating and not decelerating.

无论属于哪一类，除非你只是需要一份薪水，否则投入其中几乎都是浪费时间。

Whichever category, it's a waste of time almost unless you just need a paycheck to be with either of the other categories.

我根本不关心你是AI、SaaS还是非AI、金融科技公司。

I don't even care whether you're AI, SaaS or non AI, FinTech.

你是否像野兽一样在增长？

It's are you growing like a beast?

像野兽一样。

A beast.

我们正在观察这些差异巨大的增长率。

We're looking at these growth rates that are so divergent.

就像哈里说他感觉从他所学到的来看，世界已经完全不同了，我们从未见过如此高的增长率，也从未见过如此明显的放缓。

Like they're almost When Harry said he feels like it's a new world from what he learned, we never saw growth rates like this and we never saw deceleration like this.

对吧？

Right?

以前的范围要窄得多。

It was a much narrower band.

现在简直疯狂至极。

Now it's utterly insane.

简直疯狂至极。

It's utterly insane.

对于风投来说，真正棘手的问题是：什么时候该放弃投资组合中的公司？

And the really tough question for Venture is when do you give up on the portfolio companies?

什么时候该请走廊另一头刚加入20天的初级员工来接手这些投资？

When do you bring in a Lemkin from down the hall, the junior kid that just joined 20, and have him handle the investments?

什么时候该放弃？

When do you give up?

因为在过去，我们不愿意放弃，但现在你会放弃吗？

Because in the old days, we didn't wanna give up, but do you give up now?

不会。

No.

事实上，无论是在大学还是其他领域，你都会更早地放弃。

The the truth is across the board, university, you give up much sooner.

这就是你在风投领域看到的趋势。

This is the trend that you're seeing across venture.

我不是说对我们而言，但你在整个风投行业都能看到这种现象，因为错过下一个哈维、下一个拉戈尔、下一个快速爱碗的机会成本太高了。

I'm not saying for us, but you see it across venture because the opportunity cost of missing the next Harvey, the next Lagore, the next rapid Love Bowl is so high.

你必须这么做。

You have to.

我觉得这就是我所看到的。

I think you that's what I see.

再说一遍，也许是因为

Again, maybe it's

我不是说这样很好，罗里。

I'm not saying it's good, Rory.

我说的是很多人确实会这么做，因为在大型合伙企业的猎杀式游戏中，作为初级到中级合伙人或主管，你的业绩是根据你完成的交易来评判的。

I'm saying many do because you have to in a heat seeking missile game of large partnerships where you're judged for the deals that you do as a young to mid level partner or principal.

该死。

Fuck.

我只需要投中哈维或利瓜拉就行。

I just need to get in Harvey or Liguora.

我不在乎我有没有那些项目，但得继续前进。

Doesn't matter if I have the dogs, but move on.

你是不是觉得这并不重要。

It doesn't matter if you yeah.

我只是停一下，因为从某种程度上说，这种行为如果被推向极端，最终会导致这样的结果：如果你连续明确放弃五个项目，却指望第六个能救你，而且你根本不在前五个上花时间，那么如果第六个也没救成你，你就全输了。

I'm just pausing because at some level, that behavior in the end go when pushed to extremes just results in remember, if you blatantly write off five in a row and assume the six will save you, and you don't spend any time on the five, well, if the six doesn't save you, you've lost all six.

我不认为你能把一家差公司变好，但我确实觉得，收回一到两倍的投资和直接全部注销之间，有着天壤之别。

I don't think you can make a bad company good, but I do think there was a whale of a difference between getting one to two times your money back and just writing it all off.

我知道这个逻辑。

And I know the logic.

只要你能获得10倍或20倍的回报，其他都不重要。

Provided you get the 10x or 20x, nothing else matters.

但有时候你能达到10倍或20倍，有时候市场却达不到，你只是需要想想这一点。

But there's times when you get to 10x and 20x, and then there's times when it's not there in the marketplace, and you just want to think about that.

你也要考虑你与创业者的关系。

You also want to think about your relationship with the entrepreneur.

我承认，我坚持到错得离谱。

I admit, I hang on to the point of being wrong.

我承认。

I admit it.

这是一种人性的缺陷。

It's a it's a human failure.

我讨厌放弃。

I hate to quit.

但你看到同样的创始人表现出同样的行为，他们比以前更早地放弃了，因为他们意识到了机会成本。

But you were seeing the same founders have that same behavior, which is quit so much earlier than they did before, too, because they're aware of the opportunity cost.

是的。

Oh, yeah.

现在创始人放弃也毫无压力。

Founders have no problem quitting now either.

毫无压力。

No problem.

这并不意味着必须像罗里所说的那样。

It doesn't mean that it has to be the same to Rory's point.

对。

Right.

如果你是个创始人，除非你知道，否则在任何给定时间你都不会拥有很大的投资组合，除非你的基金里有几亿美元。

And if you're a founder, unless you know, you don't you don't have a big portfolio at any given time, unless you've got a few 100,000,000 in your fund.

当结果没那么分化的时候，你更容易相信它们之后会爆发，对吧？

When results were a little less divergent, you could believe more that they would pop later, right?

当最好的公司以100%的速度增长，而你却只能挣扎在50%的增长时，你还能看到复苏的希望。

When the outcome When the best ones grew at 100 and you'd be struggling at 50% growth, you could see reignition.

当差距变成20%对500%的增长时，优秀的产品总能脱颖而出。

When it's 20 versus 500% growth, a great product can always come.

这些大语言模型对每个人都是开放的。

These LLMs are available to everybody.

我想说的是，这就是为什么我对这么多创始人失去了信心。

Here's the thing that I This is why I've lost confidence in so many founders.

对我来说，去年是严厉的爱。

To me, last year, it was tough love.

对我来说，今年只是太艰难了，因为大语言模型对每个人都是开放的。

For me this year, it's just tough because the LLMs are open to everybody.

你没有任何借口。

You have no excuse.

你没有任何借口在法律科技领域里，却不打造一个能与Harvey、Liguora或任何你想比的人竞争的产品。

You have no excuse to be in legal tech and not have built a competitive product to Harvey or Liguora or whoever you want.

你有充足的时间。

You had plenty of time.

或者对于AIGC，你也有时间。

Or to AIGC, you had time.

做

Do

我认为LLMs。

think I LLMs.

是的。

Yeah.

安全护栏。

Guardrails.

万岁。

Hooray.

构建你自己的安全护栏。

Build your own guardrails.

杰森，正如罗里常说的，我正在掌握，这很棒，但那我呢？

Jason, as Rory always says, I'm mastering, that's great, but what about me?

如果你在风投的语境下应用‘这很棒，但那我呢？’这句话，是不是意味着我只在资助那些用AI取代人类的劳动力替代产品？

If you apply that's great, but what about me in venture's day, does that mean I only have a job funding labor displacement products that use AI to replace humans?

不，他根本没这么说。

No, he's not even saying that.

他只是说，事情在变糟。

He's just saying, shit go up.

你只资助那些能增长十倍的项目。

You only have a job funding things that grow 10x.

投资超高速增长的项目。

Fund mega growth.

在一个最佳公司增长三倍的世界里，增长两倍的公司看起来也像是有竞争力的候选者。

In a world where the best companies grow 3x, a 2x growth company feels like it could be a contender.

杰，你到底想说什么？

What are saying, Jay?

和往常一样，你表达得非常清晰。

And as usual, you express it very crisply.

在一个最佳公司增长10倍的世界里，如果你只增长2倍，就会觉得何必费劲呢？

In a world where the best companies grow 10 x, if you're doing two x, you just feel like, why bother?

现在你正在

Now you're

成长，尤其是当你身家性命都押上时，尤其是当你不是基金的管理普通合伙人、性命攸关时，你可能连这种奢侈都没有。

growing Especially if your neck is on line, especially if you're not the managing general partner of the fund and your neck's on the line, you might not even have the luxury.

这正是哈里的观点。

That was sort of Harry's point.

因为公司发展得更快，你的时间更少。

Because the companies go faster, you have less time.

我认为，在风投领域，你已经没有十年时间来证明自己了。

You don't have ten years to prove yourself anymore in venture, I don't think.

也许你还有。

Maybe you do.

我不这么认为。

I don't think so.

我认为你在风投领域只有二十二个月的时间来证明自己。

I think you've got twenty two months to prove yourself in venture.

我觉得你只有二十二个月。

I think you get twenty two months.

当我们能以超过三年的长远视角来看待这次繁荣时，将会很有趣——看看哪些快速增长的公司持续成长了，哪些虽然快速增长却因商业模式错误而失败了，以及哪些缓慢增长的公司最终通过复利效应变得更强。

It will be interesting when we have a longer perspective than three years of this boom to see which were the fast growers that kept growing, which were the fast growers that flamed out because their economic model was wrong, and which were the slow growers that kept compounding and improved in the end.

我只想说，尤其是从互联网泡沫时期来看，不要假设快速增长的公司一定会完全等同于长期成功的公司。

And all I'll say is from the .com boom in particular, don't assume that the former will entirely track the latter.

会有一些公司实现十倍增长，但当你退一步看时，会发现它们的商业模式根本站不住脚。

There will be companies that have 10x growth where you step back and go, oh my god, the economics were just wrong.

也会有一些公司以两倍或三倍的速度增长，并持续保持这种节奏，五年后，它们的业务会变得极具吸引力。

There will be companies that grow 2x or 3x that keeps compounding at 2x or 3x and fast forward five years, have very compelling businesses.

我算是个温和的增长爱好者，但我不会只做一个增长狂。

I'm a mild growth junkie, but I'm not gonna be just a growth junkie.

我只是觉得，2026年面临的根本性挑战是，我已经对这一点失去了信心。

It's just I think the existential challenge for 2026 is I've lost faith in that.

听我说，稳定复利没什么不好。

Listen, nothing wrong with steady compounding.

对吧？

Right?

我们被付钱并不是为了盲目跟风。

There's nothing especially we're paid to go along.

对吧？

Right?

有时候我们有十年、十五年，甚至二十年的时间，对吧？

We have ten, fifteen, even twenty years sometimes, right?

没关系的。

It's okay.

我的意思是，这并不完美。

Like, it's not perfect.

你当然希望在24个月内获得一百亿美元的回报，但作为早期投资者，最终如果初创公司花上十年时间，其实也没关系，对吧？

Like, you'd love to have a $10,000,000,000 outcome in twenty four months, but it's okay if the startup takes, at least as an early stage investor, a decade's fine at the end of the day, right?

问题是，我看到所有没有以异常速度增长的事物都在衰退。

The problem is I see everything decaying that isn't growing at abnormal rates.

我看到一切都在衰退。

I see everything decaying.

我能闻到这种气息。

I can smell it.

我不只是从数据中看到这一点，还从各种前兆中看到了。

Not only do I see it in the numbers, I see it in the precursors.

我在潜在客户中看到了。

I see it in leads.

我看到成交率在下降，而不是上升。

I see it in close rates going down, not up.

我看到当你的竞争对手以十倍的价格收费时，你却无法提高自己产品的价格。

I see it in inability to charge more for your product when your agentic competitors are charging 10 times as much.

所以我只是闻到了衰败的气息，而不是持续的复利增长。

So I just smell decay rather than constant compounding.

我愿意以合理的估值投资一个年化80%增长的公司。

I'll take an 80% compounder at a decent valuation.

我甚至不需要知道他们具体需要做什么。

I don't even need to know what they need to do.

那就是2021年的Tiger。

That was Tiger in 2021.

对吧？

Right?

我就是闻到了它们身上所有的衰败气息，罗里。

I just smell decay in all of them, Rory.

我在董事会会议上就能闻到这种气息。

I just smell it at the board meeting.

我能在投资中闻到它，当投资者更新报告在月底后28天才发出来时，我也能闻到。

I smell it in the invest I smell it when the investor update comes twenty eight days after the after the end of the month.

我到处都能闻到腐烂的气息。

I just smell it everywhere, decay.

我不想评论什么腐烂的气息，但我确实明白两倍的增长是不够的。

I think I'm not going to comment on the smell of decay, but I do understand 2x growth isn't enough.

你必须始终关注的是，你的相对市场地位正在衰退。

One of the things you always have to look at is your relative market position decaying.

而你现在就处在这种情况中。

And that's where you are right.

如果你的增长是两倍，而且没人做你做的事，你还处于领先地位，那就没问题。

If you're growing at 2x and there's no one else doing what you're doing and you're in the lead, it's fine.

别慌。

Don't panic.

这就是正在发生的事。

This is what's happening.

你越接近那些增长十倍的东西，就越容易被吸入黑洞漩涡。

The closer you are to something that's growing at 10x, the more likely you are to be sucked into the black hole vortex.

所以，如果你有一个直接竞争对手正在以10倍的速度增长，那么从定义上讲，你每天都在失去优势。

So and if you've got a direct competitor going 10 x, you are by definition losing every day.

所以，这部分问题完全关乎你相对于竞争对手的位置。

So some part of this is all about where are you relative to competition.

是的。

Yeah.

当然。

For sure.

如果我进一步追问，回到你的框架，罗里，你说过有些公司能维持高增长率，而有些则会迅速衰落。

If if I push you, going back to your framework, Rory, you said there's ones which will sustain those high growth rates, and there's ones which will flame out.

如果让我请你举出当今两个典型例子，分别属于这两个阵营，你会选哪些？

If I were to push you on well known examples today for one in each camp, what would you put as one in each camp?

你知道，我尽量避免当一个喷子，所以对于后者，我可能会谈一些类别。

You know, I try to avoid being a hater, so I might do categories on the latter.

对吧？

Right?

而前者，虽然老生常谈，但你还是会这么说。

And the former, it's trite, but you'll say it.

当然，核心模型公司都在做。

Any of obviously, core model companies do it.

是的，说Entropic很容易。

Yes, it's easy to say Entropic.

你几乎因此得不到任何加分。

You almost don't get any points for that.

我相信我会选一些更难的例子。

I do believe so I'll pick some harder ones.

我相信，那些所谓的说唱类机会——我故意用了贬义词——但我相信在这些基础上可以构建许多有吸引力的企业级应用，所有这些应用即使增长率只有3倍、4倍、2倍，甚至更低，也能持续复利增长。

I do believe the many of the rapper type opportunities I'm deliberately using the pejorative I do believe there's many compelling enterprise applications to be built on top of those and all the apps that will compound durably, even if the growth is 3x, 4x, 2x even.

你提到过Liu这个案例，我们在GCI投资了Harvey、Lagoon。

You've mentioned the Liu case we have an investment there at GCI, Harvey, Lagoon.

所有这些都会持续复利增长。

All those things are going to compound.

可能会有竞争，但这些都是很好的领域。

There might be competition, but those are great categories.

我认为，我们不得不转型商业模式以实现复利增长的明显例子，是许多消费者主导的超高增长但利润率不佳的领域，比如创意工具，类似这样的东西，它们可能不得不改变模式。

I think the obvious example of things that we'll have to transition the business model to compound are many of the consumer led ultra high growth but uncompelling margins, creative tools, some of the stuff like that maybe where they will have to morph the model.

我不是说它们不会，但最终你呈现出来的会和现在大不相同。

I'm not saying they won't, but what you show up with in the end will be different than what you're showing up with now.

你们能帮我理清我一直困惑的分歧吗？就是关于HubSpot的外汇经常性收入（ARR），你们提到过，当我们讨论挑战型公司时，我其实是在想HubSpot的外汇ARR以及传统CRM提供商的困境。

Can you guys help me understand the divergence that I've been struggling with, which is the Forex ARR on HubSpot, where you said I was kind of thinking of that when we were talking about the challenge companies, like the Forex ARR on HubSpot and the challenge position of traditional CRM providers.

然后你再看下一代CRM提供商的数量，无论是你们的Days、Atios，还是我们投资的Xero、Revo。

And then you look at the number of CRM next gen providers, whether it's your Days or your Atios or we're in one called Xero, Revo.

我的意思是，这样的公司名单足足有50家，真的有这么多。

I mean, list is is 50, literally.

帮我解释一下这个现象。

Help me understand that.

我的想法是，现在有两件事在同时发生。

Well, thought my thought was there's two things going on.

一个是我说的‘爸爸风投’。

One is what I call dad VC.

这是一种有时甚至脱离时代的风投，要么投资他们熟悉的东西。

It's the VC that sometimes even gets out of touch that either invests in what they know.

比如我懂CRM、懂ERP，或者投资他们孩子觉得酷的东西。

I know CRM, know ERP, or they invest in what their kids say is cool.

我孩子回来后说他们在用Snap，我就投了，做了Snap的种子轮。

My kids came back and I invested, I did the seed round at Snap because my kids are using it.

以前我刚入行时，还经常嘲笑这种‘爸爸风投’。

And Dad VC is not actually, I used to mock it when I started investing.

但现在我发现它真的有效。

Now I see it works.

它确实有效。

It does work.

就像风投们不断投资的无限多的高管助理项目。

You know, like endless executive assistant investments VCs would do.

每天要和30位创始人安排会议实在太难了。

It's so hard to schedule meetings with 30 founders a day.

我有自己的日程安排。

I've got a schedule.

这正是典型的爸爸风投投资。

That's classic dad VC investment.

所以我认为，有些风投理解CRM，知道这是一个巨大的市场，他们觉得它有问题——用引号说，就像Salesforce有问题一样。

So I think some of it is that VCs understand CRM, and they understand it's a large market, and they think it's broken, in air quotes, like Salesforce is broken.

我要告诉你，我觉得我用过的最没毛病的应用就是Salesforce。

And I will tell you, I think the least broken app I use is Salesforce.

它非常强大。

It's very powerful.

所以我认为这确实是一种爸爸风投。

So I think there's a dad VC.

但我要回答你的问题，我们在这里使用并正在使用的那些产品都是高度自主的。

But I will say the one to answer your question, the products that we use and are using here are hyper agentic.

所以有时候CRM是一个非常宽泛的术语，有时候我们实际上只是指销售自动化，有时候我们只需要其中一部分功能。

So sometimes CRM is such a broad term and sometimes we really just mean SFA and sometimes we really just need one part.

如果这个产品只是出去自动获取新客户，那和Salesforce可不是一回事。

If all this product does is go out and automatically acquire new customers, that's not the same thing as Salesforce.

而很多这类产品正在做的就是它们的自主式客户获取。

And that's what a lot of these products are doing is their agentic customer acquisition.

这非常强大。

And that's very powerful.

事实上，这是最容易销售的东西之一。

In fact, that is one of the easiest things to sell.

嘿，嘿，Rory，我是Jason和Harry。

Hey, hey, Rory, it's Jason and Harry.

我们来自NextGen CRM。

We're from NextGen CRM.

只要5万美元，我们就为你打造了一个代理，能帮你获得500万的新预订。

For $50,000 we've built an agent that will get you 5,000,000 of new bookings.

你想试试吗？

Would you like to try it?

Rory是新任CMO。

And Rory's the new CMO.

他还有十个月就会被解雇，但他有500万美元的预算。

He's got ten months until he's fired, and he's got a $5,000,000 budget.

你知道吗？

You know what?

Rory可能会流失，但如果我们是最棒的销售团队，并且有一些成功的案例，Rory会给我们50到100美元，而Pipedrive却只能勉强拿到每月8美元。

Rory may churn, but if we're the best sales guys and have some good case studies, Rory's going to give us $50 to $100, while Pipedrive struggles to get $8 a month.

这正是今天正在发生的事。

That's what's happening today.

即使这一切最终都会流失，但‘代理可以替代多人工作’这个想法非常强大。

And it may all churn, but this idea that the agents will do work of many humans is very powerful.

我认为，这些最优秀的CRM初创公司真正做的，就是用一个代理取代十人、二十人甚至五十人。

And I think that's what the best of these CRM startups are really doing, is replacing ten, twenty, 50 humans with an agent.

我

I

我喜欢你的回答，杰森。

like your answer, Jason.

我想仔细分析一下。

I want to pick it apart.

因为我们实际上是在说，Salesforce 和 HubSpot 在 CRM 领域是被极度低估的上市公司。

Because what we're basically saying is Salesforce and HubSpot are extraordinarily low valued public companies in the CRM space.

有很多风投正在为一大批下一代 CRM 公司注资，估值达到收入的 50 到 100 倍。

There's a whole bunch of VCs funding a whole bunch of next generation CRM companies at 50 or 100 times revenues.

你怎么调和这两者之间的矛盾？

How do you reconcile those two things?

杰森，你提出了这两种理论。

And what, Jason, you put forth these two theories.

第一种理论是‘老爸理论’，意思是可怜的老风投们，像我这样的人，只懂 CRM，所以我们只是在做一款新的 CRM。

One theory is the dad theory, which is poor old VCs, me, all they understand is CRM, so we're just doing a new CRM.

你所说的隐含着一个观点，而我也同意：如果新的CRM只是在旧CRM基础上加了一些AI功能，它将因为两个原因而失败。

And implicit in what you're saying is, and I agree with you, by the way, if the new CRM is pretty much the old CRM but with some AI bells and whistles, it will fail for two reasons.

第一个原因是，旧CRM的增长率已经放缓，因为市场已经饱和。

The first is the old CRM growth rate has slowed because the market saturated.

现在你试图推销一种稍微好一点的产品作为替代品，即使它有一些很酷的AI功能。

Now you're trying to do a replacement sale of a slightly better product, even if it has some pretty AI features.

这根本行不通。

That's just not going to work.

我同意你的观点。

I agree with you.

你所说的另一点是，你可以设想一种独立的CRM初创公司类别，即下一代CRM公司，它们不只是复制旧CRM的工作流程，而是真正地完成工作，比如生成线索，甚至可能直接促成销售。

What you're saying also is, but you could envisage a separate category of CRM startups, next generation CRM startups, which don't just replicate the workflow part of old CRM, but literally do the work, including generate pipe, maybe even generate sales.

在某个时刻，这种模式会变得极具吸引力。

And at some point that would become compelling.

而正是这些公司能够抢占市场份额。

And those are the guys who can take market share.

这就是艺术。

That's the the art.

是的。

Yeah.

对吧？

Correct?

我的意思是，你看，我投资了一家还没上线的公司，但有一家我没投资、我们却在使用的公司叫Artesyn，它就是一个AI销售开发工具，还包含更多功能。

I mean, look, I'm an investor in one that hasn't launched, but one that I'm not investor in that I know that we use is called Artesyn, which is just an AI SDR tool and more.

他们上个月实现了200万美元的收入。

They did 2,000,000 last month.

这是从零开始、一年内增长到200万美元的。

It's $2,000,000 up from nothing twelve months ago.

其中一部分原因就是我们会为你带来客户。

And part of that is just we'll get you customers.

现在你当然可以批评这个产品之类的，但今天要卖出价值50或100美元的产品，这其实并不难，对吧？

Now you can criticize the product or whatever, but that is not the hardest sell for $50 or a $100 today, right?

这并不是一种威胁。

And that is not a threat.

它对HubSpot和Salesforce构成间接威胁。

It is an indirect threat to HubSpot and Salesforce.

我们可以讨论这一点，因为每有一美元流向那里，就让追加销售变得更加困难，但同时它也运行在Salesforce之上。

We could talk about it because for every dollar that goes there, it makes the upsell just that much harder, but it also runs on Salesforce.

因此，当Salesforce作为核心平台时，它反而让Salesforce变得更强大。

So at the same time, it makes Salesforce more powerful when Salesforce is the hub.

但这些都很容易推销。

But those are easy sell.

说实话，在当今世界，尤其是当你向增长型公司销售时，这很容易推销。

Honestly, in today's world, especially if you're selling to growth companies, it's an easy sell.

我不是说产品不需要好，但天啊，这不需要20通电话、七次拜访和办公室关门才能成交。

I'm not saying the products don't have to be good, but man, this is not a 20 call, seven visits, and the office close.

这确实很容易推销，而且我们也有投资。

It is an easy sell, though, and we have investment as well.

你学到的一个经验是，每个B2B公司的市场策略都有其独特性，有时候你承诺能带来500万美元的PIPE，但实际上却做不到。

One of the learnings you've had is there are nuances around every B2B company's go to market such that one time in two, you promise I'll generate $5,000,000 in PIPE, but you don't.

因为也许目标市场非常狭窄，只有两三百个潜在客户，根本找不到未被发掘的客户。

Because maybe it's a very tight TAM where there's only two hundred-three 100 and target there's no undiscovered customers to call.

也许这取决于销售周期或其他因素，根据我的观察——见过大量公司在这个市场中进出——这纯粹是一种承诺。

Maybe it's depending on the sales cycle or whatever it is, there's dynamics that quite often my observation, having seen a lot of companies come in and out of this market, purely a promise.

那种‘轻松成交’的假设——即能带来客户——结果并不总是成立。

The easy sale, which is that will make customers happen, turns out to not always be true.

而这些客户流失的可能性很高。

And those customers have a high propensity to churn.

你知道吗，这确实是真的。

You know what, it is true.

我只想补充一点细节，因为我们确实已经向Artisan和Qualify输送了数百万个商机，因为这是我们最先使用的两个平台。

I will just add one nuance just just to this, because we have literally sent and I'm not an investor either, we have literally sent millions and millions of business to Artisan and Qualify because those are the first two ones we use.

这并不是因为我是个投资者，我并没有在为任何东西做推广。

Was not because I'm an investor, I'm not shilling anything.

我真的从中得不到任何好处。

I really don't get any benefit out of it.

他们只是在早期帮了我们。

They were just the ones that helped us in the early days.

所以我选了他们，因为他们帮了我们。

That's why I picked them, they helped us.

明白吗？

Okay?

但我可以告诉你，从这些人的经验来看，看看未来的发展，他们拒绝了我们发送的大部分线索。

But I can tell you from these guys, just seeing where the future is, they turn away most of the leads we send them.

他们拒绝了这些线索。

They turn them away.

所以之前那种感觉，哦，我们会……我并不是说所有的部署都完美无缺。

And so that feeling before of, oh, we're going to And I'm not saying that all the deployments are perfect.

我不是说你找不到不满意的客户。

I'm not saying you won't find unhappy customers.

别结婚。

Don't get married.

但我可以量化地告诉你，他们拒绝了我认为非常好的潜在客户，作为创始人，我肯定会毫不犹豫地签下这些客户，因为他们知道自己的数据不足、网站流量不够、分析工具不完善、CRM系统不够丰富，所以他们拒绝了这些客户。

But I can tell you quantitatively is they turn away leads that I think are quite good, that I certainly would have closed as a founder in a heartbeat because they know they don't have enough data, there's not enough web traffic, there's not enough analytics, the CRM isn't rich enough, and they turn them away.

他们没有足够的数据来服务这些客户。

They don't have the data to service them.

他们并不想要那100美元或50美元。

They They don't want the $100 or the $50.

这不值得，而且他们不仅不想获得这笔收入，也不愿意浪费前端工程师和客户导入的资源。

It's not worth And not only do they not want the revenue, they don't want to waste the FDE and the onboarding resources.

这才是真正的问题。

This is the real thing.

如果只是纯软件，他们或许会做，但这里涉及人力——你所能拥有的前端工程师数量是有限的。

If it were pure software, they might do it, but there's a human You only can have so many FDEs, so many forward deployed engineers.

所以你不想把资源投入到一次失败的部署上。

So you don't want to put it on a failed deployment.

但我真的在领英上收到私信，说：‘杰森，他们不回我电话。’

But literally I get DMs on LinkedIn, Hey, Jason, they won't return my call.

我该怎么办？

What should I do?

我说：‘他们确实回了。’

I'm like, Well, they did.

他们已经把你筛选掉了。

They qualified you out.

抱歉，但他们不希望失败。

Sorry, but they don't want it to fail.

这只是一个有趣的演变，对吧？

It's just an interesting evolution, right?

这确实是一种演变。

It is an evolution.

因为，没错，你说得对。

Because, yes, you're right.

第一代会让魔法无处不在。

The first generation will make magic happen everywhere.

我们在自己的公司里也看到了这一点。

We've seen that in our companies too.

第二代我们必须明确自己能在哪些地方创造价值，并且把这些事情做到极致，不要承诺什么AI魔法尘埃，否则只会带来收入增长和客户流失。

The second generation is we have to be very clear on where we can add value and do that really well, And don't promise AI magic pixie dust because you'll just end up getting revenue and getting churn.

所以顺着这个思路，你提到一个观点，与一些从零开始重构CRM的公司不同，你提到的那两家公司更像是艺术家和工匠。

So kind of pushing on this thread, you made a comment, unlike some of the reengineer CRM from the ground up, the two you cited qualified an art an artisan Yep.

工匠、MindRegi以及许多其他公司都构建在Salesforce之上。

Artisan, MindRegi, and a bunch of others run on top of Salesforce.

那么，你认为哪种模式才是正确的呢？

So which of those plays do you think is the right one?

是从底层重新设计CRM，就像我认为Day和Adio正在做的那样，构建整个技术栈，还是在Salesforce之上运行，做好智能代理部分，但依赖它们提供核心CRM功能？

Reengineer CRM from the bottom up, like I think Day is doing, like Adio is doing, you know, build a whole stack or sit on top of Salesforce, do the agentic part well, but rely on them for the day for the core CRM functionality.

哪种才是正确的选择？

Which is the right bet?

我实际上采访了Podium的创始人Eric Rare，这是一家基于Excel的公司，他昨天表示，不可能在CRM之上构建智能代理层。

I actually interviewed the founder of Podium, Eric Rare, which is an Excel backed company, and he said yesterday that there is no way that you can make the agentic layer on top of a CRM work.

你必须掌控整个技术栈，而这正是他们所做的。

You have to own the full stack, and that's what they do.

他们说的是对的。

They because true.

听好了。

Listen.

我很喜欢Eric，但这是他在推销自己的观点。

I love Eric, but that's talking his game.

这明显是错误的。

That's that's patently wrong.

我们已经售出，而且我们将会以两个人的速度卖出一千万个运行在Salesforce之上的智能代理。

We've sold We're gonna sell 10 Even a little faster with two people is gonna sell 10,000,000 running agents on top of Salesforce.

Eric非常聪明。

Eric is very smart.

我很喜欢他，但这是在替你自己的立场说话，我真想把鼠标扔到屏幕上。

I do like him a lot, but that's a talking your book, whatever comment that I want to throw my mouse at the screen.

这根本一点都不对。

It's not even remotely true.

在AI时代，吹嘘自己的那一套几乎是危险的。

Talking your game is almost dangerous in the age of AI.

每个人都在胡说八道。

Everyone's talking out of their ass.

你为什么不告诉我们，你们公司里什么行不通？

Why don't you instead tell us what doesn't work at your company?

我认为，就我们录制这段对话时的有限经验来看，你越垂直、越面向中小企业，这种说法可能越成立。

I do think my limited experience today as we record this, the more vertical you are and the more SMB you are, the more that might be true.

好吧，所以Podium应该还是面向中小企业的，对吧？

Okay, so Podium I think is still SMB, right?

它的业务非常垂直。

It's very verticalized.

除非你是Shopify，否则要构建一个基于你自身业务的庞大第三方代理平台是非常困难的。

It's very hard unless you're Shopify to build a massive platform of third party agents on top of what you're doing.

而对于Salesforce这样面向企业的产品来说，吸引这类人才却非常容易。

And it's very easy for Salesforce, which is very enterprise, to attract that talent.

我认为这正是答案，杰森，因为你确实如此。

I think that's actually the answer, Jason, because you are.

你并不是在Salesforce之上运行并仅仅充当代理层，而是像SaaS时代那样，要做到这一点，你需要在进入时拥有干净的数据，或者有能力承担清理数据、梳理数据结构的成本。

It's not that you run on top of Salesforce and just be the agentic layer, but just as in SaaS era, to do that, you need either the data to be clean when you come in or you need to be able to afford the cost of cleaning the data, figuring out the data structure.

如果你在500万美元的Salesforce实例之上做20万美元的交易，这一切都行得通。

If you're doing a $200,000 deal on top of a $5,000,000 Salesforce instance, that all works.

但如果你在数据混乱的小型Salesforce实例之上销售10美元的代理服务，你就根本没有预算去做这件事。

If you're selling a $10 agent on top of messy data structure from a small Salesforce instance, you just don't have the budget to do it.

所以我认为你说得完全正确。

So I think you're exactly right.

在中小型企业阶段，你大概会直接购买集成好的解决方案。

At the SMB stage, you probably will buy the integrated thing.

但我确信，未来很长一段时间内，仍会存在一个非常有吸引力的商业模式，即依赖CRM Salesforce基础设施来销售智能代理。

But I do believe there will be a very compelling business selling agents relying on the CRM Salesforce infrastructure for a long time to come.

我认为公开市场只是恐慌了。

I think the public market just panicked.

但当我看到Shopify以15倍收入估值交易，而Klaviyo只卖5倍，且Klaviyo的增长速度同样迅猛时，我不认为市场有这么精细的判断。

But when I look at Shopify trading at 15 times revenue and Klaviyo selling at five and Klaviyo growing just as quickly, I don't think markets are this nuanced.

但我确实好奇，他们最终是否在考虑Eric Ray的观点，即Shopify将掌控一切。

But I do wonder if what they're thinking at the end of time is going to Eric Ray's point is that Shopify will just own everything.

它和它的代理将彻底摧毁、吞噬整个生态系统，因为对于中小企业来说，这些代理必须包揽一切。

It and its agents will just destroy, eat its entire ecosystem because for SMBs, the agents that just have to do everything.

第三方代理根本没有生存空间。

There isn't any room for third party agents.

我不知道市场是否如此明智，但或许正是基于这种想法，Shopify将吞噬其整个合作伙伴生态系统。

Like, I don't know if the markets are this smart, but it may underpin it that Shopify is just going to eat its entire partner ecosystem.

我认为有

I think there's

这有一个利润空间的理由。

an margin argument for it.

下跌了25%，但基于这种信念，Shopify的股价应该上涨才对。

Down 25% and surely Shopify should be up on that belief.

但它相比同行的表现还是更好。

Well, it's still doing better than its peer set.

Klaviyo下跌了38%，对吧？

Klaviyo's down 38%, right?

听我说，我觉得Shopify确实是被过度抛售了，对吧？

Listen, I do think Shopify is the oversold one, right?

你知道，15倍的ARR仍然算是

You know, 15 times ARR is still a Is it still

15倍ARR？

15 x ARR?

我觉得确实是。

I think that it is.

我可能弄错了。

I might have it wrong.

不。

No.

120亿美元的年收入，1720亿美元的市值。

A 12,000,000,000 run rate, 172,000,000,000 market cap.

好的。

Okay.

而且这还是金融科技的毛利润率。

And that's with FinTech gross margins.

并不是80%的毛利润率。

That's not with 80% gross margins.

对吧？

Right?

它只是太贵了，但有一段时间，Klaviyo和Shopify的交易价格几乎一样。

It's just super expensive, but for a while they traded almost the same Klaviyo and Shopify.

这很合理，因为Klaviyo的收入有80%来自Shopify，但它的毛利率更高，增长也更快。

It made sense because 80% of Klaviyo's revenue is off Shopify, but it's higher gross margin and growing as faster quickly.

你可以说应该打个折扣，但这种情况很有趣。

You could argue for a discount, but it was interesting.

然后它们开始分化了。

Then they started to diverge.

现在我认为市场认为Shopify将吸收一切。

Now I think the markets have said Shopify is going to absorb everything.

我不认为Salesforce能做到，但我确实觉得存在一条界限。

I don't think Salesforce will, but I do think that is there's a line.

为中小企业构建第三方代理还有空间吗？

Is there any room to build third party agents for SMBs?

也许没有。

Maybe not.

也许你必须构建整个平台。

Maybe you have to build the whole platform.

可能吧，至少今天是这样。

Probably true, at least today.

HubSpot触底了吗？还是会继续下跌？

Has HubSpot hit the floor, or is that further to fall?

听好了。

Look.

我们是HubSpot的早期投资者。

We were early investors in HubSpot.

我非常喜欢那些人。

I love those guys.

即使Brian已经不在了，我也依然喜欢他。

I love Brian even though he's not there.

我非常喜欢Damesh。

I love Damesh.

我不了解新团队。

I don't know the new team.

我不是董事会成员。

I was not the board member.

我一直觉得他在技术上非常聪明。

I've always thought he's super smart technically.

所以，如果他们能完成需要完成的事情，他们应该没问题，因为分销仍然有优势。

So if they get done what they need to get done, they should be fine because distribution still has an advantage.

所以，我不会再说我不会了，我不会扫大家的兴。

So I'm not gonna say I'm not gonna again, I'm not gonna rain on anyone's parade.

我尤其不会去扫那些曾慷慨地赚了数亿美元的人的兴。

I'm particularly not gonna rain on the parade of someone who was kind enough to make as many hundreds of millions of dollars.

对吧？

Right?

他们得到了我的支持，或者 whatever 那个支持。

They have my vote or whatever it was.

对吧？

Right?

我只是觉得杰森说得对。

I just think Jason's right.

最终，你必须把事情搞定。

In the end, a, you have to just get it done.

你得拨开迷雾，进入新产品领域。

You have to cut through the noise and get to the new product universe.

然后，B，成效会体现在收入上，事实会说话。

And then b, the proof would be in the pudding and it'll show up in revenue.

如果没做到，那也没办法。我记得多年前，一位CEO在上任第一周开公司会议时，有人问了其他事情。

And if it's not, there's no know, I remember years ago, a CEO doing a company meeting in the kind of his first week when he took over, and someone asked about something else.

他说，好消息是我相信平衡计分卡。

And he said, know, the good news is I believe in a balanced scorecard.

坏消息是，收入增长率占了平衡的95%。

The bad news is revenue growth rate is 95% of the balance.

现金占剩下的部分。

Cash is the rest.

所以最终，这些公司有没有机会成功并变得重要？

So in the end, is the opportunity there for all these companies to pull it off and become relevant?

是的，我相信有。

Yeah, I believe so.

我不认为任何中小型企业客户会说：我不会从一家有二十年历史的公司购买。

I don't believe anyone in an SMB end customer says, I won't buy from a twenty year old company.

我必须从一家新公司购买，但你就是得把事情做成。

I need to buy from a new one, but you just got to get it done.

事情就是这样的。

That's what it was.

这是我的观点。

Here's my theory.

而且，我认为市场并没有我们想象的那么广阔。

And again, I don't think the markets are as deep as we might be.

对吧？

Right?

但并不是只有HubSpot承受了所有领导者中最大的压力，对吧？

But it's not just HubSpot has been under the most pressure of any of the leaders, right?

但Monday也承受了巨大的压力，而且它的增长速度要快得多。

But Monday has been under a ton of pressure too, and it's growing much more quickly.

它并没有达到峰值增长率，但仍然保持着30%以上的增长，市值为收入的五倍。

It's not growing peak rates, it's still growing over 30%, trading at five times revenue.

这是我的观点，而且我觉得这有点过度分析了，因为我不认为市场会这样发展，但SMB领域的难点在于，你必须增加用户席位才能增加收入。

And here's my theory, and again, think it's overthinking it because I don't think the markets go this way, but the tough part about SMB is, man, you got to grow seats to grow revenue.

就这么简单。

It's that simple.

对于HubSpot来说，即使拥有100%的净留存率，你也无法靠提价来掩盖问题。

For HubSpot to grow with 100% NRR, you cannot hide in price increases.

所以，如果你认为席位数量正在承压——而且已经有大量证据表明它确实在承压——我接触的越多的人，即使是企业级B2B玩家，也都会说：每次续约时，他们只想要去年90%的席位，对吧？

And so if you believe seats are under pressure, and there's a lot of evidence that already is under pressure, the more folks I talk to, even in enterprise B2B players, they're like, every time I go to a renewal, they only want 90% of the seats they had last year, right?

如果你相信这种趋势正在加速——数据也表明确实如此——那么SMB公司将受到最严重的冲击。

If you believe that is accelerating, which the data suggests, then the SMB guys are going to get the hardest hit.

HubSpot和Monday将受到最严重的冲击。

The Hubspots and the Mondays will be hit the hardest.

对吧？

Right?

你说得对。

You're right.

我完全同意你的观点。

But I totally agree you.

但有趣的是，我们的话题跳来跳去。

But it's just interesting that we jump around.

我们提到的三件事是，我们认为SMB市场无法支撑一个独立的产品，因为智能代理必须与CRM系统集成。

The three things we said are we don't believe SMB can support a separate product in the sense that the agent needs to be integrated with the CRM.

你提到的第二点是，我们有充分证据表明智能代理能带来巨大价值，比如旧金山的SMB所有者杰森·伦库姆，他的AI代理已经赚了数百万美元。

And second thing you said was, we totally have proof that agents can deliver huge value because Jason Lenkum, SMB owner in San Francisco, is making millions of dollars from his AI agent.

因此，任何拥有数万乃至数十万SMB客户的现有SaaS公司，都应该赶紧行动起来，开发一个智能代理，去交付这种价值。

Therefore, an existing SaaS company that has tens and hundreds of thousands of SMB customers should just get its ass in gear, make an agent, and go deliver that value.

我相信我刚才提到的那些公司的首席执行官们也这么认为。

I believe the CEOs of those companies I just named believe that too.

我相信这些产品目前还不够好。

I believe the products just aren't good enough yet.

你无法否认Agencik的产品尚未带来收入这一事实。

You can't argue with the fact that the Agencik products have not inflicted revenue.

说到底，证据就在收入上。

Like, the proof is in revenue.

我同意。

I agree.

接下来我想说的是，顺便提一下，我之所以现在即兴发挥，是因为这根本不在议程上。

And then the next thing I wanna say just and by the way, the reason I'm making this up as I go along is this was just not on the agenda.

我刚参加完一次董事会会议。

So I just came off a board meeting.

我对此毫无准备，但我们继续吧。

I had no preparation on this, but we'll keep going.

我们之前有一个说法，有人很久以前提过：在每一个这样的市场中，都是一场竞赛——一方是拥有渠道的现有企业，需要增加产品；另一方是拥有产品的新兴企业，需要建立渠道。

The soundbite that we had, someone gave it a while ago, which is every one of these markets, it's a race between the incumbents who have distribution needing to add product, and the new guys who have product need to add distribution.

如果你想想看，我拿Monday、HubSpot，然后再说Salesforce。

And if you think about I'm gonna take Monday, HubSpot, and then say Salesforce.

对吧？

Right?

如果你认为现有企业是拥有渠道的一方，那么问题就归结为这种优势能持续多久。

If you think of if the incumbent is the guy who has the distribution, it boils down to how long does that last.

现有产品越黏性越高，你就越有时间弥补差距，这就是为什么我认为许多会计类产品会有很长的时间来推进。

The stickier the existing product, the more time you have to cover the gaps, which is why I think many of the accounting products will get a long time to push on it.

像Monday这样的产品，从某种程度上说属于项目管理工具，让我觉得，那些不属于系统记录、最终不汇总到会计系统的工作任务，越有可能被颠覆。尽管Salesforce不是会计系统，但它却是大多数公司利润表的核心信息来源，因为合同都存在那里并汇总起来。

Something like a Monday, which is past management at some level, strike me as the kind of thing where the more of the knowledge worker task y that's not system of record, that's not ultimately rolling up to an accounting system of record, because even though Salesforce is not an accounting system, it is the core information for most companies' P and L because it's where the contracts sit and roll up.

你越不像一个系统记录，而越像一个工作系统，就越容易被颠覆，也越没有时间把事情搞定。

The less you are like that, a system of record, and the more you are some kind of system of work, the more likely you are to be disrupted, and the less time you have to get shit done.

我认为这些产品可能没那么有优势。

I think those products are probably a lot less.

而正如你所说，中小企业本身就比企业级市场更难形成粘性。

And then obviously SMB is just inherently less sticky, to your point, than enterprise.

但同样，正如你所说，Shopify 尽管面向中小企业，但其自然流失率是你无法控制的。

But again, to your point, Shopify, even though it's SMB, the natural churn there's nothing you can do about it.

但只要你作为中小企业还活着，你大概率不会轻易更换 Shopify，因为不清楚 AI 版的 Shopify 能给你带来什么优势。

But provided you're still alive as an SMB, you're probably going to be slow to change your Shopify because it's not clear what an AI Shopify would give you.

而且它对你所做的事情至关重要。

And it's pretty core to what you do.

这就是我们又回到了同一个问题。

That's why we get back to the same thing.

这里很可能把婴儿和洗澡水一起倒掉了，这意味着应该存在交易机会，但什么时候呢？

It's probably much babies being thrown out with the bat wolf here, which means there should be trades to do, but when?

另一件有趣的事情是，你刚才也看到一些评论提到这一点。

The other thing that's interesting, and you saw some of the commentary here, is this.

这也是事情变得困难的地方。

This is where it also gets hard.

当公司从以收入倍数估值、不扣除亏损、完全忽略期权稀释，转变为完全以自由现金流估值，并且开始扣除 SPC 时，这种转变巨大到可以说是一次彻底的估值重置。

When companies shift from being valued on revenue multiples, no deduction for loss, totally ignoring option dilution, to all the way to being valued on free cash flow and where they're counting on, where they're basically deducting the SPC, it's just such a huge change that it takes for I mean, it's such a valuation reset.

如果你进入这种估值重置，市场突然不再把你视为一家增长型公司，而是开始将你与银行、公用事业或工业公司作比较，那么你需要经历多年股价持平、稳定增长，才能达到10到15倍自由现金流的估值水平。

If you go into that valuation reset and the market suddenly stops thinking of it as a growth company and starts thinking of, let me compare you to a bank or a utility or an industrial company, it takes a lot of years of flat stock price and reasonable growth before you can be worth 10 or 15 times free cash flow.

而这里正在发生的就是这种情况。

And that's what's happening here.

在这些公司的自由现金流倍数（扣除稀释，而非SBC）达到合理水平之前，你不会看到底部。

You don't see a bottom until these things are at free cash flow multiples net of dilution, not SBC, but dilution.

当这种情况发生时，那就是你的底部。

And when that happens, that's your bottom.

真是个糟糕的时期。

What a shitty time.

这对萨提亚来说也很糟糕。

It's pretty shitty for Satya too.

这对比尔来说也很糟糕。

Pretty shitty for Bill too.

但我们还是继续谈萨提亚吧，因为我们不谈政治。

But let's stay on Satya because we're not going into politics.

老比尔现在正在澳大利亚忙活呢。

Good old Bill, he's busy right now in Australia.

顺便说一下，那件事有很多方面，但绝对不是政治。

For the record, that is many things, but it's not politics.

我们还是回到萨提亚身上。

Let's get back to Satya.

无论如何，微软经历了历史上第二大单日市值蒸发。

Anyway, Microsoft's second largest market cap loss ever.

操，我是微软的买家。

Fuck, I'm a buyer of Microsoft.

我手里有大量微软股票。

I have a shitload of Microsoft.

你能帮我理解一下吗？为什么一天之内市值蒸发了3600亿美元，问题到底出在哪？

Can you please help me understand why with 360,000,000,000 loss in market cap in a single day, what is the problem?

为什么这么低迷？

Why is this so depressed and down?

我的意思是，你看看这些数据，说清楚一点，业绩达标了，Azure增长可能只差了1%，37对38，都在误差范围内。

I mean, you you look at it and you go, let's be clear, made the numbers, maybe missed Azure growth by 1%, 37 versus 38, all within a margin of error.

我觉得这是几个因素的综合结果。

I think it's a combination of a couple of things.

一是，正如你所说，对你们大量的未来RPO（未确认收入）存在一些抵触情绪。

One is, as you say, some pushback on much of your future RPO, which is future revenue unrecognized.

比如，其中50%来自OpenAI。

Like, percent, 50% of it was from OpenAI.

突然间，人们开始质疑这些收入是否真的能变成真金白银。

And suddenly people are questioning whether that's going to turn into money.

如果这真的是主要担忧，那么当OpenAI融资一千亿美元时，按理说这部分估值应该会有所回升。

And if that really is the concern, then when OpenAI raises $100,000,000,000 logically some of that jump should come back.

但我认为更广泛的原因是，他们再次强调了杰森的观点，即推理和GPU是赚钱的，但他们把更多GPU资源分配给了内部产品研发，因此在Azure上能出售的GPU就变少了。

But I think the wider again, they also reiterated the Jason point, which is inference or GPU is money, and they allocated more of their GPUs to internal product development and therefore were able to sell less of them in Azure.

他们明确表示，如果我们把更多GPU用于Azure，就能实现那额外的1%增长，那样我们就‘没问题’了。

And they basically explicitly said, if we'd use more of our GPUs in Azure, we could have made that extra 1% of growth, so we'd have, quote unquote, been fine.

在董事会会议上听到这种说法，我真喜欢。

Love it when I hear that at a board meeting.

如果不是东海岸的风暴，我们本来会没问题的。

If it wasn't for the storms on the Eastern Seaboard, we would have been fine.

不是因为那1%的问题。

It wasn't for the 1%.

从根本上说，你看到的是，我们之前就跟你说过，在AI领域，微软的企业发展团队执行得非常出色。

Very fundamentally, what you're seeing is we said this for you, in the world of AI, they haven't as we said, the corporate development team at Microsoft have executed brilliantly.

他们拥有OpenAI三分之一的股份。

They own one third of OpenAI.

但微软的产品团队并没有出色地执行。

The product team at Microsoft have not exited brilliantly.

他们没有任何自己拥有的、令人信服的AI产品，无论是像谷歌那样的大语言模型，还是令人眼前一亮的应用。

They don't have any compelling AI products that they own, either an LLM, which Google has, or even compelling apps.

他们就是没有。

They just don't have it.

所以之前有一点风声说，他们因为是OpenAI的供应商并向其销售Azure，而获得了一些 perceived 的增长。

So a little bit of the buzz was they were getting some perceived lift because they were a vendor to OpenAI and selling them Azure.

但现在市场对这种说法已经冷淡了，因为他们认为，第一，这可能本来就是一个低利润的业务。

But now the market is kind of soured on that because they're saying, A, maybe it's a low margin business anyway.

我更愿意拥有模型，而不是提供算力。

It's not I prefer to own the model than provide the compute.

第二，如果你向OpenAI销售，这真的是好生意吗？

And then B, if you sell into OpenAI, is that really money good?

所以我认为，当市场转向时，就会出现这种情况——存在一种叙事和势头，按有效市场假说本不该存在，但确实存在，因为我们是人。

So I think it's kind of what happens in these markets when they turn against there's an element of narrative and momentum that you think shouldn't be there in the efficient market hypothesis, but it is there because we're human.

这是我学到的一件事。

That's one of the things I've learned.

我曾经是个彻底的有效市场假说支持者。

I used to be a total efficient market hypothesis guy.

我认为从长远来看我确实是，但在短期内，叙事决定一切。

I think in the long run I am, but in the short one narrative shapes everything.

过去两年里，微软的叙事一直非常强势。

And there's been two years of Microsoft narrative being really strong.

他们都拥有OpenAI。

They all have OpenAI.

他们表现得非常出色。

They're killing it.

我们，quote-unquote，记住，我们要让谷歌跳舞。

We're, unquote, remember, we're gonna make Google dance.

还记得吗？

Remember that?

自从他这么说之后，你看过思维图了吗？

Have you seen the thought chart since he said that?

这是一种负相关。

It's an inverse correlation.

负面的舞蹈。

Negative dancing.

我认为发生的是，叙事发生了转变，曾经被视为优势的东西突然被认识到是弱点。

And I think what's happened is the narrative shifts, and suddenly the things that were perceived as strengths are realized as weaknesses.

叙事转变确实有其道理，因为他们缺乏一个引人入胜的情节。

There's truth to the shift of the narrative, which is they don't have this compelling plot.

然后有一天，你错过了1%的Azure增长，所有人都惊呼：天哪。

And then one day you miss Azure growth by 1%, and everyone goes, oh my god.

叙事一变，他们就被淘汰了。

Narrative change, and out they go.

我认为这就是这里发生的事情。

I think that's what happened here.

我能理解一下吗？

Can I just understand?

他们之前是不是定价过高，现在才被正确地定价了？

Were were they overpriced previously and they're now correctly priced?

有意思。

Interesting.

在所有这些巨头中，除了显然以天文倍数交易的特斯拉，其他七家都大致处于23倍左右的前瞻市盈率区间。

Of all the mega caps, the bag seven with the exception of Tesla, obviously, which trades at an astronomical multiple, all are kind of in that 23 plus or minus forward PE.

曾经有一段时间，谷歌的估值相对于这个水平有相当的折扣，如果你那时买入，现在几乎已经涨了100%。

And there was a period where Google was at a compelling discount to that, and if you bought, you're up almost 100%.

而现在，它们的估值都差不多了。

And now they're all much of a muchness.

所以，如果人工智能持续吞噬一切，那么如果没有在这一领域推出相关产品，就越来越难打造一个有吸引力的软件业务。

So if AI keeps eating everything, it gets harder and harder to have a compelling software business without having relevant products in that space.

最终，这也是微软不得不付出的代价。

And in the end, that's a tax that Microsoft is going to have to pay too.

所以，我正在努力回答你的问题。

So I'm trying to answer your question.

它们目前的定价可能还算合理，但真正的关键是，它们必须在知识工作者层面、Azure层面和模型层面，打造出真正契合人工智能时代的系列产品，否则十年后，它们的估值就难以维持在合理水平。

They're probably appropriately priced for now, but the real truth is they have to get their act together as a set of products that are relevant in the AI world at the knowledge worker level and at the Azure level, at the model level, if they're going to be at the right price ten years from now.

如果你是萨提亚，从产品和模型层面你会怎么做？

If you're Satya, what do you do from a product and a model level?

你会在模型层面收购某家公司吗？

Do you buy someone at the model level?

你有没有去收购Cohere或Mister Al，试图在可收购的领域有所作为？

Did you go and buy Cohere or Mister Al and try and have a play there in something that's buyable?

我会很挑剔。

I'd be bitchy.

他们不是收购了Inflection AI吗？

Didn't they buy Inflection AI?

我不认为那完全是冲着模型去的。

I don't think that was exactly for the models.

没看清楚。

Didn't look out.

好吧。

Okay.

你看，奇怪的是，他们持有OpenAI 30%的股份，是除了员工信托之外最大的单一投资者。

Look, the the odd thing is they have 30% of OpenAI, which is the largest single investor other than the the employee trust.

但非常有趣的是，尽管如此，我看了数据，他们现在每年给Anthropic的收入高达数十亿美元。

But it's really interesting, despite that, I saw the numbers, they're giving literally billions a year in revenue to Entropic now.

我觉得他们给Anthropic的收入超过五亿美元。

Think I 500,000,000 plus in revenue to Anthropic.

所以，我认为，从中长期来看，如果没有自己的模型，很难成为主要的算力玩家。

So it's hard to imagine over the medium term being a major compute player without having access to some kind of model yourself, I think, as I think about it.

所以，是的，他们可能需要解决这个问题。

So yes, they probably need to figure that out.

我的意思是，当你年收入达到3200亿美元时，这很难。

I mean, it's tough when you're at a $320,000,000,000 run rate.

你必须大干一场。

You gotta go big.

这可不是个简单的问题，对吧？

It's not a simple problem, right?

购买Cursor没有帮助。

Buying cursor doesn't help.

规模不够大。

It's not big enough.

那你买什么？

What do you buy?

对吧？

Right?

我的意思是，也许你可以收购一家模型提供商，如果数学计算能说得通的话。

I mean, maybe you do buy a model provider, if the math somehow get the math to work.

但你必须做得非常大。

But you got it's gotta be just huge.

在这个规模下，结果必须足够大才能产生影响。

The outcome has to be huge to move the needle at this scale.

必须非常大。

It has to be huge.

你也不必在芯片领域也参与竞争。

You don't also have to have a play in chips as well.

而且，认真说，当你现在把亚马逊和谷歌都纳入比较时，他们都有自己的布局，老实说，如果你想保持自己作为最有价值公司之一的地位，你就需要两者兼备。

And, seriously, when you compare to Amazon and Google now both having their play, honestly, if you wanna retain your status as one of the most valuable companies, you need both.

我不认为你需要覆盖整个领域。

I don't think you need to cover the board.

我不觉得这是个风险游戏，对吧？谁覆盖的领域最广谁就赢。

I don't think this is a game of risk, right, where, you know, the person who covers most of the board wins.

我认为，那些人做芯片的部分逻辑，其实是为了在采购上对英伟达施加一些影响力。

I think the logic for some of those folks doing chips was in part your defense of trying to provide some leverage on their purchases to NVIDIA.

没错，虽然微软确实有这个问题，但我想这样讲：你可以对在英伟达上的支出感到不满，但你更应该懊恼的是，你没有自己的大语言模型。

And, yeah, while Microsoft has that issue, maybe I'd argue this way: It's okay to begrudge your spend with NVIDIA, but you should begrudge more the fact that you don't have an LLM.

如果你仔细想想，如果你处在软件栈的上层，那么掌控下一层——也就是模型层——比去优化芯片更重要。

If you think about it, if you're up here in the software stack, it's more important to own one level down, which is that model layer, than to start to optimize around chips.

那会偏离核心目标。

That would be a diversion from the core thing.

总的来说，一个重要的观点是，过去微软曾占软件行业总利润的70%到80%，收入也约占60%。

The end, one of the big picture soundbites here is, time was, Microsoft literally made 70 to 80% of all the profits made in software, and they were probably 60% of the revenue.

杰森说得对。

Jason's right.

今年他们的收入达到了3200亿美元，但未来一两年内，软件领域可能出现两家公司，每家收入达到500亿或1000亿美元，比如OpenAI和Intropic。

They're doing 320,000,000,000 this year, but it's possible that in a year or two, there will be two one or maybe even two companies doing 50 or 100,000,000,000 in the software space, which is OpenAI and Intropic.

任由这种情况发生是不可取的。

And simply letting that happen is just not great.

现在你之所以被推动，是因为你持有OpenAI三分之一或30%的股份。

Now you get somewhere compelled because you own a third of OpenAI or 30% of OpenAI.

你曾经在软件行业占据主导地位，能够攫取全部利润。

You had effectively the big dog position on the entire software industry, and you could extract all the profits.

但现在要再做到这一点会越来越难，这将是一个重大失误。

And now it's just gonna get harder to do that, and that's a miss.

你提到了利润攫取的问题。

You said about extracting profits.